Scholarly analysis of the relative merits of presidential and parliamentary

democracy has emphasized the weaknesses of presidential systems owing to

the separation of power between the executive and assembly, which often
leads to conflict between the two. Shugart and Carey, by contrast, focus on
the potential strengths of presidential systems. They identify a number of
alternative ways of balancing constitutional powers between the branches,
demonstrating that some forms of presidentialism are more likely than
others to generate conflict between the branches.
They also evaluate the role of electoral rules in presidential systems,
demonstrating that the ways in which presidents and assemblies are elected
play a critical role in determining the shape of the party system, the politi­
cal agenda, and the prospects for cooperation and conflict between presi­
dents and assemblies.
A typology of presidential systems is developed, including two hybrid
forms defined by the authors as premier-presidentialism and president­
parliamentarianism, and these various forms are evaluated according to a
set of characteristics that determine the efficiency and effectiveness of
governance. These characteristics include the manner in which the presi­
dent is elected, the legislative powers of the president, the authority of the
president over the cabinet, the timing of presidential and assembly elec­
tions in relation to one another, and the system of representation.
PRESIDENTS AND ASSEMBLIES
PRESIDENTS AND
ASSEMBLIES
CONSTITUTIONAL DESIGN
AND ELECTORAL DYNAMICS

MATTHEWSOBERGSHUGART
and
JOHN M. CAREY

. . .�! CAMBRIDGE
;:: UNIVERSITY PRESS
PUBLISHED BY THE PRESS SYNDICATE OF THE UNIVERSITY OF CAMBRIDGE
The Pitt Building, Trumpington Street, Cambridge CB2 lRP, United Kingdom

CAMBRIDGE UNIVERSITY PRESS

The Edinburgh Building, Cambridge CB2 2RU, United Kingdom
40 West 20th Street, New York, NY 10011-4211, USA
10 Stamford Road, Oakleigh, Melbourne 3 I 66, Australia

© Cambridge University Press 1992

This book is in copyright. Subject to statutory exception

and to the provisions of relevant collective licensing agreements,
no reproduction of any part may take place without
the written permission of Cambridge University Press.

First published 1992

Reprinted 1995, 1997
Typeset in Times
A catalogue record for this book is available from the British Library
Library of Congress Cataloguing-in-Publication Data is available
ISBN 0-521-41962-X hardback
ISBN 0-521-42990-0 paperback

Transferred to digital printing 2004

 To Merry Shugart
and
Mary E. and R obert J. Carey
 Contents

List of tables and figures page viii

Acknowledgments x1

1 Basic choices in democratic regime types 1

2 Defining regimes with elected presidents 18
3 Criticisms of presidentialism and responses 28
4 The premier-presidential and president-parliamentary
experiences 55
5 The constitutional origins of chief executives 76
6 Constitutional limits on separate origin and survival 106
7 Legislative powers of presidents 131
8 Assessing the powers of the presidency 148
9 Electoral dynamics: efficiency and inefficiency 167
10 Electoral rules and the party system 206
11 Electoral cycles and the party system 226
12 Electoral cycles and compatibility between president and
assembly 259
13 Conclusions 273
Appendix A Electoral rules for one-seat districts and
coalition-building incentives 288
Appendix B Theoretical explanation for models predicting
number of parties in presidential systems 293

Bibliography 301
Index 313
 Tables and figures

TABLES
3. 1 Breakdowns of democratic regimes in the twentieth
century by regime type page 40
3.2 Countries that have held at least two democratic
elections without breakdown, as of 1991 41
5. 1 Opposition influence and cabinet durability in
parliamentary democracies 79
5.2 Reelection of presidents 88
6.1 Filling and dismissing cabinets 119
6.2 A schematization of presidential powers in the
appointment-dismissal game 122
8.1 Powers of popularly elected presidents 150
8.2 Powers of popularly elected presidents by country 155
8. 3 Frequency of democratic failure, grouped by regions in
Figure 8.1 157
9.1 Strength of party leadership over rank and file in
presidential democracies 176
9.2 Effective number of parties contesting congressional
elections (NJ and effective number of presidential
candidates (Ne), Chile, 1931-1973 180
10.1 Median percentage of votes for two highest-finishing
candidates in presidential elections, according to different
electoral methods 211
10.2 Median effective number of presidential candidates (Np)
and median effective number of parties obtaining votes in
congressional elections (NJ, according to different
institutional formats 220
11.1 Effective number of parties in parliamentary systems,
according to effective magnitude 228
11.2 Type of assembly party system and associated institutional
arrangements 229
11.3 Percentage of votes for parties supporting president during
each term, Chile, 1938-1973 245
 Tables and figures IX

11.4 Percentage of votes for three largest parties in concurrent

and honeymoon elections, Venezuela, 1978-1984 249
11.5 Percentage of votes for party or parties supporting
president in systems using majority runoff for president 250
11.6 Ratios of votes in honeymoon and midterm election to
votes before presidential election, president's own party
and whole coalition 252

FIGURES

2.1 A typology of democratic regimes in two dimensions 26

6.1 Hypothetical indifference points for senate (S) and
president (P): no overlap, thus position would remain
unfilled 108
6.2 Hypothetical indifference points for senate (S) and
president (P): president nominates candidate at point /5 108
6.3 Schematized outcomes of equilibria in appointment-
dismissal game, relative to ideal points of president (P) and
assembly (A) 123
7.1 Presidential (P) and congressional (C) ideal points on
hypothetical policy: indifference point for simple majority
Uc1) and two-thirds majority Uc21) 145
8.1 Powers of popularly elected presidents 156
8.2 Separate survival and presidential cabinet power: a
comprehensive typology of democratic regime types 160
9.1 Party strength and president's legislative powers 177
9.2 Effective number of parties contesting congressional
elections (NJ and effective number of presidential
candidates (Np ), Chile, 1931-1973 181
11.1 District magnitude and number of seat-winning parties,
Costa Rica, Dominican Republic, and Venezuela 234
11.2 District magnitude and effective number of electoral
parties, Costa Rica, Dominican Republic, and Venezuela 235
11.3 District magnitude and effective number of presidential
candidates, Costa Rica, Dominican Republic, and
Venezuela 236
11.4 Median ratio of votes for president's party and coalition in
honeymoon and midterm elections to votes for president's
party in last election before presidential election
(base = 1.00) 254
B.1 District magnitude and effective number of seat-winning
parties, Costa Rica, Dominican Republic, and Venezuela 297
B.2 District magnitude and percentage change in effective
number of parties, without vanity parties, 1978 and 1988 298
 Acknowledgments

The completion of this book was made simpler and more pleasant by the
generous help of colleagues who offered their criticisms of all or part of the
book. Shugart's involvement with questions of elections and electoral
systems in presidential democracies originated in his Ph.D. dissertation,
completed in 1988 at the University of California, Irvine. There, Rein
Taagepera was the model of what a graduate student's mentor should be.
His assistance and encouragement since the completion of the Ph.D. and
his comments on this project have been a constant source of inspiration.
Arend Lijphart, who also served on Shugart's graduate committee and
later on Carey's, has offered cogent analyses and criticisms throughout.
Many times we were compelled to rethink an argument as a result of
exchanges with him.
Michael Coppedge, Gary Cox, and Scott Mainwaring offered detailed
comments and criticisms that helped us sharpen many parts of the book.
We also thank Bernard Grofman, Juan Linz, and Mathew McCubbins for
comments and intellectual support.
Truly indispensable for both authors was the intellectual atmosphere
fostered at the University of California, San Diego, by the faculties and
graduate students in both the Graduate School of International Relations
and Pacific Studies (IR/PS) and the Department of Political Science. In
particular, the students of Shugart's seminar on Politics and Policy in Latin
America in 1990 and 1991 deserve thanks both for serving as "guinea pigs"
for many of the concepts developed in Chapters 8 and 9 and for proving
that one really does learn from one's students! Special recognition is due to
a few of them: Richard Rifkin, Michael Thies, and John Warren.
As with any project of this size, there were tasks that the authors were
only too happy to delegate to others. Harold Colson of the IR/PS Library
graciously handled many anxious requests for hard-to-get data and docu­
ments. We are indebted for excellent work in computer graphics, correc­
tions in the text, and bibliographic assistance to several research assis­
tants: Steve Bingham, Dan Nielson, Mary Reiter, Gail Sevrens, and Tony
Westerling.
 1
Basic choices in democratic regime types

The 1980s were a time of growth in a subfield of political science that has
come to be known as the "new institutionalism" (see March and Olsen
1984; Grofman 1987). Often drawing in a matter sometimes more, some­
times less formal, from microeconomic understandings of rational behavior
and individual responses to incentive structures, this subfield has placed
political, rather than social or economic or cultural, variables at the center
of explanation for political outcomes. There is a renewed focus on the
importance of political institutions in accounting for the success or failure
of democracy. Recent advances of democracy in Central and Eastern Eu­
rope and other parts of the globe have given impetus to the study of
designing constitutions and the consequences of institutional choice. Old,
long unchallenged assumptions about the efficacy of presidentialism in
Latin America have been seriously challenged in recent years. Mindful of
the spirit of resurgent interest in constitutional design, we have launched
this book as a treatise on how various institutional designs for representa­
tive democracies affect the ways in which the political process operates.
This book focuses on a set of regimes that, in our assessment, have
received too little attention from comparativists. It deals primarily with
systems in which there is an elected president. These regimes differ from
the common parliamentary type in that there are two agents of the elector­
ate: an assembly and a president. In a parliamentary system, there is only
one, the assembly, with the executive being an agent of the assembly, 1
rather than of the voters directly. As this book shows, there are myriad
ways to design constitutions that vary the relationship of the voters' two
agents to one another, as well as to the electorate. Regimes with elected
presidents vary in the ways in which the president may check, cajole,

1 For purposes of this book, we do not consider bicameralism to represent two agents of the
electorate. In contemporary democracies, both chambers of the assembly are to serve the
function of representation, although often the basis of representation differs according to
the chamber. In parliamentary systems, ordinarily only the lower house has the power of
confidence over the government, and in many cases the lower house may override objec­
tions of the upper house on legislation. Lijphart (1984) offers one of the most concise
discussions of bicameralism.
2 Presidents and assemblies
confront, or simply submit to the assembly majority. We even find that
some systems give the president so little power relative to the assembly that
they are effectively parliamentary. We thus do not see a presidential regime
as being the polar opposite of parliamentarism, as much of the literature
implies; but the development of this point awaits future chapters. While
this book is fundamentally about comparing systems with presidents, under­
taking such a study requires starting with the dichotomy that most of the
literature deals with: presidential versus parliamentary. These two forms
are not only the two most discussed in the literature, especially in the
polemics regarding the desirability of certain constitutional designs; they
are also the two original types of democratic constitution. It was only later
that many other regime types or hybrids emerged and gave us the broad
range of regime types that we compare in this book.
In a parliamentary system, the executive is chosen by and may be re­
moved by the elected assembly. In a presidential system, the process of
forming the executive is institutionally distinct from the process of filling
seats in the assembly, as both branches are popularly elected. In each of the
other types and hybrids that we assess in this book, there is some combina­
tion of presidential and assembly authority over the composition of the
executive. As we shall argue and make one of our principal themes, in
none of the types is it easy to divorce the process of executive formation
from the process of voters' choices for assembly representatives. Typically
in parliamentary systems, even though it is only the assembly that comes
before the electorate for votes, voters make their choices to a significant
degree on the basis of what kinds of policies they want addressed by the
executive. In presidential systems, too, there may be "coattails" from the
presidential races that affect the vote for the assembly. Where the voting
for the two branches is most clearly separated, as in some presidential
institutional designs, often the lack of policy agreement between the execu­
tive and the assembly majority causes stress in the regime. These are
among the principal trade-offs that we shall discuss as the book proceeds.

THE PRESIDENTIAL-PARLIAMENTARY DICHOTOMY

For us, the authors, a primary factor motivating the decision to undertake
this project was our observance of a sharp polemic on the subject of whether
presidential or parliamentary democracy is the "better" form of representa­
tive government. Most of the scholarly literature on the subject comes out
quite squarely behind parliamentarism as the preferred alternative. How­
ever, among practicing politicians, the message is getting through slowly, if
at all. Nearly all new democracies in the 1970s and 1980s, and in 1990, have
had elected presidents with varying degrees of political authority.
Among newer democracies, true parliamentarism remains largely a phe­
nomenon of the former British Empire, as in transitions to democracies in
 Basic choices in democratic regime types 3
the late 1980s and early 1990s in Pakistan and Nepal. Yet even in the British
Commonwealth the appeal of presidentialism among elites has led to the
replacement of parliamentary systems with presidential regimes in Guy­
ana, Nigeria, and Zimbabwe, among others.
After the collapse of the Soviet bloc in Central and Eastern Europe, only
in Czechoslovakia (and, very briefly, East Germany, before it united with
West Germany) did the political elite initially opt for a pure parliamentary
system with no popularly elected president. Encouraging for academic par­
tisans of parliamentary democracy, the Hungarian electorate refused to
endorse a powerful elected presidency in a referendum.
For the supporters of parliamentarism, the bad news is that no existing
presidential system has ever changed to a parliamentary system, while
several have made the reverse move. As of this writing, there was at least
serious discussion of adopting regime types other than presidential in some
Latin American countries, notably Brazil, where, as in Hungary, voters
were expected to be given an opportunity to express themselves on the
topic.
W hat we have, then, is a considerable divergence between a virtual
academic consensus, on the one hand, and actual political practice, on the
other. In the remainder of this introductory chapter, we seek to outline the
very basic choices involved in the choice of either the presidential or the
parliamentary form as well as sketch briefly the historical origins of these
regime types. In so doing we shall be introducing two of the basic themes of
the chapters that follow: (1) Especially in regimes with elected presidents,
there are many institutional choices that matter besides the choice of what
means to employ for the constitution of executive power; and (2) that, even
where powers are formally separate, the processes of election to the two
branches interact in ways that affect the functioning of democracy.

Assemblies and executives: constituting political power

A major dilemma in democratic regimes concerns a divergence between
what representative assemblies do best and what executives must do if
democracy itself is to function well. Assemblies, or at least lower houses of
assemblies, are intended to be representative of the population. 2 A typical
2 The common generic term "legislature" is often misleading. In many parliamentary re-
gimes, most actual lawmaking takes place within the cabinet. Similarly, as we shall see in
later chapters, the executive in presidential systems also has important legislative powers.
more in some regimes than in others. Hence, instead of the term "legislature," we shall use
the term "assembly" throughout this book. We believe this term more accurately reflects a
common feature of democratic regimes: the existence of an institution in which elected
representatives assemble for whatever constitutional function they may be granted. In cases
in which that function is to sustain confidence in a government, we shall use the term
"parliament." For presidential systems, in which executive power is separated from the
assembly, we shall use the term "congress." We shall eschew the term "legislature" entirely.
4 Presidents and assemblies
democratic assembly is elected for the purpose of giving voice to the inter­
ests of localities or to the diversity of ideological or other partisan divisions
in the polity and society. That is, assemblies are ordinarily expected to be
parochial in nature. Executives, on the other hand, are charged with acting
to address policy questions that affect the broader interests of the nation,
as well as to articulate national goals.

Geographic representation. Historically, representative assemblies emerged

for the purpose of offering a link between the central authority and smaller
geographic subdivisions of the state. Typically a locality would elect one,
two, or slightly more members of the assembly by means of some variant of
plurality or majority rule.
The definition of the basis of representation would change over time in
some systems, particularly those that adopted proportional representation
(PR). For the time being, however, we shall concern ourselves only with
the historical period in which the two basic regime types emerged, and we
emphasize that both did so in a context of assemblies constructed to repre­
sent localities.

Parliamentary confidence. Over a period of many decades and through a

process that will not directly concern us here, the principle of cabinet
responsibility to parliament developed in Britain. In Chapter 9 we review
this process in more detail. The important point here is that what was
originally the Crown's ministry (and in strictly formal terms remains so),
became a ministry subject to the confidence of a majority of the assembly.
Thus developed what would come to be known as parliamentary democ­
racy, although it must not be forgotten that this system was not originally
"engineered." Rather, it emerged gradually, and to this day no specific
written constitution has been drafted in Britain. Modern parliamentarism
was born as a political system in which an assembly constructed to repre­
sent localities coexisted with a ministry that gradually shifted from the
control of the monarch to the control of the assembly itself. We shall return
below to the consequences that parliamentarism holds for the nature of
elections for and representation in the assembly. For now, let us turn to that
other means of constituting executive power that emerged early in the
development of democracy, presidentialism.

Popular election of the executive. W here and how did presidentialism

emerge? The answer takes us back to the U.S. Constitutional Convention,
which was the first instance of either type of regime's having been con­
sciously "engineered." The Framers opted for neither the monarchical
form of government that the Revolution had been made against nor the
system of the short-lived Articles of Confederation, under which there was
 Basic choices in democratic regime types 5
no national executive. As possible methods of formation of the executive,
they considered an executive elected by the Congress and one elected by
direct popular vote before settling on the mechanism of the electoral col­
lege. They rejected one chosen by Congress on the grounds that corruption
would prevail as the executive would owe its very existence in office to
another arm of the government.
While the Framers considered an executive chosen by the Congress, they
did not contemplate an executive responsible to the representative assem­
bly. Indeed, such a regime type, which we would now know as parlia­
mentarism, had yet to exist. In Britain, the cabinet was still the responsibil­
ity of the monarch, whose authority, of course, did not rest upon any
connection, direct or indirect, with the electorate. Once both selection of
the executive by the assembly and a monarchy had been rejected, the
Framers were in effect replicating the essentials of a form of government
that then existed in Britain, an executive with powers separated from the
powers of the elected assembly. The crucial difference is that rather quickly
the electoral college evolved from being a mere filter for popular prefer­
ences into an automatic expression of them. Thus developed the first demo­
cratic system in which there are two agents of the electorate.
As voters effectively abandoned the intended practice of simply voting
for a delegate who would be entrusted with making the decision of who
should be president in favor of making it themselves (by way of electors
"pledged" to a particular candidate), the U.S. presidential system took on a
new character. By Andrew Jackson's era, the president would be the expres­
sion of a popular "mandate," having a special link to the whole national
constituency. It is this character of presidentialism that has brought down
so much criticism on the regime type, as we shall see. It is ironic that, while
the U.S. presidential system was a deliberate response to a problem of how
to create a national executive when there was neither a monarch nor any
preexisting models besides monarchy, the modern "presidentialism" of a
combative partisan political figure often battling the separately legitimated
assembly was an unforeseen development.
A second round of creation of presidential systems came with the inde­
pendence of the Latin American republics. Several of these countries
opted for direct popular election of the president, while in others, at least
initially, the executive was chosen by the assembly for a fixed term. As in
the early U.S. and British democracies, the Latin American systems had
assemblies designed to be the expressions of local interests. In some cases,
regionalism was by far the most important basis of the polity. We develop in
Chapter 9 a theoretical explanation for why a variant of presidentialism
with unusually strong presidential powers and weak parties emerged in
these settings of preeminent localism.
As in the United States, in Latin America the fixed-term executive was
6 Presidents and assemblies
the only available model,3 yet some early constitutional designers saw their
innovations as emulating the British model, only with the "monarch" popu­
larly legitimated. Thus, while the "British model" would later come to refer
to parliamentarism, as late as the 1820s when Latin American indepen­
dence was under way, this model indeed referred to a cabinet responsible
only to the head of state and an assembly elected from geographically
defined districts. That presidentialism as we know it emerged primarily
because there were no other obvious models is borne out by looking into
the history of European institutional development.

Further "engineering": parliamentarism and new regime

types in Europe
As in Britain, in all of the other states on the European continent that
evolved into constitutional monarchies in the nineteenth century, authority
over the cabinet shifted from the hands of the monarch to the parliament.
Unlike the cases of the new countries of the New World born of rebellion
against monarchies, in Europe the emergence of constitutional regimes did
not see the establishment of elected heads of state, since the monarch
would serve that role. Even France, which for a while oscillated among
republican democracy, dictatorship, and monarchical restoration, parlia­
mentary democracy with no elected president was established under the
Third Republic in 1871. Not until Charles de Gaulle's plebiscite under the
Fifth Republic in 1962 did popular election of the president become en­
trenched.4 By then additional alternative regime types had emerged, so
France retained cabinet responsibility to the parliament even while adopt­
ing a popularly elected presidency.
t,. second wave of democratization in Europe led to regime types other
than parliamentarism or presidentialism. These regimes emerged after the
democratization of Britain, meaning the basic parliamentary feature of
cabinet responsibility was a recognized model of constitutional design.
However, in the absence of a monarch, the head of state is popularly
elected, thus creating a second agent of the electorate. This combination of
an elected presidency with a cabinet dependent on the assembly's confi­
dence thus multiplied the number of permutations on constitutional design
that we discuss in this book. Some of these regimes constitute a regime type
we shall define as premier-presidential. Such a regime, sometimes called
3 The exception here is Brazil, which originally inherited the monarchy that had been over-
thrown in Portugal. Later on, the Brazilian monarchy was replaced by a presidential system.
4 Initially, the constitution of the Fifth Republic was simply a "reinforced" variant of parlia­
mentarism. In contrast to the Fourth Republic, the terms under which cabinets could be
replaced were changed to favor greater cabinet durability, and the president (still to be
elected by an electoral college made up of other elected officials) was granted some emer­
gency provisions.
 Basic choices in democratic regime types 7
"semipresidential," ensures that the cabinet depends only on assembly
confidence, yet provides certain powers to a popularly elected president. A
more detailed definition follows in Chapter 2; in particular, we shall see
that many regimes have "confused" control over cabinets, by granting both
of the electorate's agents considerable authority in matters of government
formation. For now, all we are concerned with is the historical fact that
Europe's "second wave" of democratization often took elements of both
the British concept of government responsible to parliament and the New
World concept of a separately elected president.
This phase of democratization begins with Finland's independence from
Russia in 1905 and accelerates after the defeat of the Austro-Hungarian
and German empires in World War I and the independence of the remain­
ing dependencies of Western Europe. 5 The nature of presidentialism on the
U.S. and Latin American models as defaults chosen in the lack of other
models is further underscored when we realize that, while many of these
"second wave" states' leaders saw as desirable an elected head of state to
replace the monarch, they nonetheless made the cabinet responsible to the
parliament. Finland, Weimar Germany, Austria, Ireland, and Iceland all
adopted such systems, although they varied widely in the amount of power
given to their presidents.
Thus the initial phase of democratization on the European continent was
set apart from that of the New World by the existence of monarchies. What
set it apart from the process in Britain were two aspects: the prior existence
of the "model" of parliamentary confidence, and, later on, the decision to
adopt proportional representation for the assembly. In these systems, there­
fore, a conscious decision was made that the assembly should be elected on
the basis of ideological or sectoral, rather than geographical, differences
within the society. The later democracies in Europe followed the earlier
ones in adopting proportional representation and parliamentary cabinets,
but some also adopted elected presidencies.

ASSEM B L I E S A N D E X ECUTIVES : I N T E RACTION A N D

E L ECTORAL CONSEQUENCES

A frequent theme of this book is that the means chosen to constitute

executive power affect the nature of elections and representation in the
assembly as well. In particular the interaction and its consequences mani­
fest themselves as trade�offs between efficiency and representativeness.
"Efficiency" refers to the ability of elections to serve as a means for voters
to identify and choose among the competing government options available
5 The correlation is not perfect, as some of the new (and short-lived) democracies of Eastern
Europe in the early twentieth century chose to have their presidents elected by parliament.
One, Estonia, had no head of state, thus being arguably the "purest" of all parliamentary
systems.
8 Presidents and assemblies
to them. "Representativeness" refers to the ability of elections to articulate
and provide voice in the assembly for diverse interests. We argue that
parliamentarism virtually requires that a choice be made; either there will
be an efficient choice of policy options to be addressed by the executive or
there will be a broadly representative assembly, or a compromise must be
struck. With only one agent of the electorate, it is not feasible to have both
efficient and representative elections in the same system. Having an
elected president, on the other hand, at least holds out the promise of
serving both ends through elections. The ability of a presidential regime to
do so and how it handles the inherent tensions of doing so are the themes of
most of the rest of this book. Before looking at how parliamentarism and
presidentialism affect efficiency and representativeness, let us elaborate on
these terms.

Efficiency. It was Walter Bagehot who described the British system as

having both "efficient" and "dignified" components. A major message of
his famous essay, The English Constitution, was to call attention to a then
largely unacknowledged transformation that had occurred over time in
Britain. The cabinet no longer was under the sway of the monarch; indeed,
the portions of the (unwritten) English constitution in which the monarch
now had a role were merely ceremonial, hence the term "dignified." The
cabinet had fallen under the sway, instead, of the parliament. The cabinet
had become the "connecting link" between the "fused" powers of the
legislative and the executive authorities. This linkage and its importance in
the actual workings of British government made the institution of parlia­
mentary confidence the "efficient" component.
Growing out of this fusion of powers was another form of efficiency and
it is in this sense, discussed by Cox (1987) in The Efficient Secret, that we
use the term in this book. Because governments were made and unmade
according to that composition of the majority of parliament, a voters'
choice of parliamentary candidate was also a choice of executive. Elections
thus came to turn on the voters' preferences for government and, by exten­
sion, on policy rather than on more purely local concerns (such as "pork"
and patronage). As a result, neither elections nor the legislative process
turned on distribution of particularistic goods by means of logrolling across
districts. Instead elections offered an "efficient" choice from among com­
peting policy options, and legislation is the domain of the majority party
and its cabinet.
An inefficient system, on the other hand, would be one in which elec­
tions turn more on the provision of particularistic services to constituents
than on policy. Such a legislative process is also likely to be inefficient, as
members of congress seek to be able to bring back to their districts proj­
ects, jobs, or other services that will help them be reelected.
Efficiency, when centered on elections to the executive, is closely re-
 Basic choices in democratic regime types 9
lated to identifiability - the ability of voters to identify the choices of
competing potential governments that are being presented to them in
electoral campaigns. As developed in more detail in Chapter 2, concepts
such as efficiency and identifiability relate to models of what Powell
(1989) calls "Citizen Control" that provide for governmental accountabil­
ity and policy mandates. In a highly efficient system like that of the
United Kingdom, identifiability is high. The more inefficient an election
is, the lower will be its identifiability.

Representativeness. Electoral systems may be scaled by the degree to which

they represent diversity. That diversity may be expressed in geographic
terms as the parochial interests of localities, or it may be expressed sec­
torally as the narrow interests of groups within the society, whether or not
they are geographically concentrated.
If representation takes place on a strictly local basis, it may be associated
with a highly inefficient electoral and legislative process, but it need not be
so. Local representation can also mean the ability to articulate policy con­
cerns that are specific to a region, as, for example, in polities with a geo­
graphically concentrated ethnic minority, as in Spain, or significant regional
variations in economic standing, as in Canada. Locally defined representa­
tion may even be a combination of particularistic services, such as casework
and pork barrel, and policy interests of special concern locally, as in the
United States.
The articulation of local interests depends on both the electoral system
and the constitutional design. Such representation is likely to be maximized
under a system of small districts such that each member of the assembly will
hail from a clearly defined region. Much of the discussion that follows will
consider the relationship between local representation and whether a sys­
tem is presidential or parliamentary.
If representation is to be based upon group interests, such as clashing
ideologies or distinct ethnic or religious segments, it need not be at all local
in nature. It must, however, be a form of proportional representation (PR)
based on party lists. 6 Indeed, the extreme of such an electoral system
would elect all members in a single nationwide district, as is the case in
Israel and the Netherlands. Many other countries, including Sweden and
Denmark, elect a major part of their assemblies in nationwide districts.
Some that have rather small geographic districts nonetheless have a second
tier designed to ensure faithfully proportional representation of diverse
parties, as in Austria and Germany and, in their first democratic elections,
Bulgaria and Hungary.

6 Some PR systems also allow (or require) candidate voting, which often weakens the party­
centered nature of elections and may even make elections tum on local particularism. For
discussion of one such system, see Ames (1987).
10 Presidents and assemblies

Parliamentarism
As cabinet responsibility to parliament came to be the basis for constituting
executive power in Britain, what happened to the geographic representa­
tion in the assembly? And how did the adoption of proportional representa­
tion in continental European parliamentary regimes affect the institutions
of cabinet responsibility? Efficiency is best served when there are two
principal options from which voters may choose, so that one of the options
will be certain to hold executive office. Thus efficiency in a parliamentary
system would be best served by a low-magnitude electoral system, where
magnitude refers to the number of seats elected per district. The reason
that low magnitude is tied to efficiency under parliamentarism is that as
magnitude is decreased, so ordinarily is the number of parties, while the
probability of a "manufactured majority" for a party with less than a major­
ity of seats is increased (Rae 1967; Taagepera and Shugart 1989). Since
efficiency as it relates to government formation implies a categoric choice
among two contenders, one of whom will gain control over the govern­
ment, a small number of parties and a majority of seats for one of them are
required to maximize efficiency. On the other hand, as district magnitude is
decreased, so is the opportunity for the political institutions to articulate
minority viewpoints.

Efficiency. Owing to the workings of the plurality rule in single-seat dis­

tricts, one of the British parties usually obtains a majority of parliamentary
seats and thus forms the government on its own. Cabinet responsibility led
to the emergence of disciplined, programmatic parties. The development
of the party system around the institutions of parliamentarism and geo­
graphically defined representation led to the basis of election of members
ceasing to be primarily local concerns and instead becoming national issues
(Cox 1987).
Parties must differentiate themselves on the basis of national policy.
Politics in the districts from which members of parliament are elected can
no longer focus so much as in the absence of party discipline on the stuff of
legislative "logrolling" and compromise - the ability to deliver services to
the constituency. Instead it must focus much more on nationally divisive
but often not divisible concerns. W hile this focus on national issues in
elections even within locally defined districts is the source of the system's
efficiency and is justifiably much admired, the system contains potential
problems.
Two principal criticisms of parliamentarism on the British model thus
emerge. First, an assembly formally constructed to represent local interests
is compromised in that function and instead becomes principally an "elec­
toral college" for determining which party holds executive power. It thus
emerges as neither a legislature, as legislative authority is concentrated in
 Basic choices in democratic regime types 11
the cabinet, nor very representative, at least on the level at which its
members are chosen, since the national policy concerns that are expressed
by parties capable of winning national power become paramount.
Second, and stemming from the first, where elections are contested al­
most entirely on the basis of which of two major parties shall receive
unchallenged executive power for a prescribed period, electoral contests
may become excessively polarized, perhaps raising the stakes of elections
so much as to destabilize the system. The collapse of British-style parlia­
mentary regimes in countries as diverse otherwise as Grenada, Nigeria, and
Pakistan is an example of potential pitfalls of parliamentary systems. 7

Representativeness. The other model of parliamentarism is more inclusive

of broad interests and multiple parties because it employs proportional
representation. The extent to which the system will be representative of
diversity depends on district magnitude. Previous studies (Rae 1967 ;
Taagepera and Shugart 1989; Lijphart 1990) have shown a compelling rela­
tionship between district magnitude (M) and the number of parties: the
greater M is, the more "relevant" parties there are. 8 Since under PR, multi­
seat districts must be used, in any district the number of possible winners
will be greater than in M = l systems such as Britain. The result is that,
especially as M gets to be very high, representation is defined less by
geography and more by ideological or sectoral interests in a multiparty
system.
If the dilemma of parliamentarism on the British model is the sharp dimi­
nution of local representation, under multiparty parliamentarism there is a
different dilemma. W ith multipartism, coalition or minority governments
are likely to prevail. Such a system is thus more representative than the
British model, since divergent (parochial) concerns of sectors, classes, or
other segments of society may be present at once in the assembly and even in
the cabinet. Such constitutional designs are what Powell (1989) called "repre­
sentational." However, maximizing representativeness through PR almost
ensures minimizing cabinet durability, since parliamentary confidence
means the cabinet must assuage the interests of diverse parties. In such
systems, then, the degree to which voters may assess accountability and
confer mandates through elections is limited (Powell 1989). Thus, while this
form of parliamentarism enhances representativeness and consequently the

7 Even in the United Kingdom, that most venerable of parliamentary systems, the recent
decline in the certainty with which cabinets resign upon losing votes that are not explicitly
votes of confidence or else major spending bills indicates some dissatisfaction with the strict
linkage between MPs' votes and the fate of the national executive. Continuing calls for PR are
also a manifestation of this dilemma, as PR and the coalitions that would ensue would enable
parties to represent more faithfully sectoral, if not necessarily locally based, constituencies.
8 Further along in the book we shall define terms more precisely, such as "effective magni­
tude" and "effective number" of parties.
12 Presidents and assemblies
role of the assembly, it limits the possibility that the executive can articulate
national policy goals transcending parochial partisan interests. Examples of
parliamentary systems that have functioned in this way are the French
Fourth Republic, and the current regimes of Belgium and Denmark. In the
1980s, Israel's difficulties in forming governments that had to appease the
demands of numerous small parties also manifest the same problem.

Balancing efficiency and representativeness. In fairness to those who advo­

cate parliamentary democracy, few today encourage the adoption of either
the British model or high-magnitude PR for emerging democracies. Some,
such as Linz (1987), specifically refer to PR designed so as not to encourage
too fragmented a party system and perhaps with a "constructive vote of no
confidence," by which cabinets cannot be replaced by purely negative coali­
tions but only by a majority coalition in favor of a new alternative cabinet.
Of course, the dilemma of introducing such mechanisms is that they tend
to reduce the number of parties and thus the representativeness of the
system, compared to one with higher M and lower barriers to changing
cabinets. Thus the attempt to balance efficiency and representativeness,
like walking a tightrope, is precarious. Indeed, Powell (1989), who consid­
ered mostly parliamentary systems, noted that his scores on elements of
accountable and decisive (hence efficient) governance were weakly related
to his scores on representativeness. Whichever is the formal basis on which
the assembly is constituted, geographical representation in Iow-M districts
or ideological representation in higher-M districts, parliamentarism leads
to a fundamental conclusion: It is difficult to maximize at once the func­
tions of the two branches. Either the choice of government becomes so
important in elections that the more parochial interests of localities or
sectors matter little in elections and representation, or else cabinet durabil­
ity becomes imperiled by an overly representative assembly. The reason
that this choice is forced is that only one branch comes directly before the
voters for "democratic legitimation" (Linz 1987): the assembly.

Systems with popularly elected presidents

If parliamentarism makes difficult the maximization of the fulfillment of
both executive and assembly roles, what about systems with two agents of
the electorate? At first it would seem that systems with presidents would
perform extremely well because the choice just described - pursuit of more
parochial interests versus supporting the executive - need not be con­
fronted at the level of electoral campaigns. The separate election and
possibly independent constitution of executive and assembly powers do not
have to compel the essential trade-off between conducting elections on the
basis of efficiency or conducting them on the basis of representativeness.
 Basic choices in democratic regime types 13
Efficiency. In a presidential election, the choice of executive power can be
very efficiently presented to voters, provided that the president has real
executive powers. If only one of the candidates contesting elections can win
the right to fill executive positions or designate who shall perform this
function, the ability of voters to identify and choose among options for
executive office is quite well achieved when there is an elected president.
There are many variations on the presidential role in these matters, but for
now the basic point is simply that elections for president tend to be rela­
tively efficient for executive formation to the extent that the executive does
not depend entirely upon postelection negotiations, as in multiparty parlia­
mentary systems.

Representativeness. An efficient choice for executive under parliamentar­

ism necessarily renders assembly elections less representative than they
might be, since the electoral process for executive and assembly is fused.
W ith a presidency, however, the assembly election need not present voters
the same narrow range of choices that confronts voters in the executive
election. Or, looked at from the opposite angle, even a high-M propor­
tional representation system for the assembly need not compromise the
efficiency of executive election or render basic cabinet composition depen­
dent on nonelectoral shifts in party alignments. Thus it is theoretically
possible to have representational elections that give voters a wide range of
party choices without compromising efficiency in executive formation.

Balancing assembly and executive power. Having noted that, in theory,

systems with popularly elected presidents do not require the compromising
of either branch's electoral connection does not, of course, mean it all
works well in practice. Indeed, a reader initiated in the debates about
presidential and parliamentary regimes may be surprised to see us pointing
to such features as the different composition of assembly and executive
powers as a strength. The very benefits we have suggested have also been
identified as the regime type's basic flaws, as discussed in Chapter 3.
In setting up the discussion in this way, we are calling attention to the
existence of regimes in which the electorate has two agents: one that re­
sponds to the desirability of efficiency in executive formation, and the other
that responds to the desirability of giving voters choice of parties and policy
programs. How these two goals are reached and the degree to which they are
achieved - or in less successful designs, the potential for conflict - depends
on a number of variations that are discussed in this book. We thus undertake
the comparative study of presidencies and their interaction with assemblies
from a perspective of "political engineering": the concept that constitutional
design matters for the way in which it shapes politicians' pursuit of their
interests. Before proceeding with detailed definitions and analysis, we
14 Presidents and assemblies
should consider in more detail what we mean by political engineering. In
doing so we take as our starting point The Federalist Papers.

CONSTITUTIONAL DES I G N AND POLITICAL I NTERESTS :

B RI N G I NG THE FEDERALISTS BACK IN

This book fits within the school known as the "new institutionalism" because
it is founded on the premise that politicians' "preferences can be understood
only in the context of the institutionally generated incentives and institution­
ally available options that structure choice" (Grofman 1989:1). The new
institutionalism draws upon analytic tools from microeconomics, game
theory, and social choice as a way to understand how the design of institu­
tions conditions political outcomes. It is appropriate, then, that we begin
with The Federalist Papers, for the men who drafted and then campaigned
for the ratification of the U.S. Constitution well understood that political
outcomes could be shaped by the "fine details" of institutions (Schwartz
1989:38).
In seeking to understand better those constitutions that are founded
upon the notion that there should be two agents through which the elector­
ate exercises control over the polity, we are drawn into political engineer­
ing (Sartori 1968). We agree with Grofman (1989:4) that the Founding
Fathers were "the greatest (if not the first) political engineers." In stating
our appreciation for the authors of The Federalist Papers, however, we are
not implying that the Founders had the last word on constitutional design.
Rather, we are saying that they provided an essential starting point, as
relevant today as in the late eighteenth century, for understanding how
institutions shape politics.
Any study that is rooted in the principles of rationality, as is this one, and
that employs such tools as principal-agent relations and spatial modeling
follows upon the same concerns as The Federalist Papers: how the rules of
the game can be devised so as to structure incentives in such a way that
leads to good government. 9 Madison's underlying assumption is that
elected officials will be influenced by the self-interested motive of reelec­
tion (Cain and Jones 1989:22). We share that assumption. It was in that
vein that the Founders sought to pit "ambition against ambition," in their
famous phrase, in order to thwart tyranny. In The Federalist Papers, the
Founders develop one of the most profound defenses of the promise of
political engineering ever written. We would not be undertaking this pres­
ent study were we not also convinced of the promise of institutional engi­
neering. W hile we do not proceed mainly with the aim of prescribing
"optimal" constitutional designs, we do recognize some as preferable to
others. W here we have a clear preference, we shall say so explicitly. Our
9 Federalist No. 51 defines "Justice" as "the end of government."
 Basic choices in democratic regime types 15
main purp ose, h owever, is to c onsider the implicati ons of the myriad ways
in which the p owers of presidents and assemblies have been c onstituted in a
wide variety of c ountries.
Besides their c ontributi on t o a theory of c onstitutional design and ra­
ti onal ch oice, The Federalist Papers are als o useful f or our purp oses be­
cause they devel oped an elab orate c onstituti onal design based on the
concept of there being tw o agents of the elect orate: a president and an
assembly. Or rather, they set the stage for the devel opment of constitu­
ti ons with tw o such agents, even if the pr oduct of the C onstituti onal
C onventi on c omprised the ideal of separate agency c onsiderably. As
Chappel and Keech (1989:47) observe, Federalist N o. 51 argued that each
department of g overnment was meant to be c onstituted such that "the
members of each sh ould have as little agency as p ossible in the app oint­
ment of the members of the others." A rigor ous adherence t o this princi­
ple w ould require that all t op officeh olders "be drawn from the same
f ountain of auth ority, the pe ople." Obvi ously, the original elect oral col­
lege, which was n ot dependent on p opular elections, was one of several
ways in which the original d ocument c ompr omised the ideal of separate
agency; but over time the principle came t o be m ore fully realized.
All of the c onstituti ons c onsidered in detail in this b o ok similarly pr ovide
for tw o agents of the elect orate. They vary greatly, h owever, in the extent
to which there are als o agency relati ons among the elect orate's agents, f or
example by the inv olvement of the assembly in app ointing or dismissing
members of the executive. M ore over, as The Federalist Papers p oint out,
even a maximum degree of separati on in the c onstituti on of auth ority in the
executive and assembly d oes n ot preclude overlap in their functi ons. Here,
t o o, as we shall see, there is en ormous r o om f or variati on. These variati ons
are the subject of the next chapters.

PLAN OF THE BOOK

Chapter 2 lays out specific definiti ons of three types of regime that c onsti­
tute our primary basis of c omparis on. We define presidentialism as a re­
gime type based on the ideal of maximum separati on of p owers and full and
exclusive resp onsibility of the cabinet t o the president. We define premier­
presidentialism as a type in which the president has certain significant
p owers, but the cabinet is resp onsible only t o the assembly. The third type
is president-parliamentary, a c omm on type with shared - or confused -
resp onsibility over cabinets between president and assembly.
In Chapter 3 we review the criticism of presidentialism as c ontrasted with
parliamentarism. We expl ore several advantages of presidential over parlia­
mentary dem ocracy and then c onsider h ow premier-presidentialism might
offset s ome of the criticisms launched against presidential systems while
still retaining the advantages of having tw o agents of the elect orate.
16 Presidents and assemblies
Chapter 4 provides some case studies of premier-presidential and presi­
dent-parliamentary regimes. We see that the more successful regimes
have ensured that the president's powers did not so dominate that the
assembly majority could not maintain a cabinet when president and as­
sembly differ. The president-parliamentary regimes have frequently been
conflict-ridden, as we shall see.
Chapters 5-8 are studies in the comparative presidency. First, in Chapter
5, we explore the means of constituting chief executive power. We look at
cases in which the assembly is involved in choosing the chief executive and
consider when such involvement vitiates the ideal of two agents of the elector­
ate and when it does not. The chapter also considers vice-presidencies versus
interim elections as means to replace chief executives who cannot fill out
their fixed terms. Finally, we discuss a little-known but potentially viable
variation on presidentialism: the collegial executive.
Chapter 6 discusses several ways in which powers can be less than sepa­
rate. Some regimes leave it only to the president to fill and dismiss members
of the cabinet, while in others the assembly plays a role in appointments,
dismissal, or both. We offer a simple heuristic, based in spatial modeling,
that predicts the equilibrium points at which cabinets will rest, depending on
the relative symmetry of appointment and dismissal powers. The chapter
also considers the powers of dissolution held by some presidents.
In Chapter 7 we look at the legislative powers of presidents, exploring
variation on veto and override procedures as well as decree powers. We
argue that students of presidential systems have frequently overstated the
importance of decree legislation. We distinguish among different forms of
decrees and offer a spatial model of conditions under which assemblies are
likely to delegate decree powers and with what consequences.
In Chapter 8 we develop a system of scoring presidential powers, with a
table of the powers of presidents in forty-four constitutions. We also pro­
vide two graphs of the regimes. One graph shows presidential power in two
dimensions: legislative and nonlegislative. We find that the most powerful
presidencies also have been the most problematic regimes. The second
graph locates the regimes according to degree of presidential authority
over cabinets and the degree of separation of powers. This graph allows us
to see at a glance the most important characteristics of a regime's constitu­
tional design and to what extent each conforms to the types defined in
Chapter 2.
In the four subsequent chapters we focus on elections, parties, and elec­
toral rules. Chapter 9 develops theoretical constructs of inefficient and
efficient presidentialism, based on the relation between presidential legisla­
tive powers and the form of party organization, as shaped by the electoral
rules.
Chapter 10 looks at empirical relations between various features of insti­
tutional design and the number of parties. Chapter 11 offers a theoretical
 Basic choices in democratic regime types 17
model for why we observe two-party systems in some cases (despite PR)
and multipartism in others. These chapters show the importance of the
electoral cycle - the timing of presidential and assembly elections in rela­
tion to one another - on party system outcomes.
Chapter 12 takes up the question of designing regimes, focusing on the
role of the electoral cycle in shaping the likelihood of compatibility be­
tween president and assembly majority. The chapter considers conditions
under which such compatibility is desirable as well as when it may be
desirable not to have compatibility. We discuss how different constellations
of variables can be created to maximize particular political goals. Finally,
Chapter 13 offers general conclusions.
 2
Defining regimes with elected presidents

Any thorough analysis of presidential systems must first establish clear

criteria for what presidentialism is. In this case, the task is especially critical
because we shall later examine a number of variations on the institutional
arrangements of presidential government, as well as proposals for modifica­
tions as yet untested.
The tasks of this chapter, then, are as follows: First, we provide a defini­
tion of presidentialism, drawing on the theoretical heritage of The Federal­
ist and contrasting ours with other institutional definitions of the regime
type. Next, we define premier-presidentialism in contrast to pure presiden­
tialism. 1 We then discuss some hybrid regime types that exhibit combina­
tions of qualities from both the presidential and premier-presidential types.
Finally, we introduce a simple typology of regimes based upon two dimen­
sions: (1) the degree of separation of powers, and (2) the nature of the
cabinet. We return to this typology in greater detail in Chapter 8.

P R E S I D E N T I A L I S M A C C O R D I N G TO I T S FO U N D E R S :
SEPARATE ORIGIN AND SURVIVAL

Beginning with The Federalist, the central defining characteristic of presi­

dentialism has been the separation of legislative from executive powers.
Indeed, this theme predates presidential government. The authors of The
Federalist based much of their faith in the separation of powers on Montes­
quieu's arguments for legislative control over a king's ministers. Later, the
idea of presidentialism as the separation of powers in a republican govern­
ment is clearly delineated first in The Federalist, and this idea will remain
central to our definition of presidentialism.
Nevertheless, by itself, the phrase "separation of powers" does not tell us
much. Even in the United States, legislative and executive powers are not
entirely separate. The presidential veto represents an executive intrusion in
the legislative process; and the requirement of Senate ratification of trea-
1 Throughout this study, the term "presidential" regime refers to the "pure" type defined in
this chapter; we do not use the term to encompass premier-presidentialism or various
hybrids.
 Defining regimes with elected presidents 19
ties allows for legislative influence in the executive domain of foreign pol­
icy. Indeed, without such overlap, it would be difficult to imagine how one
branch could check actions of the other. Along with the separation of
powers, checks and balances were central to the conception of presidential
government elaborated in The Federalist. 2
If we were to define presidentialism as the absolute separation of execu­
tive from legislative power, then any veto requiring an extraordinary congres­
sional majority for override would constitute a deviation from presiden­
tialism. We argue that this is not the case. Powers are never entirely separate
under presidentialism - nor were they intended to be. In fact, Madison's
arguments in The Federalist suggest that the rationale for separating the
sources of the origin and the survival of the executive and Congress was to
ensure that each branch could impose checks on the other without fear of
jeopardizing its own existence. Thus, separation in some respects serves to
ensure interdependence - that is, checks - in others.
The critical distinction is between provisions pertaining to the origin and
survival of the two branches and provisions pertaining to their actions.
Regarding origin and survival, maximum separation is characteristic of
presidentialism. But regarding actions, presidentialism seeks to protect
mutual checks, which in turn requires that powers overlap considerably.

C O N T E M PO R A R Y D E F I N I T I O N S : I N S T I T U T I O N A L C R I T E R I A

The definition we use throughout this book when we refer to presidential

government (or, at times, "pure" presidentialism), is the following:
1 . the popular election of the chief executive;
2. the terms of the chief executive and assembly are fixed, and are not contingent
on mutual confidence ; and
3. the elected executive names and directs the composition of the government.
This three-part definition captures the essence of separate origin and
survival of government (executive) and assembly in addition to specifying
that the president be elected by voters (or an electoral college chosen by
them for that sole purpose). We also identify a fourth criterion that, we
maintain, logically follows from the above:
4. the president has some constitutionally granted lawmaking authority.
The latter criterion does not concern the formation and maintenance of
powers, but if presidents do not have lawmaking power (such as a veto),
then they are chief executives only in the most literal way: they execute
laws the creation of which they had no way of influencing. Granting presi­
dents lawmaking powers is a way of ensuring that the popular endorsement
of a policy program through a presidential election can be translated into
2 See especially Federalist No. 5 1 .
20 Presidents and assemblies
actual policy output. There is only one regime that is fully presidential by
the first three criteria yet fails to meet the fourth: Venezuela.
Having provided a working definition of presidentialism, we now review
briefly some other contemporary definitions, in order to highlight ways in
which our definition differs. In 1959, Douglas Verney offered a list of
eleven criteria by which to distinguish presidentialism from parliamentar­
ism, as well as from what he called "assembly government. "3 But Verney's
definition suffers from ambiguity and overlap among his criteria, and from
the fact that he does not establish the primacy of any of them. In his book
Democracies, Lijphart (1984:66-71) argues that two criteria are essential
to presidentialism, while a host of others (some of which belong to Verney's
list) are associated with - but not necessary for - presidential government.
Lijphart's two essential criteria are that the chief executive shall:
1. not be dependent on legislative confidence, but rather shall sit for a fixed term;
and
2. be elected by popular vote. 4
In a recent essay, Lijphart (1989) identifies a third criterion, which he
regards as essential to presidentialism as well:
3. a one-person executive.
3 Verney's conception of assembly (or convention) government is much more like parlia­
mentarism than presidentialism. In effect, it is parliamentarism in which the assembly.
rather than the cabinet, is dominant. His distinction between the two results primarily from
the fact that he regarded the Westminster model as the prototype for parliamentary govern­
ment, and was therefore forced to define another category for parliamentary forms that
vary significantly from this model. We shall propose a somewhat different definition of the
term "assembly regime" below. Verney's eleven criteria overlap considerably, so we have
distilled Verney's institutional characteristics of presidentialism down to six:
1. the head of state is also the head of government;
2. the president appoints the cabinet;
3. legislative and executive personnel are distinct;
4. the executive is not dependent on confidence of the assembly;
5. the president cannot dissolve the assembly; and
6. the assembly is the supreme branch of government. (Verney 1959: 39-57, 75-7)
4 Students of democratic institutions may object that the formal election of the president by an
electoral college (as in the United States, Argentina, and Finland) violates the principal of
popular election for the president. Nevertheless, the election of members of an electoral
college, who fulfill no other role than casting votes for president, is a vastly different phenome­
non from the election of assembly members who also elect a head of government. It would be
a mistake to equate the indirect election of a president, via an electoral college . with the
selection of a head of government by the assembly. For this reason, we reemphasize the
distinction between popular election and direct election for president. In most presidential
systems, presidents are chosen by direct, popular election . Where an electoral college is
chosen for the sole purpose of naming the president, without conferring or negotiating
compromise candidates, then we still regard the election as popular, but not direct. In
parliamentary and premier-presidential systems, where a prime minister must be approved
by the assembly, the election of the head of government is neither popular nor direct.
 Defining regimes with elected presidents 21
Another way of saying this is that the electoral district magnitude - the
number of seats elected in a district (Rae 1967) - for the executive equals
one (M = 1), where the district is the entire nation. Lijphart's primary
criticism of presidentialism is that it is inherently majoritarian, providing
poor representation for minorities. W here M = 1, electoral dispropor­
tionality is bound to be high unless one candidate wins virtually all the
votes. To define presidentialism by virtue of district magnitude, however,
and then to argue that presidentialism generates majoritarianism, obscures
the issue that majoritarianism is generated by electoral formula, and does
not add much explanatory power to the definition of presidentialism. It
also requires the classification as wholly distinct regime types systems such
as Uruguay (1952-67) and Cyprus (1960-3), which have had popularly
elected collegial executives while retaining separation of powers and
checks and balances. If a regime has these latter features, we prefer to
classify it as presidential, regardless of whether or not its elected executive
is unitary or collegial.
In addition, to define presidentialism by the most majoritarian character­
istic of most presidential systems seems inherently to contradict the spirit of
the first two criteria described by Lijphart. Both of these criteria emphasize
the relative independence of the legislative and executive branches. This
independence is, as we shall see, essential to protecting the mutual checks
in presidential systems - a distinctly antimajoritarian tendency. Thus, aside
from the empirical problem of collegial executives, to define presiden­
tialism by its unitary executive seems to be theoretically contradictory.
It could be that the contradiction over the number of persons constitut­
ing the elected executive is less a problem of definition than an indication
of internal inconsistency in the regime type itself. Indeed, judging from the
tensions between Madisonian and Hamiltonian conceptions of democracy
evident in papers of The Federalist, this point is well taken. 5 We shall
explore in the chapters ahead many of the institutional tensions and per­
haps contradictions in a wide array of systems generally classified as presi­
dential. For the purposes of definition, however, we seek consistency in
addition to generality, and we have chosen to embrace the Madisonian
conception of presidentialism over the Hamiltonian.6 Madison's logic, we
believe, generally prevailed over Hamilton's in shaping the first presiden­
tial constitution, and better reflects the themes of separation of powers
with checks and balances that we deem to be illustrative of presidentialism
across all cases.
5 While recognizing the problems of Lijphart's third criterion, we must also acknowledge his
point. The overwhelming majority of presidential systems have had unitary executives, and
this characteristic does generate majoritarianism in the executive. Indeed, because of the
preponderance of single-member presidencies, criticisms of majoritarianism are central to
the criticisms of presidentialism that we review in Chapter 3.
6 For an example of the contrast between the two, one could examine Madison's Federalist
No. 5 1 versus Hamilton's Federalist No. 70.
22 Presidents and assemblies
In yet another effort to pin down the institutional nature of presiden­
tialism Giovanni Sartori (1992) suggests as a third criterion that:
the president heads the cabinet, which he or she appoints.

Since the president, rather than the assembly, names the cabinet, this
criterion implies that the entire executive is the president's domain, and is
responsible to him or her. We regard this as a necessary defining characteris­
tic of presidentialism, as it ensures consistency with the principle that the
origin and survival of the executive should be separated from the influence
of the assembly. 7
In fact, the three criteria we have identified entail a complete descrip­
tion, in institutional terms, of this principle. These three criteria, in them­
selves, however, do not provide a complete definition of presidential gov­
ernment. As we have suggested, the rationale for separating the origin and
survival of executive and legislative powers is to ensure the viability of
mutual checks. In particular, Madison sought to temper the sweeping legis­
lative authority of assemblies. An independent executive with no real law­
making power, however, would pose no obstacle to parliamentary sover­
eignty. Therefore, in defining presidentialism we must add the criterion
that the elected executive must have some lawmaking authority. We do not
offer a more specific description of this lawmaking authority because, as we
shall see, the authorities of the executive vary widely among presidential
systems, and there is no single legislative power common to all indepen­
dent executives. We shall see, moreover, that when the executive and
assembly are of different political tendencies, this overlap in lawmaking
authority can generate legislative deadlock in presidential systems; and it is
where deadlock becomes chronic that presidentialism as a system has been
subject to the most severe criticisms.
We prefer the definition of presidentialism given at the start of this
section because it clarifies the major point distinguishing presidentialism
from other regime types that feature popularly elected presidents, as well
as from parliamentarism: the separation of origin and survival. Moreover,
it clarifies who is responsible for the executive: the president, hence the
term presidential government, while in parliamentary regimes, it is parlia­
ment that is responsible for the composition of the executive. We now turn
to defining the other type that we shall devote considerable attention to in
this book, premier-presidentialism.
7 As we discuss in Chapters 5 and 7, there have been a few examples of otherwise presidential
systems in which cabinet ministers were directly accountable to the assembly. The dramati­
cally different performance of these regimes from pure presidential systems has prompted
many observers to label them parliamentary, notably in the cases of Chile 1891- 1924, Brazil
1961-3, and even Cuba in the 1940s. We share the consensus that these regimes were not
purely presidential; but we shall consider whether they are most accurately classified as
parliamentary, or as hybrid systems.
 Defining regimes with elected presidents 23

P R E M I E R - P R E S I D E N T I AL I S M

A premier-presidential regime is, as its name implies, one in which there is

both a premier (prime minister), as in a parliamentary system, and a popu­
larly elected president. The credit (or blame) for first characterizing some
regimes as "semipresidential" and describing the constitutional characteris­
tics of these regimes belongs to Maurice Duverger. 8 Duverger's concern is
primarily with the nature and performance of the French Fifth Republic.
He describes this regime as alternating between presidential and parliamen­
tary phases, according to whether the assembly majority supports the presi­
dent or not, instead of as an intermediate institutional arrangement some­
where between presidentialism and parliamentarism. Duverger does not
characterize the Fifth Republic as purely presidential nor as purely parlia­
mentary in either phase ; but he prefers the concept of alternation between
phases to that of an intermediate regime type for France. Yet the term
"semipresidential" implies a regime type that is located midway along some
continuum running from presidential to parliamentary. We find such an
implication misleading, owing to the special characteristics such regimes
exhibit. For the other European cases he discusses, there appear to be less
distinct regime phases, although Duverger does not address this point. As
we have stated, we consider such systems neither intermediate nor alternat­
ing regime types. Thus, what Duverger refers to as semipresidential, we
designate premier-presidential.
Despite our differences with Duverger, we recognize his definition as
concise and generalizable. Under premier-presidentialism :
1 . the president is elected by popular vote;
2. the president possesses considerable powers; and
3. there also exist a premier and cabinet, subject to assembly confidence, who
perform executive functions (Duverger 1980:161).

Our first criterion is the popular election of the chief executive, but, unlike
under presidentialism, the president under premier-presidentialism is not
necessarily the "chief" executive, but rather must coexist with a premier,
who is head of the government. As we shall see, the relative status of each
office within the executive may vary both across and within regimes.
Our second criterion is that the president has some political powers. The
distinction with presidentialism here is that in premier-presidential sys­
tems, presidential powers are not necessarily legislative. There may be
authorities such as submitting a bill to the electorate instead of to the
assembly or referring legislation for judicial review. The presence of other
legislative powers, such as veto or unilateral decree, might well lead the
8 The first explicit definition of semipresidentialism published in English, that we are aware
of, is Duverger (1980). Nogueira ( 1986) also provides a theoretical discussion of semi­
presidentialism.
24 Presidents and assemblies
president into conflict with the rest of the executive itself, which is depen­
dent on assembly confidence. Such powers, however, have existed and do
exist in some premier-presidential regimes. More typical - and consistent
with the regime type - are powers relating to the formation of govern­
ments, such as nominations for ministerial portfolios in addition to the
premier, as well as nonministerial appointment powers. Typically, presi­
dents in premier-presidential regimes have the power to dissolve the assem­
bly. The important point here is that premier-presidentialism does not
guarantee a legislative check by the president on the cabinet or the assem­
bly. There is, however, one presidential power that moves a regime outside
of the category of premier-presidentialism: the power to dismiss ministers
unilaterally. Such power would contradict the third criterion of the defini­
tion, namely, the dependence of the cabinet on the assembly.

OTHER REGIME TYPES

We have defined two ideal types of democratic regimes with popularly

elected presidents. Real-world regimes come in a variety of shadings and
variations on these types. W hile it would be pointless to identify numer­
ous types in an attempt to encompass all imaginable variations on the
criteria of separation of powers and presidential authorities, nonetheless
there are significant cases that clearly do not meet the criteria of either
definition we have presented. The most common type is one in which
both the president and the parliament have authority over the composi­
tion of cabinets. If the president both appoints and dismisses cabinet
ministers, and if the ministers are subject to parliamentary confidence, we
have another distinct type of regime. We designate such regimes, of which
there have been several, president-parliamentary. Just as the names presi­
dential and parliamentary for common regime types identify what elected
institution has authority over the composition of the government, and just
as the term "premier-presidential" indicates the primacy of the premier as
well as the presence of a president with significant powers, so does the
term "president-parliamentary" capture a significant feature of the re­
gime: the primacy of the president, plus the dependence of the cabinet on
parliament. Such a regime is defined thus:
1. the popular election of the president;
2. the president appoints and dismisses cabinet ministers;
3. cabinet ministers are subject to parliamentary confidence ;
4. the president has the power to dissolve parliament or legislative powers , or both.

The definition captures two senses in which these regimes represent

neither of our two ideal types, presidentialism and premier-presiden­
tialism. First, these president-parliamentary regimes provide equal author­
ity to dismiss members of the cabinet, unlike the other types. In either a
 Defining regimes with elected presidents 25
presidential or a premier-presidential regime, while both president and
assembly may play a role in cabinet formation, by nominating or confirm­
ing candidates for ministerial positions, only one of the powers may dis­
miss ministers. The importance of this asymmetry of dismissal powers,
alongside possibly shared appointment powers, will become clear in Chap­
ter 6. The second feature captured by this definition is the lack of indepen­
dent survival of assembly and executive powers, despite the great author­
ity of the president over the cabinet. In a presidential regime, maximum
separation of both origin and survival is the norm; under a president­
parliamentary system, the ability of the parliament to censure ministers
means that the survival of executive power is not separate. This latter
point is obviously true in a premier-presidential system as well, but the
difference is that under the latter type, it falls to the parliamentary major­
ity to reconstitute the government after a censure, albeit in a process
usually initiated by presidential nominations. Under president-parliamen­
tary regimes, presidents themselves reconstitute the government, subject,
of course, to the possibility of further censures. Moreover, many of these
president-parliamentary regimes provide for the power of dissolution in
addition to the powers over cabinets, thus meaning that separation of
survival is nonexistent.
In Chapter 8, after we have discussed the various dimensions of presiden­
tial power, it will be possible to rate each regime according to the powers of
presidents and the degree to which separation of survival characterizes
each system. For now, we can provide a schematic that suggests where
these regime types would be located in a two-dimensional space (Figure
2.1). There are two axes, one concerning the authority that presidents have
over the composition of cabinets, the other concerning the degree to which
the survival of assembly and executive are separated. In the upper right,
with maximum presidential cabinet authority and maximum separation, we
have the ideal-type presidentialism. In the lower left, with a minimum on
both dimensions, we have the ideal-type premier-presidentialism, in which
the president's powers are restricted to calling new elections to determine
the makeup of the parliament that alone determines the composition of the
cabinet. Slight deviations from these corner-points do not necessarily ren­
der the regime a wholly different regime type. For example, we have
already defined a regime in which the president has some leeway in making
the initial recommendation for premier or in which her or his powers of
dissolution are restricted as premier-presidential. Similarly, a regime that
requires assembly confirmation of presidential cabinet appointments would
still be presidential.
The more significant deviations are those that would place a regime near
the upper left. A regime with minimal separation of survival but maximum
presidential authority to name and remove cabinets is a distinct regime
type. As we shall see in Chapter 8, many regimes that are encountered in
26 Presidents and assemblies

Separation of assembly and cabinet survival

None Maximum
Maximum President­ Presidential
parliamentary

President's
authority
over
cabinet

Premier­ Parliamentary Assembly­

None presidential (censure only) independent

Figure 2 . 1 . A typology of democratic regimes in two dimensions

this region have experienced serious crises over the authority of presidents
and parliaments.
The lower right corner of Figure 2.1 also needs to be addressed, although
only briefly. Maximum separation of survival combined with no presiden­
tial role in naming or dismissing cabinets would be an approximation of the
Swiss regime: an executive chosen by the assembly but not removable by it.
The obvious qualifier is that Switzerland has no elected president as head
of state, but such a president would have no power over composition of the
government anyway, thus (hypothetically) such a regime would be located
at that part of the figure. We shall have little to say about this rare combina­
tion of powers, except to provide a contrast to presidentialism (see Chapter
5). We propose to call such regimes assembly-independent, to indicate the
source of executive power, as well as the lack of mutual confidence of
assembly and executive.
Even parliamentary regimes are encompassed in these two dimensions.
Normally, parliamentary government is contrasted with presidential "sepa­
ration of powers" systems as an opposite, in which the survival of the
assembly and executive are mutually dependent. In parliamentary govern­
ment, however, there is no president, or at least none with real powers.
Thus, there is nobody from whom the assembly might require separation ,
except, of course, the cabinet, which in most but not all parliamentary
systems is permitted to dissolve the assembly. The cabinet, however, is a
mere agent of the assembly to begin with. So, unlike assembly-independent
systems, the survival of the assembly and cabinet between scheduled elec­
tions is not inviolate. Yet no institution outside the assembly itself may
 Defining regimes with elected presidents 27
shorten the mandate of the assembly, as is the case under a typical premier­
presidential regime. Thus parliamentary regimes are located at the mid­
point of the dimension of separate survival.

CONCLUSION

We have now provided straightforward definitions for both presidential and

premier-presidential government, the two forms with which this book is
principally concerned. Our discussion, moreover, has generated a schema
for locating these regime types that is generalizable to hybrid systems, as well
as to parliamentary government itself. The task of establishing definitions is
important so that key terms are clarified early, but we now move on to the
more difficult and interesting job of evaluating the performance of different
types of regimes.
 3
Criticisms of presidenti alism and responses

Having defined what presidentialism is, as well as what other types of

regimes have popularly elected presidents, we now turn to the scholarly
debate about the merits of presidential regimes. In this chapter we concern
ourselves primarily with the ideal type of presidentialism that we defined at
the beginning of Chapter 2. We address the criticisms of this regime type,
then offer responses. Our purpose here is to determine whether it is even
worth considering presidentialism as a viable option for democracies, espe­
cially new democracies. For, if scholars such as Di Palma (1990) are correct
that regimes with elected presidents are "dangerous," especially for new
democracies, then there would be little point in considering possible advan­
tages of such regime types. In short, we do not find the criticisms of
presidentialism, which have been launched in a nearly one-sided debate
thus far, to be unassailable. After responding to these criticisms as they
relate directly to pure presidentialism, we turn to the ways in which
premier-presidentialism addresses the criticisms, and perhaps offers reme­
dies for some of presidentialism's more dubious qualities.

THE CASE AG AINST THE PRESIDENT

The myriad criticisms brought against presidentialism have been elabo­

rated in various combinations in a number of recent essays criticizing the
system.1 We understand presidentialism's problems, as described in the
current literature, to fall into three broad categories. The fundamental
deficiencies of the system are its:
1 . temporal rigidity;
2. majoritarian tendencies;
3. dual democratic legitimacies.

Temporal rigidity
The first group of problems, those of temporal rigidity, refer to the set
length of presidential, and congressional, terms. Although most presiden-
1 Mainwaring (1989, 1992a) presents the best current synthesis of the arguments in two essays
on presidentialism, multiparty systems, and democracy in Latin America.
 Criticisms ofpresidentialism and responses 29
tial constitutions make provisions for impeachment, they are invariably
unwieldy procedures, requiring large extraordinary majorities, which for­
mally can be used only when there is evidence of malfeasance or disregard
for constitutional procedure on the part of the chief executive. In sum,
presidents are elected for a fixed period, from four to six years, and cannot
be removed for political reasons. While constitution builders promoted
such a provision as generating stability in the executive, critics of the sys­
tem argue that is is politically crippling - and the critics have a strong case.
The principal dilemma presented by the set presidential term is what to do
with a highly unpopular chief executive; or even one who may enjoy some
popularity but who faces stiff majority opposition in congress (Mainwaring
1992a; see also Linz 1987). Such presidents are unlikely to be able to initiate
and pursue effectively their legislative programs. They may be at constant
loggerheads with the majority of members of the assembly, who are attempt­
ing to pursue a rival legislative program. These presidents may even be
widely repudiated by the majority of citizens. But they are heads of govern­
ment (as well as heads of state) until their fixed terms expire, barring im­
peachment or extra-constitutional removal. W ithout the option of a no­
confidence vote, presidential systems have no institutionalized means of
removing an unpopular - and possibly feckless - chief executive.
The opposite manifestation of the same problem occurs when a highly
popular and competent president must step down at the end of a fixed
term. W here constitutions forbid reelection (at least immediate reelec­
tion), as in most of Latin America, the country is faced with the ironic
problem of deciding between losing an effective government and tamper­
ing with the constitution. Most presidential systems - and even each party
that hopes to compete at the highest level - must generate a viable national
leader every four or five years (Linz 1987:31). Critics of presidentialism
point out that this is a substantial, and unnecessary, burden to place on the
system.
The problem of temporal rigidity in presidentialism is not restricted to
the executive's term. Assemblies, too, are elected for fixed terms, and as
with presidents their terms cannot be shortened for political reasons. The
problem just described of a weak president can alternatively be described
as one of an intransigent congress. Unlike many parliamentary systems,
presidential systems do not afford the head of state the alternative of
dissolving congress and calling new congressional elections, even in cases in
which a new president with strong popular support is elected to face a
hostile congress elected at an earlier date.
The inabilities of the assembly to remove the executive and of the execu­
tive to remove the assembly prevent either branch from resolving political
crises based on fundamental mutual opposition. In systems where presi­
dents cannot be reelected, moreover, they are effectively lame ducks imme­
diately on taking office (Mainwaring 1992a). The date of their demise -
while it may be years away - is known. And for a legislator disinclined to
30 Presidents and assemblies
compromise with the executive on legislation, the certainty of this date
might be enough to turn stubbornness to intransigence.

Majoritarianism
Another group of problems associated with presidentialism is caused by the
majoritarian tendencies within most presidential systems. The various mani­
festations of this presidential majoritarianism have been most central in
analyses of Lijphart (1989), Lijphart and Rogowski (1991), and Linz
(1987).
The majoritarian critique seems odd at first glance, in light the Madi­
sonian rationale for presidentialism as a system that offers the best protec­
tion against the tyranny of factions - be they of the minority or of the
majority. 2 The critics of presidential majoritarianism, for their part, are
concerned with the effects of electoral disproportionality - both in the
executive and in the assembly - which distort the accuracy of representa­
tion in the system and can contribute to political extremism. Even ma­
joritarianism may be an insufficient description of the extent to which
presidentialism distorts the proportional representation of voters' choices
in the executive, where M = l. Adherents of PR will point out that no
presidential election - even the biggest landslide - begins to approximate a
proportional representation of voter preferences. Since the winner by defi­
nition wins 100% of the seats when only one seat is at stake, deviation from
proportionality (Loosemore and Hanby 1971) is 100% minus whatever
share of votes the victor received. If the presidential election is a multicandi­
date affair, deviation may be 60% or even higher. Even in a two-candidate
race, deviation is seldom lower than 40%. Then the result is that the head
of state and of government may accurately represent only a minority of
voters.
Not only is the balance of voter preferences distorted in the person of the
presidents themselves, but the composition of the cabinet will likely re­
inforce the effect, rather than temper it. Unlike parliamentarism (and
premier-presidentialism), which encourages compromise among parties in
the construction of executive coalitions,3 presidentialism does neither.
W ithin the cabinet, the president is by no means first among equals, as the
prime minister ostensibly is (Lijphart 1989:4). Even in those few cases
where cabinet members are subject to congressional approval for appoint­
ment, they can be dismissed only by the president. They are generally
members of her party, or nonpartisan experts, but less often are opposition
2 See especially Federalist No. 10.
3 In parliamentary executives, deviation from proportionality is thus unlikely to be as high as
in presidential systems. However, given that cabinets are small compared to assemblies and
that many parties may be shut out of the cabinet, deviation is still higher than the corre­
sponding figure for parliament as a whole.
 Criticisms ofpresidentialism and responses 31
partisans. In most presidential systems, they cannot be members of the
assembly, thus reducing the potential for the congressional opposition to
participate in shaping the executive agenda.
Finally, as Lijphart and Rogowski (1991) have convincingly argued, mi­
nority representation in presidential cabinets is characterized by tokenism,
or "descriptive representation." Blacks, Hispanics, and women in U.S.
cabinets, for example, have tended not to be aggressive supporters of the
political agenda promoted by the minority groups to which they belong.
Rather they have generally been moderate and acquiescent to the pro­
grams of their presidents. 4 Because of the dominance of presidents over the
rest of the cabinet, and the paucity of constraints on their ability to choose
ministers, minority representation is more symbolic and less meaningful in
presidential than in parliamentary systems.
Lijphart contends that this lack of meaningful representation in cabinets
hinders the ability of presidential systems to resolve conflicts based on
multiple or deep political cleavages - those divisions that mark plural soci­
eties. Below, we examine the obstacles to consensus politics between the
assembly and executive in presidential systems. At any rate, Lijphart's
argument is that the disproportionality of representation within the execu­
tive makes politics based on consensus within that branch less likely in
presidential than in parliamentary systems. It is important to note, how­
ever, that at the core of the majoritarian criticism of presidentialism is the
presupposition that M = 1 for the executive. This point is viable for the
vast majority of presidential systems, but, as argued above, it is not essen­
tial to the nature of presidentialism.
Because of the exclusive nature of the unitary executive, the importance
of capturing the presidency becomes paramount, dwarfing all other elec­
toral goals for parties in presidential systems. This is what Linz (1987: 13-
15) describes as the "winner-take-all" nature of presidentialism. The high
stakes, and the certainty that control of the executive will not be open to
question again for a set period, raise the tension of electoral politics. In the
wake of presidential elections, moreover, the winners have no reason to try
to make amends with the losers. The ultimate prize has been secured, and
those who contributed to the victory are clamoring for compensation, per­
haps patronage or cabinet positions. The losers, moreover, have no reason
to try to cooperate with the new incumbent. There is little to be gained,
given the exclusive nature of presidential executives.
In addition to these manifestations of majoritarianism in the executive,
the representation in the assembly of parties not winning the presidency
4 Polsby (1983) has argued that American party and electoral reforms of the late 1960s and
1970s have increased the feasibility and the incentives for presidents to rely on "descriptive
representation" of minorities in their cabinets; and for cabinet members in general to act
more as personal ambassadors of the president than as representatives of distinct group
interests. See also Pitkin (1967).
32 Presidents and assemblies
can be dampened when assembly and presidential elections occur concur­
rently or in fairly close succession. The effects of electoral cycles will be
developed fully in Chapters 10 and 11. For now, the point is that the
relative timing of presidential and assembly elections can benefit the presi­
dent's party in congress considerably. This phenomenon reinforces Linz's
argument regarding the urgency of presidential elections. Evidently, the
winner takes all of the executive, and depending on the timing of congres­
sional elections, can expect to take more than her share of congress as well.
We do not universally condemn the pull of presidentialism on assembly
elections. In countries with multiple, factional, and personalized parties,
the impetus to coalesce for presidential competition may counteract the
tendency toward fragmentation. The point for now is that those who con­
cur with Lijphart on the desirability of consensual over majoritarian democ­
racy have lodged serious complaints against presidentialism. To the extent
that certain institutional designs in presidentialism may generate a congress
dominated by the president's party, it is majoritarian. Even where the
timing of congressional elections prevents this effect, the unitary executive
in most presidential systems is highly majoritarian.
Moreover, in addition to distorting representation, this phenomenon can
inspire a false sense of popular mandate (Blondel and Suarez 1981:63-5;
Suarez 1982:137). Although presidents may be elected with a mere plural­
ity of the votes - and even a small one, as Salvador Allende was in Chile in
1970 - those who are personally dominant over a homogeneous cabinet are
likely to develop an inflated notion of the support for a presidential pro­
gram and the inevitability of its implementation. This is both a function of
and a contributor to presidentialism's majoritarian problems. The nature of
elections and of the executive branch generates the false sense of mandate,
and the misperception in turn inhibits cooperation and consensus-building
between president and opposition, both within the government and with­
out, according to these critics.

Dual democratic legitimacy

The remainder of the criticisms against presidentialism share common
roots in the existence of the separate and competing claims to democratic
legitmacy made by assembly and executive. These claims are based on the
first two defining criteria of presidentialism. Because both the assembly
and executive are popularly elected, both can claim a unique popular man­
date. Moreover, because the tenure of members of each branch is unaf­
fected by relations with the other, the need for cooperation between presi­
dent and the congress is not urgent. This category of criticisms is the most
central in that the problems generated by temporal rigidity and ma­
joritarianism become practically irresolvable when presented to institutions
that lack built-in incentives to achieve compromise (Mainwaring 1989).
 Criticisms ofpresidentialism and responses 33
In the first place, presidents in multiparty systems who do not have
majority party support in congress have far less incentive to seek and
maintain lasting coalitions in congress than do parliamentary executives.
The success of any given piece of legislation may depend on putting to­
gether a majority coalition, but the survival of the executive does not. For
this reason, coalition building in many presidential regimes tends to be
piecemeal, and the incentives for cooperation offered to congresspersons
particularistic (Stockman 1985; Birnbaum and Murray 1987; Kaminsky
1989; Jacobson 1990; Mainwaring 1992a). Since the need for presidents to
cultivate support in the assembly is not as pressing as for prime ministers,
presidents tend to be less receptive to dissatisfaction with the executive
than prime ministers do, thus increasing the chance of political crisis.
A complicating factor here is that the electoral process in presidential
systems facilitates the selection of "political outsiders" as chief executives.
While this is particularly dramatic in the United States, where primary
elections determine party nominees, it is true as well where parties are less
subject to grass roots pressures. Direct election of the executive dictates
first and foremost that candidates have widespread popular appeal, even at
the expense of political experience. Moreover, presidential aspirants who
are well known independent of any leadership position in party or govern­
ment can run for highest office with the support of parties formed expressly
for the purpose of conducting presidential campaigns. As a result, we can
expect presidents to have less political experience than prime ministers. 5
Data presented by Suarez (1982) supports this proposition, at least as
regards Latin America. Examining aggregate data on heads of state and
government around the world between 1940 and 1976, Suarez found that
Latin American presidents had on average far less ministerial experience
than leaders in the Atlantic democracies. Adherents of parliamentarism
regard this as a disadvantage, in part because of the lack of experience, but
also because political outsiders are likely to be less disposed than "insiders"
to coalition building in the assembly.
In addition to producing leaders who are unable or unwilling to cooper­
ate with congress, the system's dual democratic legitimacies generate no
incentives for members of the assembly to seek rapprochement with the
executive, according to the critics. For representatives of parties other than
the president's the logic of opposition is clear, and no different from that of
opposition members in parliamentary systems: There is nothing to gain

5 The 1990 presidential election in Peru is perhaps the most striking manifestation of this
tendency ever. The novelist Mario Vargas Llosa exemplifies the phenomenon of a nation­
ally known figure who assembles a party that is purely a personal vehicle to achieve the
election of an individual to the presidency. Even more remarkable, however, was the rise of
the eventual winner, Alberto Fujimori, whose "party" was no less personalistic, and whose
platform was even more dramatically defined by its single-minded opposition to, and con­
trast with, traditional Peruvian politics.
34 Presidents and assemblies
from cooperation with the executive. Even for congresspersons from the
president's own party, however, the separation of powers weakens incen­
tives to support the president, and can create incentives to oppose him.
Valenzuela (1989b) argues that presidentialism fostered polarization of
the Chilean party system in congress. His logic is that even members of
congress from the president's own party have less to gain than to lose
through unconditional loyalty to the president. This is especially true in
systems where the president cannot be immediately reelected. Although
members of congress have an interest in their own party retaining the
presidency, the success of the incumbent does not guarantee this; nor does
his or her failure preclude it. Congresspersons from the president's own
party, then, do not have an overwhelming incentive to support a lame
duck. On the other hand, while the president may not have an election
pending, the congressperson does. Dips in presidential popularity, in par­
ticular, prompt congresspersons to distance themselves as much as possible
from the chief executive. The result, according to Valenzuela, is polariza­
tion in the assembly, both among parties, as opposition congresspersons
attack the president, and within parties, as the president's own party mem­
bers try to stake out some space far enough from the president that they
can divorce their political fates from hers or his. 6
This is an argument that presidential systems generate weak - or at least
vulnerable - executives. Still, the case is not incompatible with Lijphart's
charges of presidential majoritarianism. The paradox of presidentialism is
that, while its majoritarian characteristics create great expectations for the
potency of the executive, the competing incentives within the system pre­
vent these expectations from being realized (Moe 1987; Lijphart 1990a:76-
7; Mainwaring 1992a).
In a further argument that combines the themes of temporal rigidity and
separate democratic legitimacies, Linz suggests that presidentialism suf­
fers, in effect, from too much electoral accountability. Because it is so
difficult to build support coalitions for the executive after presidential elec­
tions, he points out, these coalitions must be presented before the election,
offering to the voters the choice of providing the president with a congres­
sional majority. But once the electoral decision is made, according to Linz,
political leaders are left with less room to maneuver, to make "deals diffi­
cult to defend in public but that might be necessary" (Linz 1987:33).
Linz wants to allow room for coalition building away from the public
spotlight. But the same process can equally be described as private deal­
making by political elites. The argument is less than compelling in that it
implies a proclivity toward elitist politics and dismisses the principle of
6 The literature on midterm elections in the United States examines at length the proposition
that voters cast congressional ballots based on presidential performance, and that congres­
sional candidates position themselves accordingly. See Campbell et al. (1964) ; Kernell
(1977) ; Erikson (1988).
 Criticisms ofpresidentialism and responses 35
accountability. Furthermore, the critics of presidentialism take aim primar­
ily at systems in which reelection is prohibited. Proponents of such a rule
argue that the lame duck president should not be constrained by short-term
electoral considerations in pursuing the national interest. Indeed, we see
no reason to believe that parliamentary leaders should be any more willing
to take action that is "difficult to defend, but necessary" than should a
president barred from reelection.
At any rate, the critics of presidentialism rightly point out that the diffi­
culties of building and sustaining support for the president's legislative
agenda pose a serious problem for effective government. The principle of
separating legislative from executive powers as a check against the abuse of
either can go awry when neither branch is able to exercise power effec­
tively. This dilemma prompts the question of whether a president lacking
congressional support would pose a problem if legislative initiative rested
primarily in congress. If coalition building is not primarily an action taken
by the executive to coordinate the congress, then the disincentives to co­
operation between the two branches should not present a problem. The
evidence from the United States seems to support this proposition. The
U.S. Congress has functioned most of the time since World War II with
opposition majorities. It is not the most decisive and effective legislative
body imaginable, but it has avoided chronic deadlock with the executive.
Further, the Chilean Congress was, for much of the twentieth century, one
of the most independent and activist of Latin America's assemblies. Al­
though minority presidents were a regular phenomenon in Chile, immobil­
ism and deadlock did not reach crisis levels until the 1970s.7 The evidence is
not conclusive; but it is suggestive that the critics of presidentialism might
qualify their attack on the separation of powers, to direct it at those sys­
tems in which congresses are primarily reactive, rather than initiative
(Mainwaring 1992a). Such a qualification would then direct attention to the
question of whether the problem of presidential immobilism could be allevi­
ated by strengthening congresses, a topic to which we devote considerable
attention in Chapters 5, 7, 8, and 9. However, where multiple parties
exhibit substantial ideological differences, confrontations between con­
gress and a president who perforce represents only one political tendency
may be chronic.
The last criticism of presidentialism we shall address holds that chronic
legislative-executive conflict presents dangers far greater than that of dead­
lock; it invites rule by executive decree (Mainwaring 1992a) and regime
instability - even military coups.
7 Indeed, one could argue that, had the Eduardo Frei government not obtained a constitu­
tional weakening of the Chilean Congress, the center-right opposition in the early 1970s
might have curbed Allende's program, and possibly even prevented the crisis of 1973 from
escalating to the point where the military was able to seize power with substantial civilian
support. These points are addressed in Chapter 9.
36 Presidents and assemblies
The case that immobilism invites rule by decree is straightforward and is
supported by the example of Colombia in the years during and after the
National Front (Hartlyn 1989:16-18). Intense factionalism in a congress
with limited powers of legislative initiative made coalition building virtually
impossible, even though both major parties nominally supported the
Front's common presidential candidate at election time. Owing to their
inability to secure support in the congress for their legislative programs,
Colombian presidents ruled by decree under the State of Emergency as a
matter of course throughout the 1960s and 1970s (V asquez 1979).
Some authors have extended the criticism of presidential immobilism to
conclude that the executive's inability to maintain consistent support has
directly caused regime instability and encouraged military coups. Scott
(1973:290) argued that civilian political elites would encourage and wel­
come military intervention in the face of congressional-presidential immo­
bilism. Furthermore, using worldwide data from 1940 through the 1970s,
Suarez (1982) demonstrates that Latin American governments have en­
dured for less time on average than those in other parts of the developing
world, and that these presidencies have had an inordinate proportion of
unconstitutional transitions of power. He goes on to argue that in at least
three cases - the overthrows of Presidents Fernando Belaunde in Peru,
Joao Goulart in Brazil, and Salvador Allende in Chile - congressional­
presidential deadlock was the primary factor triggering the military coups.
Valenzuela's (1978) account of polarization and crisis in Chile under Al­
lende certainly supports Suarez's thesis.

PRE S I D E N T I A L I S M A N D A UTHORITA R I A N ISM

The suggestion that presidential democracy is prone to breakdown, leading

to authoritarian government, is the most troubling of all the criticisms
leveled by adherents of parliamentarism. Before offering a defense of presi­
dentialism, then, we suggest reconsidering the logic by which presiden­
tialism is tied to authoritarianism. Two points, in particular, need to be
addressed:
1 . the difference between presidential constitutions that are formally democratic
but fail, and those in which formal presidential dominance over other actors
precludes workable checks and balances to begin with; and
2. the question of whether parliamentary democracy has in fact performed as well
relative to presidential democracy as parliamentarism's adherents have claimed.
One key criticism of presidentialism stems from a failure to differentiate
presidential constitutions developed in an authoritarian context from those
in which power is less concentrated in the executive. A fundamental criti­
cism of presidential systems concerns the extent of presidential powers and
the independence of presidents from party and assembly. Lijphart (1989) is
 Criticisms ofpresidentialism and responses 37
especially clear on this point when he argues that parliamentary cabinets
are necessarily more collegial than presidential cabinets. Even a prime
minister who appears to exercise dominant power over his or her col­
leagues, as did Margaret Thatcher in Britain, is ultimately responsible to
the party. The Conservative Party's removal of Thatcher from office in
November 1990 after some of her erstwhile cabinet colleagues publicly
criticized her policies only underscores the point. Under a presidential
system, on the other hand, the cabinet, and especially the head of govern­
ment, is not responsible to other elected officials for survival.
Even in an authoritarian context, a formally parliamentary constitution
may allow the party to retain significant political power over the authoritar­
ian head of government. Thus, where a party organization matters, formal
parliamentary rules may be expected. Indeed, the Communist systems of
Eastern Europe and the Soviet Union before 1990 were generally parlia­
mentary in form. That this feature from time to time worked is shown by
the replacement of such leaders as Nikita Krushchev and Edward Gierek
by their own parties.
More personalized authoritarian systems where party either does not
exist or is less developed, such as those of Chile under Augusto Pinochet
and Paraguay under Alfredo Stroessner, have tended to be presidential. In
Mexico, where the official party is not a mass party in the Leninist model
but rather was created as a means of circulating the top leadership among
several competing caudillos (Craig and Cornelius 1991), a presidential sys­
tem was adopted. And, tellingly, when Mikhail Gorbachev sought a way to
protect his political power from elements of the party that he feared might
want to replace him, he opted for a presidential constitution. In each case
just mentioned, not only was executive power protected from party and
assembly by means of presidentialism, it was also given significant powers
over legislation.
This latter point, the extent of legislative powers in the hands of the
president, affects the policy content of a presidential system's legislative
process more than the degree to which executives are responsible to their
own parties. Additionally, the more the president is given substantial legis­
lative authority, the more the outcome of conflicts between the assembly
and executive is skewed in favor of the latter. Further, many presidential
democracies have inherited constitutions from an authoritarian period and
those constitutions were developed at the service of a dictator. This has
also been the case in some transitions where new constitutions have not
been written. 8 W herever such resulting democratic regimes prove trouble­
some, we have further evidence for our proposition that it is the balance of
presidential-congressional powers, more than presidentialism per se, that
8 See Geddes (1990) for a discussion of transitions in Latin America in which new constitu-
tions were written versus those in which an existing or prior constitution was retained.
38 Presidents and assemblies
has hampered democratization in many countries. In Chapters 7 and 8, we
undertake a systematic analysis of the many possible ways of arranging the
relative powers of presidents and assemblies, and the implications of such
arrangements for the balance of powers.
From this postulated affinity of dictators and constitutions with strong
presidential powers, then, flow some of the exaggerated criticisms of presi­
dential political powers in the democratic context. Precisely because a presi­
dential constitution allows the head of government to be sheltered even
from his own party and can be - but need not be - designed as well to give
superior legislative powers to the president, authoritarianism and presiden­
tialism have tended to go together. However, the category denoted by the
term presidential democracy includes wide variation in the extent to which
presidents have legislative powers. No criticism of authoritarian tendencies
in presidential democracy is adequate without reference to authoritarian
legacies that remain in many presidential constitutions in the form of supe­
rior legislative powers accorded the executive.

I S P R E S I D E N T I A L I S M E S PE C I A L L Y P RO N E TO B R E A K D O W N ?

The criticisms of presidential democracy that we have reviewed are more

than simply arguments about the quality of democracy. Indeed, even Linz,
perhaps the harshest of the current critics of presidentialism, concedes that
direct election of the executive is highly democratic. So does Lijphart
(1984). Rather, as we have noted, the critics tend to express their argu­
ments in terms of the likelihood of stable democracy. Parliamentary democ­
racy, because it does not suffer from maladies such as temporal rigidity and
competing claims to legitimacy and because it has the vote of confidence to
resolve interbranch conflicts, is more conducive to stable democracy, the
critics say. Interbranch conflicts under presidentialism are likely to be re­
solved through coups or other nondemocratic measures. Thus, according
to critics, presidential systems are especially prone to breakdown.
Mainwaring (1992a) provides a list of stable democracies, where stability
is defined in terms of longevity, that appears to be a clear vindication of the
argument that presidential systems are inimical to the continued function­
ing of democracy. He lists thirty-two democracies that, as of 1991, had been
democratic for at least twenty-five years. Of those countries, twenty-three
(72%) are parliamentary. But is this the final word on the relationship
between regime type and democratic longevity? We should like to try a
different approach. Rather than take a "snapshot" of democracies as of a
specific date, we shall look at democratic failures throughout the twentieth
century.
Why should our method be a more reliable measure of the proclivity (or
lack thereof) of different democratic regime types to break down? There are
two basic reasons, both of which stem from historical observations. The first
 Criticisms ofpresidentialism and responses 39
historical reason is that there have been two waves of breakdowns of democ­
racy in this century, one between World War I and World War II and the
other in the 1960s. The first wave claimed mostly parliamentary regimes
(and no true presidential regimes). The second claimed mostly, but not
exclusively, presidential systems. Of the three paperback volumes contain­
ing case studies on The Breakdown ofDemocratic Regimes (Linz and Stepan
1978) that were originally published, only the volume on Europe between
the wars was out of print by 1990. Apparently there is not as much of a
market for studies of failures in an earlier period - when the failed regimes
happened to have governments based on parliamentary confidence - as
there is for the contemporary period, when several Latin American presiden­
tial systems broke down.
The second historical reason that our method of counting breakdowns is
more valid than counting twenty-five-year-old (or older) democracies as of
one point in time is that there have been three waves of democratization in
this century (cf. Huntington 1991). The first was in the aftermath of World
War I. The second occurred after decolonization in the post-World War II
period (and extended up to the early 1960s). The third began in the 1980s
and continues into the 1990s. Of these, only the third wave has given birth
mostly to presidential (and some premier-presidential) democracies. Thus
the method of counting successes, as Mainwaring does, cannot take ac­
count of new presidential democracies that, for all we know, might prove to
be long-term successes, any more than it can account for failures (whether
presidential or parliamentary) in earlier periods.
We have, in Table 3. 1 compiled a list of democratic failures. We define
a failure as a breakdown that led to a replacement of the regime by a
nondemocratic regime. We thus do not count cases in which there was a
brief lapse of democratic authority that was followed by either a new
democratic constitution or a renewal of the former one. Such cases are
not regime failures and replacements, but rather what Linz (1978) calls
"cases of reequilibration." Cases of reequilibration include both presiden­
tial and parliamentary systems that suffered crises: Chile in 1925 and
1933, France in 1958, and India in 1972. In the Chilean cases and in
France, democratic authority was restored in less than a year. And in
India, where an election led to the ruling party's defeat despite emergency
rule, there was never a break in the constitutional order. Indeed, in these
cases arguably democracy (of whatever institutional form) reemerged
stronger from the crisis. Another criterion we have used is to count only
those cases in which at least two consecutive general elections were held
before the breakdown. In so doing we avoid ambiguous cases in which
one election was held under the watchful eye of a departing colonial
power or as a mere "demonstration" for foreign consumption (Herman
and Brodhead 1984). If such an election was followed by another and only
subsequently by a breakdown, we count it.
40 Presidents and assemblies

Table 3 . 1 . Breakdowns of democratic regimes in the twentieth century by regime

type

Parliamentary Presidential Other

systems systems types

*Burma 1962 *Argentina 1930 Austria 1933 (premier-presidential)

Estonia 1934 *Bolivia 1 964 *Ecuador 1962 (president-
*Fiji 1 988 *Brazil 1964 parliamentary)
Greece 1936 *Chile 1973 Germany 1933 (pres.-parl.)
Greece 1967 *Colombia 1953 *Korea 1 961 (pres.-parl.)
*Guyana 1978 *Cuba 1954 *Peru 1 968 (pres.-parl.)
Italy 1922 *Guatemala 1954 *Sri Lanka 1982 (pres.-parl.)
*Kenya 1969 *Korea 1972
Latvia 1934 *Nigeria 1983
Lithuania 1 926 *Panama 1968
*Nigeria 1966 *Philippines 1 972
*Pakistan 1954 *Uruguay 1973
*Pakistan 1977
Portugal 1926
*Sierra Leone 1967
*Singapore 1 972
*Somalia 1 969
Spain 1936
*Surinam 1 982
*Thailand 1976
*Turkey 1 980
*Third World cases

We have identified twelve presidential regimes and twenty- one parlia­

mentary regimes that have br oken d own in the twentieth century. Additi on­
ally, we identify six other types that have broken d own. Except for the
premier-presidential regime of interwar Austria, these six " other" break­
d owns were all president-parliamentary regimes.
Our greater number of parliamentary failures than presidential failures
is hardly what one w ould expect fr om reading m ost of the c omparative
literature. Perhaps parliamentary regimes have received m ore credit than
they are due as means t o resolve p olitical conflicts. Some refinement of
the sample needs t o be made, h owever. F or example, if parliamentary
regimes were far m ore c omm on, especially in the Third W orld where the
success of any type of dem ocracy has been minimal, then our simple
counting of breakdowns might obscure a greater rate of presidential
breakd owns.
Table 3.2 sh ows dem ocratic regimes that have not br oken d own in the
twentieth century or, f or newer regimes, that have passed the thresh old of
tw o electi ons that we referred t o ab ove. Included are only th ose c ountries
with at least 200,000 p opulation, t o av oid micr ostates. The table lists twelve
presidential systems, twenty-seven parliamentary, and eight other types,
 Criticisms ofpresidentialism and responses 41

Table 3.2. Countries that have held at least two democratic elections without break­
down, as of 1991

Parliamentary Presidential Other

systems systems types
Australia *Argentina Austria (premier-presidential)
*Bahamas *Brazil *Bolivia (assembly-independent)
*Barbados *Colombia *Ecuador (president-parliamentary)
Belgium *Costa Rica Finland (premier-presidential)
*Botswana *Dominican Republic France (premier-presidential)
Canada *El Salvador Iceland (premier-presidential)
Denmark *Guatemala *Peru (president-parliamentary)
Germany *Honduras Portugal (premier-presidential)
Greece *Senegal Switzerland (assembly-independent)
Ireland United States
Israel *Uruguay
Italy *Venezuela
*Jamaica
Japan
Luxembourg
*Malaysia
Malta
Netherlands
New Zealand
Norway
*Papua New Guinea
*Solomon Islands
Spain
Sweden
*Trinidad and Tobago
*Turkey
United Kingdom
*Third World cases

indicating that parliamentary regimes are more common. The breakdown

rate is higher for presidential regimes, 50. 0%, compared to 43.8% for
parliamentary regimes.
When we consider only Third World cases, however, we find that just
over half (52.2%) of the presidential regimes in the less developed coun­
tries have broken down, while a higher percentage (59.1%) of the parlia­
mentary regimes have. A remarkable fact revealed by the tables is the
extent to which presidentialism is a Third World phenomenon, as only the
United States is a developed nation with a presidential regime. Also strik­
ing is that most of the presidential failures from Table 3.1 reappear as
democracies subsequently in Table 3.2. Among the parliamentary failures
in the Third World, on the other hand, only Turkey reappears. While we
shall not attempt to provide a theoretical justification for this observation,
which could simply be spurious, it does indicate that the parliamentary
failures in the Third World have tended to be irreversible, at least as of
42 Presidents and assemblies
1991. More importantly, parliamentarism has not fared any better in the
Third World than has presidentialism; arguably, it has fared worse.
Indeed, we find no justification for the claim of Linz and others that
presidentialism is inherently more prone to crises that lead to breakdown.
Nor is it our purpose to stand that argument on its head and argue that
parliamentarism is inherently prone to breakdown. On the other hand,
these caveats about the link between regime type and regime breakdown
does not mean that institutional choice does not matter; some presidential
democracies that broke down might have survived had they been parliamen­
tary (Chile in 1973, perhaps). It is also possible that some failed parliamen­
tary democracies might have survived had they been presidential instead
(the first republics of Nigeria and Pakistan, perhaps). Some crises may be
so severe that no constitutional system could endure (perhaps the Weimar
Republic). 9 Our central point here is that the argument against presiden­
tialism has often been expressed universalistically, as if presidentialism as a
regime type were prone to breakdown, while parliamentarism contained
conflict-regulating mechanisms that would ordinarily shield it from break­
down. The historical evidence presented in Tables 3.1 and 3.2 suggests that
the critics of presidential systems should rethink their case.
The presence of so many young democracies in our list of failed parlia­
mentary systems might appear to be stacking the deck against that regime
type. Ironically, however, much of the criticism of presidentialism and
support for parliamentarism centers around the superior appropriateness
of a parliamentary system for new democracies (Di Palma 1990 ; Main­
waring 1992a). So a list that includes young democracies is hardly inconsis­
tent with the critics' means of conceptualizing the problems and choices at
hand. There is also a theoretical reason for not excluding new democracies,
as long as they have survived long enough to hold a second general elec­
tion. Conceptually we agree with Di Palma that a lengthy period of "habit­
uation" (Rustow 1970) is not necessary to prove the "consolidation" and
"legitimacy" of a democracy. Once a transition to a democratic regime is
complete, the new institutions either obtain what Di Palma calls "behav­
ioral compliance" or they do not. Among new democracies, cases of both
regime types have failed to obtain the behavioral compliance of political
elites. Indeed, some of the new presidential democracies established since
around 1980 have already outlived several of the young parliamentary
systems that failed in the 1960s or earlier.
The critics of presidentialism might counter that the failures of parlia­
mentarism in the 1920s and 1930s were concentrated in one region and
resulted from international circumstances, such as economic disaster and
the imposed peace of the Treaty of Versailles. However, as we foreshad-
9 However, Adolf Hitler three times ran for the presidency and lost. He was ultimately
appointed chancellor because only a cabinet headed by him could count on parliamentary
confidence, which the Weimar constitution required.
 Criticisms ofpresidentialism and responses 43
owed above, a similar argument resting upon regional concentration of
breakdowns and international determinants can be made for Latin Amer­
ica in the 1960s (Therborn 1979). Besides, the proponents of parlia­
mentarism have not framed their case in terms of anything other than the
domestic political process and the ability of the regime to handle pressures,
whatever the source of those pressures.
To be fair, Linz (1987) has stated his defense of parliamentary democracy
in terms of other institutional choices besides the basic choice of parlia­
mentarism or presidentialism that can affect the chances for stable democ­
racy. Specifically, he cites the adoption of electoral systems less conducive
to party system fragmentation and the constructive vote of no-confidence
when comparing the interwar German and Spanish regimes to those coun­
tries' current systems. We wholeheartedly agree that other choices besides
the basic regime type matter. We simply argue that, to be fair to presiden­
tial democracy, the critics should recognize that choices within presidential
democracy as well can make a difference. Neither parliamentary nor presi­
dential democracy is a unified regime type. Nor, as we have argued, do
these commonly recognized types constitute an exhaustive typology. In­
deed, that is one of the principal points of this book. Before going on to
discuss institutional choices within presidential democracy that may be
more or less conducive to democratic success, let us first address the ques­
tion of why one might favor any variant of a presidential system.

I N T H E PRESI D E N T ' S DEFENSE

The criticisms of presidentialism have been discussed at length and refined

in a number of recent analyses. Defenses of presidentialism, on the other
hand, have been few and far between. 10 We should reemphasize that our
defense of presidentialism is not meant to imply that we regard presiden­
tialism in all cases to be a superior regime type to parliamentarism. Indeed,
we acknowledge most of the criticisms of presidentialism reviewed above
as penetrating. Nor do we reject the common empirical assertion that
presidential institutions, as they have been realized, have often performed
poorly. Much of our project in the chapters ahead will consist of discussing
the variations in institutional arrangements found within the presidential­
regime type, as well as premier-presidentialism and other forms. In some
cases, we shall suggest alternative arrangements that might mitigate prob­
lems of less successful systems. At any rate, for now our discussion remains
abstract; the only qualities of the presidentialism we discuss here are those
outlined at the beginning of this chapter. We suggest four areas in which
10 Hartlyn (1989) offers intermittent and strictly qualified praise of Colombia's National
Front as a consociational adaptation of presidentialism. Rather than a defense of presiden­
tialism. however, Hartlyn's response to the system's critics is that even parliamentarism
would not have prevented the conflicts that made the National Front necessary.
44 Presidents and assemblies
presidentialism offers at least potential advantages over parliamentary de­
mocracy. In some respects, these advantages can be understood to reflect
merely the flip side of presidentialism's deficiencies. For example, vetoes
built into the legislative process can be defended as protection against the
capricious use of state power, rather than attacked as the roots of
immobilism. Broadly, then, in response to the advocates of parliamentar­
ism we offer the following ideal-type advantages of presidentialism. What
presidentialism provides, to a degree that parliamentarism does not, is:
1. accountability;
2. identifiability;
3. mutual checks; and
4. an arbiter.

Accountability
Accountability describes the degree and means by which elected policymak­
ers are responsible to citizens. The more straightforward the connection
between the choices made by the electorate at the ballot box and the
expectations to which policymakers are held, the greater accountability.
Accountability is closely related to concepts such as retrospective voting
(Fiorina 1981), or "throwing the rascals out. " Under these conceptions of
electoral connection between constituents and government, voters need
not be able to assess prospectively the policy directions with which they
wish to endow a mandate on the government (Riker 1982a). Rather, voters
need only have the opportunity to impose sanctions on elected officials at
the next election. Officials can be expected to anticipate such sanctions if
they stray too far from voters' wishes between elections (Eulau and Prewitt
1973; Pennock 1979). As Powell (1989) has observed, this form of account­
ability is weakened if governments tend to change between elections, and
especially if the government changes shortly before an election, as happens
frequently in parliamentary systems. 1 1 Under such circumstances, the re­
sponsibility for policy in the interelection period is less clear, making it
difficult for voters to know whom to hold accountable on election day.
On the principle of maximizing direct accountability between voters
and elected officials, presidentialism is clearly superior to parliamentar­
ism, since voters vote directly for an executive that cannot be removed by
shifting coalitions in the assembly. However, the principle itself is not
universally accepted. Linz's argument regarding the inflexibility of politi­
cal alliances in presidential systems and the advantage for members of
congress to be able to make deals that are difficult to defend before the
electorate is a direct contradiction of the principle of accountability. As
we have stated, we are not convinced of Linz's case. W ithout wandering
1 1 Strom ( 1990:75) notes that only 49% of the governments in his sample of parliamentary
regimes were formed near elections.
 Criticisms ofpresidentialism and responses 45
into an extended analysis of whether the electorate - or any particular
electorate - is capable of exercising the kind of citizen control implied
here, we simply suggest that the principle of maximized opportunity for
accountability is complementary with the principles of representative de­
mocracy. The debate between presidentialism and parliamentarism is over
the most effective institutional means of maintaining representative de­
mocracy. Therefore, the direct election of the executive by the electorate ,
and the accountability of the president to the public that this entails, will
be considered an advantage of presidentialism.
In addition to its direct relationship to executive accountability, presiden­
tialism clarifies the motivations behind the votes of individual legislators on
important issues. 12 W here legislative voting does not have implications for
government survival, a legislator or a party leadership can be held directly
accountable for votes on legislation.

Identifiability
An aspect of the connection between voters and the electoral outcomes
that is related to accountability is the degree to which voters can identify
before the election the likely alternative governments that may emerge
after the election. The concept of identifiability is closely linked to theories
of democratic government that call on voters to be able to provide a policy
mandate for their elected representatives (Ranney 1962). Unlike account­
ability, which requires only that voters be afforded clear governmental
responsibility in order to punish or reward their representatives retrospec­
tively, maximizing identifiability requires that voters have an opportunity
to make a clear prospective choice.
Strom (1990) has developed an indicator of "identifiability" for his sam­
ple of stable parliamentary democracies. The indicator can run from O to 1,
with 1 indicating that in 100% of a given regime post-World War II elec­
tions the resulting government was identifiable as a likely result of the
election at the time voters went to the polls. The average of all Strom's
sample is .39, meaning that far more than half the time voters could not
know what government they were voting for. Yet, under a parliamentary
regime, voting for MP or party list is the only way voters can influence the
choice of executive.
Among parliamentary systems there is a great range of identifiability
scores (Strom 1990:75). Not surprisingly, systems with only two major
parties, such as Canada and Britain, have scores of 1.00. Those with a clear
left-right divide, such as Sweden (1.00), Norway (.83), or Ireland (.83)
12 Of course, under list PR systems, the relationship between voter and representative will
always be less than clear. This is a function not of presidentialism or parliamentarism but
of ballot structure and electoral formula. The point remains, however, that where account­
ability is possible under presidentialism, it becomes more difficult under parliamentarism.
46 Presidents and assemblies
also score high. But multiparty systems without a two-bloc format tend to
be far lower, with extremely low figures for some, such as Belgium (.10),
Israel (.14), and Italy (.12).
In presidential systems, identifiability is bound to approach 1.00 in most
cases. As we shall see in Chapters 10 and 11, even in multiparty systems, if
the president is elected by plurality rule, ordinarily only two "serious"
presidential candidates compete in the election for chief executive. 13 Thus,
if we value the ability of voters to be able to know which alternative
executives are likely to emerge out of an election, presidentialism appears
superior to most variants of parliamentarism.
To this argument one might respond that while voters in a typical two­
candidate presidential race know what their alternatives are, their range of
choice is sharply restricted. A vote for a minority viewpoint is effectively
"wasted" in the vote for executive in a presidential system, while in a
parliamentary system, a minor party may be able to play a role in cabinet
formation. On the other hand, such a party may be shut out entirely due to
the preferences of leaders of major parties. We could sum up this trade-off
by indicating that, given multiparty competition for the assembly, a parlia­
mentary system is good at allowing voters, with their range of partisan
options, to know what they are asking for, but presidentialism makes it
clearer what they are getting. We do not wish to claim that voters are
universally able to make use of prospective voting, but we suggest that
democratic constitutions ought to give voters the opportunity to do so.
Later, we discuss some institutional variations on regimes with popularly
elected executives - specifically, premier-presidentialism, collegial presi­
dencies, and modifications to the electoral cycle - which are potential ways
to enhance both the advantages of accountability and identifiability, on the
one hand, and the ability to make a vote for a minor party that is meaning­
ful from the standpoint of policy output.

Mutual checks
In much the same way that it clarifies the relationship between representa­
tive and constituent, presidentialism clarifies the stakes between legislator
and the executive on any given issue. Following the logic above, members
of the assembly freed from the threat of a vote of confidence can ratify or
check executive initiatives based on the merit of the legislation itself rather
than on the survival of the government. Of course, it is precisely this
independence of the assembly that generates the problem of immobilism.
Is legislative independence an advantage or a flaw? The critics of presiden­
tialism unanimously see it as a drawback, although Mainwaring (1992a)
qualifies this position somewhat in his argument that presidentialism is
13 In Chapter 10 we propose an electoral method, the "double complement rule.'" which
safeguards plurality rule against the possibility of very narrowly endorsed victors.
 Criticisms ofpresidentialism and responses 47
particularly inappropriate in multiparty democracy, where minority presi­
dents are the rule. Still, the flip side of immobilism is rule by consensus in
the sense that one party is prevented from ruling alone; where a president
lacks a majority in congress, consensus may be favored, provided the bal­
ance of powers is conducive to compromising.
If, on the other hand, the president's party, or a stable coalition support­
ing him or her, has a majority in congress, assembly independence may be
a distinct advantage of presidentialism. Under such conditions, we do not
expect immobilism. Moreover, it is under such conditions that parliamen­
tary systems exhibit quasi-majoritarian characteristics. The incentive not to
jeopardize the survival of the government pressures members of parlia­
ment whose parties hold executive office not to buck cabinet directives.
The opposition will unlikely be able to solicit any defectors from among the
majority to oppose the cabinet. The ability to coerce parliamentary majori­
ties, then, can render the bloc of opposition votes in parliament meaning­
less. On the other hand, if the ability of parties to influence legislation from
outside the cabinet is high (e.g., as in Belgium and Italy), we tend to find
lower scores of identifiability and the most unstable executives (Strom
1990:73). Either condition - the impotence of the opposition, or cabinet
instability - can be regarded as a drawback to parliamentary democracy.
In contrast, by reducing the coercive capacity of the executive over
members of the assembly majority, presidentialism preserves the viability
of the opposition, without necessarily endangering stability - provided the
executive can solicit some defections on particular votes. Insofar as we
share Lijphart's preference for consensual government over majoritar­
ianism, we regard this characteristic as an advantage of presidentialism. At
the same time, Mainwaring's point is well taken. Presidentialism without a
presidential majority can be problematic. The question becomes, Is it possi­
ble to preserve the congressional check under majority conditions, while
providing for the resolution of deadlock in the absence of a majority? 14
This is a question to which we turn later in the chapter.
On the subject of mutual checks, it is worth pointing out that presiden­
tialism fortifies the executive check over the assembly as well as the re­
verse. W here legislative initiative lies predominantly in the executive, this
is not so much an issue. However, in systems where assemblies are dy­
namic, such as the United States and Chile until the reforms of the late
1960s, this check can be critical. The logic is straightforward, as follows.

14 Requiring a constructive vote of no-confidence may well free up legislators from the
parties in cabinet to oppose cabinet initiatives occasionally - provided they are confident
that no alternative cabinet can be offered to defeat the current government. However. in
two cases that use this format, West Germany and Spain, the cabinet has the authority to
make any proposal before the parliament an ordinary vote of confidence, thus making the
whip as effective as it can be in any parliamentary system. Only votes of no-confidence
initiated in the assembly must be "constructive. "
48 Presidents and assemblies
Where executive survival is based on assembly confidence, the executive is
in no position to resist assembly initiative. In those parliamentary systems
in which there are fragmented party systems and considerable initiative in
parliament, the potential for coercion of the executive by the assembly is
clear and is the key reason for frequent cabinet crises. Under presiden­
tialism, on the other hand, use of the veto to resist assembly majorities
does not jeopardize the executive's tenure. In sum, under conditions in
which mutual checks do not threaten presidentialism with immobilism ,
they act to prevent majoritarianism, and should be regarded as advantages
of the system.

Arbiter
Turning on its head the criticism that presidents have no means of enforcing
discipline on members of congress, some authors have suggested that the
president might serve as an above-partisan arbiter of political conflict. 15 In
a sense, this approach seeks a silver lining in the fact that the executive
cannot threaten legislators with a crisis of government stability in order to
secure support, as a prime minister can by declaring a given measure
before parliament to be a matter of confidence.
The arbiter president resembles Neustadt's (1960) ad hoc bargainer, or
cajoler. It is precisely the distance from congressional alliances which the
separation of powers ensures that allows the president to act as a modera­
tor of dispute in order to secure a legislative agreement. This distance,
taken to its extreme, implies both a lack of influence, and perhaps as result,
a lack of fetter. Because presidents serve a fixed term and often are not
eligible for reelection, they will allegedly not be averse to making conces­
sions even on issues central to the party platform, especially if the alterna­
tive is no compromise, and ultimate deadlock.
There are two primary problems with this view. The first is that to expect
presidents to act in a nonpartisan manner under any circumstances is not
entirely realistic. It is true that presidents are less bound by partisan con­
straints than are prime ministers, but their independence is not complete.
Most are, after all, the leaders of major - if not majority - congressional
parties. The second problem is that the case for the president as arbiter is,
in effect, a celebration of presidential weakness and, at the extreme, com­
promises accountability. Weakness over legislation itself is not necessarily a
bad thing. The Madisonian perspective, for example, prefers a weak presi­
dent. The problem is that arbitration will be most needed under pure
presidentialism at times of regime crisis, when support for and stature of
the presidency is lowest (Valenzuela 1989a: 15).
Basically, the scenarios are as follows: When the president has majority
15 See Karl (1986:213) and Valenzuela (1989a:19) ; also Neustadt (1960).
 Criticisms of presidentialism and responses 49
support, there should be no need for an arbiter. Congressional support for
executive initiatives should be regular, although the congressional check
remains viable. W hen the president has no majority, the arbiter is the role
of last recourse. A president's separation from entanglement in assembly
alliances should make her or him a better arbiter than would be a minority
prime minister. 16 If the president may stand for reelection, accountability
can still be preserved.

To this point, we have established a working definition of presidentialism,

reviewed the central criticisms leveled against presidential systems, and
outlined some tentative advantages of such systems over parliamentarism.
We stress once more that the criticisms of presidentialism are not directed
entirely against what we regard as the criteria that define presidential govern­
ment, but rather against many of its associated characteristics, such as the
unitary executive, the tendency to generate presidents who face hostile
congressional majorities, or the ineligibility of the president for reelection.
At the same time, we acknowledge that many of the criticisms directly
address the core elements of presidentialism. Therefore, let us consider
another regime type - premier-presidentialism - that retains some of the
advantages of presidentialism, while showing the potential to diminish some
of presidentialism's defects.

H O W P R E M I E R - P R E S I D E N T I AL I S M A D D R E S S E S T H E
C R I T I C I S MS O F P R E S I D E N T I A L I S M

In order to assess the potential advantages of premier-presidentialism, we

return to the framework constructed to review the criticisms and defenses of
presidentialism. We consider in turn the issues of temporal rigidity, ma­
joritarianism, dual legitimacies, mutual checks, and the viability of the presi­
dent as arbiter and as head of state.17 Using both the criticisms and the
defense of presidentialism as points of contrast, we hope to illustrate the
potential virtues as well as the ambiguities of premier-presidentialism. In the
following chapter, we shall undertake brief case studies of the performance
of premier-presidential government in five countries. We suggest there that
16 Ultimately, the case for the president's potential as arbiter is far more compelling under
premier-presidential systems. In such systems, the president without majority support
enjoys at least equal distance from involvement in partisan alliances as under pure presi­
dentialism; yet the provision for the formation of a government in opposition to the
president establishes a safety valve short of regime breakdown by which political crisis can
be quelled.
17 It seems straightforward that the accountability of the president to the electorate should be
equal in premier-presidential systems to that in presidential systems. For this reason, we
will not review the accountability argument separately in this section. The question of the
incentives for legislative support of the president is considered in the section on mutual
checks.
50 Presidents and assemblies
the performance of premier-presidential systems has been largely attribut­
able to the institutional arrangements unique to each system.

Temporal rigidity
Presidential terms under premier-presidentialism are generally fixed, as
under presidentialism. 18 The problem of a president facing an opposition
assembly, however, is not defined by the same temporal rigidity as under
pure presidentialism, for two reasons. In the first place, the president is not
necessarily the head of government. W hen an assembly will not support a
government of the president's party or coalition, the president must name a
prime minister to form an opposition government, as Frarn;ois Mitterrand
acceded to Jacques Chirac in France in 1986. Moreover, it is not required
that the president should dominate the prime minister - and so, indirectly,
the cabinet - under premier-presidentialism, even when he or she enjoys a
legislative majority. If the prime minister rather than the president is the
leader of the principal governing party, for example, then the president's
fixed term of office does not imply the potential for crises of temporal
rigidity. But even if presidents do dominate the executive, a potential
midterm loss of legislative support will not portend crisis if they name an
opposition cabinet that enjoys assembly support, and they are capable of
exercising the constitutional powers of the presidency while ceding control
over the cabinet to an opposition prime minister.
The second means of avoiding the problems of fixed terms in several
systems is to provide the president the authority to dissolve parliament and
call new elections. This provides the chief executive the opportunity to
appeal to the voters for a sympathetic assembly, and may be particularly
appropriate when a new president faces an opposition assembly elected at
an earlier date . 19

Majoritarianism
The problem of majoritarianism in presidential systems is mitigated under
premier-presidentialism primarily by the nature of the premier-presidential
18 In Austria and Iceland, extraordinary assembly majorities of two-thirds and three-fourths,
respectively, can call for plebiscites on the continuation in office of the president (Iceland,
Art. 11; Austria, Art. 6, Sec. 6). We know of no cases in which these provisions have been
exercised. In addition, in some premier-presidential systems (France, for example) presi­
dents may resign and call for presidential elections before their term is complete. Neverthe­
less, while these provisions could mitigate the problem of temporal rigidity found in pure
presidential systems, the more important characteristic of premier-presidentialism in this
respect is the provision for assembly confidence in the cabinet.
19 Of course, as Chapters 11 and 12 illustrate, this particular scenario can be avoided by
synchronizing assembly and executive terms.
 Criticisms ofpresidentialism and responses 51
cabinet. Because the cabinet is subject to parliamentary confidence, 20 it
will not be as narrowly representative of the president's interests as will a
presidential cabinet - unless of course, there is majority support in parlia­
ment for the president's narrow interests. But under such conditions, we
would expect cabinets in parliamentary regimes to be highly majoritarian
as well. Thus, if we can expect a given party to enjoy dominance in a
parliamentary cabinet, we can also expect it to enjoy dominance in a
premier-presidential cabinet.

Dual legitimacies
Premier-presidentialism replaces the problem of competing legitimacies
between the assembly and the executive with the potential problem of
competing legitimacies within the executive itself (Pierce 1990). We ad­
dress this dilemma in later chapters as we discuss the dangers of cohabita­
tion. For now we consider the advantage of premier-presidentialism in
providing an institutionalized means of resolving legislative deadlock be­
tween the president and assembly. The replacement of the president as
head of government with a prime minister with legislative confidence has
been discussed already. Only a couple more points have to be made on this
subject.
First, it is worth noting that even a majority president will have a greater
incentive under premier-presidentialism than under pure presidentialism to
anticipate and accommodate the demands of his or her coalition, since a
president's status within the executive depends so directly on the coalition's
cohesiveness (Suleiman 1981:98-103). This consideration may temper the
danger of an exaggerated sense of mandate, for which presidentialism has
been criticized.
Second, while political outsiders can still be elected president under
premier-presidentialism, the prime minister enjoys constitutional protec­
tion from dominance by the president. Thus, when the president is not the
leader of the prime minister's party, we expect the prime minister to be
able to act relatively independently. Provided that the responsibilities ac­
corded the heads of state and of government are clearly distinguished, an
outsider in the presidency should not present the problems associated with
such scenarios under presidentialism. The strict separation of powers is
reduced under premier-presidentialism by the requirement of parliamen-

20 Of course, cabinet nominees in the United States are subject to Senate approval as well.
Cabinets in premier-presidential systems are inarguably more subordinate to assembly
preferences than are those in the United States, since they are subject to collective rejec­
tion, which can be imposed ex-post as well as ex-ante and because the U.S. president
retains the right to dismiss ministers. See our discussion in Chapter 6.
52 Presidents and assemblies
tary confidence for the government. Among other things, this requirement
ensures that the head of government's level of political expertise will be
sufficient to satisfy the political insiders in parliament.

Mutual checks
If presidents were provided a veto, then their check against a legislative
majority would be no different under premier-presidentialism than under
presidentialism. However, as already noted, most premier-presidential
regimes do not provide for a veto that requires an extraordinary majority
to override. The relationship between premier-presidentialism and the
legislative check on the executive, however, is less clear. Certainly, the
requirement of parliamentary confidence in the government implies a
check of some sort. But we have noted that the same requirement con­
strains, rather than fosters, legislative independence from the executive
under pure parliamentarism. Do we have reason to expect that the major­
ity of legislators who support a government under premier-presiden­
tialism will be any more likely to vote their consciences to oppose cabinet
initiatives?
The answer must be a qualified no. No, because the logic of parliamen­
tary voting still applies under premier-presidentialism. Legislators in the
majority coalition still face a disincentive to jeopardize cabinet stability. We
qualify the answer because the disincentive may not be so great under
premier-presidentialism as under parliamentarism. If the president domi­
nates the prime minister, then the vote of no-confidence threatens this
position, but not the president's continuance as head of state. To the extent
that significant political authority is retained even by a minority president,
then independent opposition by legislators to government initiative does
not connote such high stakes as under parliamentary regimes.
Despite this conjecture , there is little evidence that legislators under
premier-presidentialism actually demonstrate more independence than
their parliamentary counterparts. During the first years of the Fifth Repub­
lic, de Gaulle relied on shifting majorities to build support for various
legislative proposals (Wood 1968:103-4; Pierce 1990). But by the mid-
1960s, as partisan divisions under the new regime became clear, so too did
discipline improve - at least among the main party in a given majority
coalition. Similarly, Kaminsky (1989:28-9) finds that Mitterrand relied on
shifting majorities on budget legislation in his second term, but the support
of Socialists was dependable. On the other hand, Suleiman (1980:103-4)
points out that Giscard d'Estang was abandoned by the Gaullists in 1976
when he attempted to pass a major reform of capital gains tax laws over
their objections. Legislators of parties peripheral to the majority coalition,
then, seem to be more inclined to vote independently. This is often true as
well under pure parliamentarism.
 Criticisms of presidentialism and responses 53

Arbiters
W hen presidents dominate a premier-presidential system, it is because they
enjoy majority support in the assembly. Under these conditions, it would
seem, the need for an arbiter to cultivate legislative support for executive
proposals ought to be minimal. Nevertheless, students of premier-presiden­
tialism have observed that when the president dominates the executive, the
prime minister becomes something of arbiter, or as Duverger (1980: 184)
describes it, chief of staff. Provided that the president and prime minister
share a similar political agenda, then the prime minister acts as the execu­
tive liaison to the assembly. Suleiman (1981: 121) points out that the prime
minister may also serve at times to shield the president from political harm.
If the executive's economic policies fail, for example, the president can
sacrifice the prime minister, provided the president can credibly stick the
PM with the blame.
W hile the prime minister may play the arbiter's role when the president
enjoys an assembly majority, the president can serve as arbiter of partisan
conflicts under divided government. When presidents face an opposition
government, or when they are not the party leader, they are necessarily
removed somewhat from the partisan politics of legislative coalition­
building. It is precisely this distance that can allow the president to arbi­
trate partisan disputes (Lijphart and Rogowski 1991).
First, and perhaps most important, the president has the prerogative of
naming the prime minister who will form a cabinet. The president-as­
arbiter can use his or her discretion in choosing among whichever candi­
dates could form viable governments to seek concessions from the future
prime minister, and possibly to ensure a broad representation of interests
in the cabinet. 21 In the second place, minority presidents can rely on what­
ever formal powers are granted them constitutionally to influence the out­
comes of legislative struggle, including (in some systems) referring govern­
ment legislation to judicial review or referendum. Finally, where presidents
are accorded the authority to dissolve the assembly, they may use this
leverage to pressure prime ministers to accept compromise with coalition
partners or face the electorate. As we shall see in the next chapter, this
means of presidential arbitration has been used effectively in premier­
presidential systems such as Finland and Portugal, but destructively in
president-parliamentary systems such as the Weimar Republic.

CONCLUSION

This chapter has weighed theoretically the advantages and disadvantages of

presidentialism and premier-presidentialism, based on their institutional
21 See Du verger (1980: 169-70) on the role of Austrian Socialist presidents in cabinet forma-
tion during the 1950s and 1960s.
54 Presidents and assemblies
characteristics. In particular, we have reviewed the central criticisms of
presidentialism made by proponents of parliamentarism, and have offered
a series of responses suggesting that the basic principles of presidential
government may be worth preserving. We argue that presidentialism offers
the advantage of allowing the clearest possible choice to voters as to the
constitution of the executive. The checks and balances inherent in presiden­
tialism are also desirable when majoritarian rule is opposed.
We have provided some arguments for why both presidentialism and
premier-presidentialism might be desirable as a format for representative
democracy, but we have yet to explore empirically the performance of these
regime types. This is the project we begin in Chapter 4, briefly considering
the performance of premier-presidential institutions in a number of coun­
tries, and contrasting these regimes with president-parliamentary regimes.
 4
The premier-presidential and
president-parliamentary experiences

INTRODU CTION

Before pr oceeding with our analysis of the relationship between presi­

dents and assemblies m ore generally, we n ow c onsider the perf ormance of
premier-presidential g overnment as well as a few systems that we deem
president-parliamentary. Such a task is necessary in order t o illustrate
m ore clearly the nature of the regime types, and their distincti on from
b oth presidentialism and parliamentarism. This is so particularly because
premier-presidentialism and president-parliamentarism have been dis­
cussed infrequently in the c omparative literature, and may be unfamiliar
to many readers. When students of dem ocratic instituti ons have sp oken of
premier-presidentialism (what others have called "semipresidentialism"),
they have rarely shared a specific, c ommon idea of its instituti onal struc­
ture. And, of c ourse, regimes that we define as president-parliamentary
have been described a number of ways: as presidential, semi-presidential,
parliamentary, and as hybrids. In Chapter 2, we pr ovided definiti ons that
we believe are exhaustive and generalizable f or all these regime types. We
begin here with a general discussi on of the th orniest instituti onal pr oblem
that is, in effect, built int o premier-presidential g overnment - the p ossibil­
ity of a divided executive.
Where the president and the cabinet are of opp osing parties or blocs,
premier-presidential g overnment faces a challenge s omewhat similar t o that
of presidentialism under divided g overnment. The specific tensions am ong
p olitical adversaries, the dangers t o regime stability, and the p otential means
of res olving th ose threats, h owever, are all different under premier­
presidential and president-parliamentary as opp osed t o presidential sys­
tems. The the oretical discussi on of divided executives will be foll owed by
examinati ons of premier-presidential and president-parliamentary regime
perf ormance. We begin by discussing s ome of the m ore well-kn own cases of
these regime types: France under the Fifth Republic, Finland, and P ortugal,
as well as Sri Lanka since 1979 and Weimar Germany. As we shall see,
although the latter tw o cases have been identified as premier-presidentialism
(semipresidentialism) elsewhere, a careful examinati on of the f ormal bal-
56 Presidents and assemblies
ance of powers, and the performance of some of these regimes, suggests that
they more closely approximate our definition of president-parliamentary. In
the case of Sri Lanka, this is so despite the influence of France's system on
the designers of the Sri Lankan constitution.
We then proceed to consider other cases in which independently elected
presidents coexist with cabinets dependent on assembly confidence. We
shall see that such an arrangement does not necessarily produce the direc­
tion of the cabinet by the assembly. Finally, we consider some new premier­
presidential regimes, and a new proposal for premier-presidential govern­
ment, in which the performance of the institutions cannot yet be assessed.
Many of these brief case studies underscore the danger of institutional
arrangements that do not provide clear mechanisms for the resolution of
conflict between presidents and opposition assemblies. But the cases also
illustrate the potential advantages of premier-presidential systems in some
of the areas in which we have suggested the regime type might differ from
both presidentialism and parliamentarism.

T H E P E R I L S OF C O H A B I T A T I O N

For any of the potential advantages of premier-presidentialism to hold, two

assumptions must hold: first, that the system will provide for a clear divi­
sion of presidential from prime ministerial responsibilities; and second,
that this division will be respected on both sides, instead of producing
competitive dyarchy. That the second condition should be met is by no
means self-evident, even where the constitution is textually clear on execu­
tive responsibilities and the division is clear and mutually exclusive. Unfor­
tunately, the ambiguity of some constitutions has been a common stum­
bling block to cooperation within the executive. Judging from three
president-parliamentary regimes that have been less than satisfactory -
those in Portugal, Sri Lanka, and the Weimar Republic - it appears that a
primary challenge of constitutional design must be to establish a clear
division between the authorities of head of state and head of government,
and to make clear within the constitution that the formal distinction be­
tween the two roles will be recognized upon the loss of the presidential
majority in the assembly. In the case studies that follow, we shall see more
clearly the problems that constitutional ambiguity, or disproportionate con­
stitutional powers, can cause.
Before examining the cases, however, we must consider the question of
whether constitutional divisions of authority will be respected when a presi­
dent and prime minister are of opposing political tendencies - a scenario
dubbed "cohabitation" to describe the French experience of 1986-8. The
fundamental problem of the "executive divided against itself" (Pierce 1990)
would parallel the problem of dual legitimacies under pure presidentialism,
as described in Chapter 3. In premier-presidentialism, the executive could
 Premier-presidentialism, president-parliamentarism 57
be divided into opposing segments - the president on the one hand, and a
cabinet with assembly support on the other - elected with separate man­
dates by distinct constituencies. Where either or both segments fail to
recognize the claims to executive authority made by the other, cohabitation
could generate regime crisis. As we shall see, the risk is especially great in
president-parliamentary systems.
We can imagine two potential scenarios of conflict under cohabitation.
First, the cabinet may refuse to accept the legislative powers of a president,
despite the fact that the president may be provided such powers in the
constitution. 1 Second, a president may refuse to acknowledge the claims to
executive leadership made by an opposition assembly majority. If this is the
case, the president might use her authority over government formation to
deny the prime ministership, or even cabinet membership, to a key party or
parties. Or she might simply ignore cabinet and assembly opposition and
attempt to legislate over the head of the assembly, if constitutional means
exist to do so. 2
Pierce (1990) has argued convincingly that where a president and an
assembly majority each have sufficient constitutional authority to challenge
the other's direction of the executive, neither can be expected to defer to
the other in the absence of quite unusual circumstances. Regarding the
French experience of 1986-8, Pierce points both to consensus between
President Mitterrand and Prime Minister Chirac on important policy dimen­
sions, and to the short prospective time horizon for cohabitation and the
electoral incentives of both sides. We believe that Pierce's second point is
generalizable; that is, that the relative timing of presidential and assembly
elections is a critical determinant not only of whether presidents and assem­
bly majorities will have incentives to make the concessions necessary to co­
inhabit the executive, but also whether the conditions of cohabitation are
realized to begin with.
Because of the requirement of confidence in the cabinet, assembly elec­
tions in premier-presidential systems will always be contested at least par­
tially as judgments on executive performance. Therefore, if either the
assembly or the president can claim a more recent electoral mandate, that

1 This scenario may be especially likely where the president is not a leader of any major party
in the assembly, and so cannot provide support in the form of votes for any cabinet or
legislative program. This was the situation in Germany during Paul von Hindenburg's final
year as president, and in Portugal under the constitution of 1976, before the curtailment of
presidential powers in 1982. Indeed, we shall see below that the conflict between President
Ramalho Eanes and a series of opposition prime ministers fueled assembly support for the
constitutional changes.
2 Again, we are thinking of here of Weimar, but in the years preceding Hitler's rise to the
chancellorship. Also relevant, while less ominous, are President de Gaulle's uses of refer­
enda in the 1960s on proposals for which he lacked assembly support. As we shall see below.
de Gaulle's extra-constitutional actions could well have prompted regime crises had he been
challenged, or had he failed to resign upon the rejection of his last referendum.
58 Presidents and assemblies
mandate may be interpreted as a valid claim to exclusive control of the
executive. 3 This being the case, the scenarios for crises of cohabitation
described above will be most likely either where:
1 . a new assembly is elected with a majority in opposition to the incumbent presi­
dent; or
2. a newly elected president faces a sitting cabinet composed of the president's
principal opposition .
In the second instance, if the presidents are allowed to dissolve the
assembly, they can do so as a means of petitioning the electorate for assem­
bly support. In the first case, assembly elections must be interpreted as a
repudiation of cabinet performance. To the extent that the president exer­
cises influence over the cabinet (as leader of the previous majority party,
for instance), then, it is a repudiation of this leadership.
The potential scenarios for crises of cohabitation just described will be
most likely when elections for one branch regularly occur in the midst of
the other branch's term. That is, where presidential and assembly elections
are concurrent or held within a short time of each other, we expect the
assembly majority to be compatible with (or at least not diametrically
opposed to) the president; but the likelihood of compatibility declines and
the likelihood of sharp division increases with midterm elections. We elabo­
rate on electoral cycles under premier-presidentialism in Chapter 12. For
now, the following points are in order. First, cohabitation presents distinct
dangers for premier-presidential regimes. Second, the most likely source of
cohabitation crisis is the claim by either the president or the assembly
majority that a recent election imparts a more legitimate mandate for
exclusive control of the executive than that of the opposing actor. Third,
although cohabitation is possible under any electoral cycle, it is more likely
under some formats than others.

THE CASES

France
In the French case, the two phenomena we consider - the use of referenda
on legislation, and the successful experience of cohabitation - demonstrate
a certain irony. One of the principal dangers associated with premier­
presidential government is that any constitutional ambiguities are likely to
foster conflicting claims to executive leadership when presidents and assem­
bly majorities oppose each other. Regarding the referendum, the letter of
the constitution is quite clear, but was disregarded by French presidents.
On the question of the division of responsibilities under cohabitation, on
3 The same claim could not be made under pure presidentialism, where elections in one
branch do not directly affect the composition of the other.
 Premier-presidentialism, president-parliamentarism 59
the other hand, some ambiguity exists, yet the two-year period produced
no crises or dangerous confrontations.
We have already suggested that where presidents do not recognize the
claim to executive leadership of the assembly majority, they may attempt to
avoid conceding leadership of the executive by exercising extraordinary
legislative authorities. W here presidents have the authority to call refer­
enda, this would be a means of bypassing assembly opposition. Even a
minority president might use the device selectively, putting together single­
issue majorities on issues that cross-cut partisan cleavages, to retain some
control over the legislative agenda.
Of course, the power to call referenda provided the French president in
Article 11 of the constitution is not unilateral. The referendum must be
proposed as well by the government or by a joint motion of the two assem­
bly chambers. Moreover, the issues which may be subject to referenda are
limited to those which will affect the organization and functioning of exist­
ing political institutions. 4 However, the referendum has been used in the
Fifth Republic without reference to these checks and limitations. President
Charles de Gaulle, who refused to recognize the assemblies as the presi­
dent's equal in legitimacy as representative of the nation, first invoked the
referendum in his direct appeal to the people, without assembly or cabinet
approval, for the authority to set up a provisional executive in Algeria in
1961. The Constitutional Council did not challenge the move. De Gaulle
used the referendum five more times in the next eight years, and it was the
rejection of his sixth attempt by the electorate that prompted him to resign
in 1969. Following de Gaulle's precedent, President Georges Pompidou
also used the referendum unilaterally in 1972, to propose British member­
ship in the Common Market. The referendum barely passed, and the reac­
tion to Pompidou's use of the device without cabinet or assembly authoriza­
tion was lukewarm at best. He did not use the referendum again (Wilson
1979).
The ability of de Gaulle and Pompidou to ignore the letter of the French
constitution underscores the limitations of formal institutional arrange­
ments as solutions to political conflict. Rather than being interpreted as
evidence that the formal rules of political contestation are meaningless,
however, these cases ought to remind us of the bounds of this type of study.
It remains true that politicians opposed to the president's actions could
have used constitutional means to challenge them, had they anticipated
political advantage to doing so. The experience of cohabitation, on the
other hand, reinforces the potential of institutional analysis. Where condi­
tions were favorable, even a somewhat ambiguous set of constitutional
provisions for the division of executive powers served to guide President
Mitterrand and Prime Minister Chirac through two years of cohabitation
4 Constitution of the Fifth Republic, Title II, Article 11.
60 Presidents and assemblies
without incident, until compatibility within the executive was restored in
1988.
For the first three decades of the Fifth Republic, there was some skepti­
cism that the system could adapt to cohabitation without competitive
dyarchy and possibly regime crisis as a result. The 1986-8 period produced
no great crisis within the executive, however, even while Mitterrand re­
tained leadership over substantive policy areas. Article 5 of the constitution
states that the president "shall be the guarantor of national independence,
of the integrity of the territory, and of respect for Community agreements
and treaties," while Article 20 determines that "the Government shall
determine and direct the policy of the nation." These two provisions have
generally been interpreted as reserving the domain of foreign and security
policy initiative for the president, although it is by no means inconceivable
that an opposition prime minister with a recent mandate would challenge
the president's authority to determine policy of any kind. During cohabita­
tion, it was to this interpretation of the division of responsibilities that
Mitterrand and Chirac referred when each made claims for the scope of his
authority. We acknowledge Pierce's (1990) point that a general consensus
on security issues, as well as the electoral cycle, created incentives for each
man to make concessions to the other. Yet it is important to recognize that
even a less than perfect constitutional division of authorities provided a
point of reference on which a mutually acceptable interpretation was
agreed.
The annual summits of the seven major industrialized democracies pre­
sent good examples of how the constitutional division of responsibilities
was implemented, since the summit agenda include both trade and security
issues. According to the interpretation just described, we should expect the
president to represent the country's interests alone on international secu­
rity, while the division of authority on trade will not be as clear. Trade
issues, it seems, could be classified as both foreign and domestic policy.
The arrangement adhered to by Mitterrand and Chirac at the summits fits
precisely with this complex schema.
At the 1986 Tokyo summit, Mitterrand arrived a day before Chirac, to
attend the formal opening dinner - the symbolic head of state representing
France. W ithout Chirac, he also attended meetings the following morning
on terrorism and security. The prime minister's arrival was carefully
choreographed so that he could join the meetings by Mitterrand's side just
as they turned to trade matters. 5 Likewise, the following year at Venice,
Chirac deferred on the symbolic and security portions of the summit, but
participated in the trade negotiations. 6 In the trade negotiations them­
selves, of course, the unified position of the president and prime minister
5 "French Power-Sharing Alters Summit Protocol," International Herald Tribune, May 5,
1986, p. 6.
6 Cover photo, International Herald Tribune, June 10, p. 1 ; June 12, p. 4.
 Premier-presidentialism, president-parliamentarism 61
had to be carefully negotiated before the summits began. Nevertheless, the
two men succeeded in speaking for France "with one voice, but through
two mouths. "7
The example of French cohabitation does not furnish conclusive proof
that a minority president with a significant reserve domain of powers will
necessarily function effectively with an opposition government. But it does
demonstrate that such an arrangement is feasible under premier-presiden­
tialism, and that the constitutional division of responsibilities within the
executive can provide a salient point of reference for politicians who ac­
knowledge the rules of the game (see Schelling, 1960). Where such divisions
have been less clear than in France - and especially in the president­
parliamentary systems - regime performance has been much more problem­
atic. We shall see this more clearly with Portugal, Sri Lanka, and Weimar
Germany. Before proceeding to those cases, however, it will be worthwhile
to examine another successful premier-presidential system: Finland.

Finland
Nousiainen (1988:232-3) has argued that the Finnish constitution does not
make entirely clear the division of authority between the president and the
government (Council of State). This assessment is most likely because
Finnish presidents have consistently played an arbiter's role, rather than
directing the cabinet through the prime minister, regardless of the cabinet's
partisan composition. 8 Actually, the interaction between the president and
the Council of State is not entirely clear-cut. The president does indeed
chair weekly council meetings, but the only subjects discussed at those
meetings are those which fall under his reserve domain. He does not attend
the cabinet meetings chaired by the prime minister, in which most legisla­
tive and administrative matters are handled. As Arter (1985:480-1) de­
scribes it, "The Prime Minister leads the government and there has been no
attempt at a 'divide and rule' strategy, with the President cutting out the
Prime Minister by sending messages to ministers." The lack of clarity
Nousiainen notes in the division of responsibilities, then, has not generated
crises in the actual functioning of the Finnish system. This is due in large
part to the clarity with which the president's reserve domains - in foreign
policy and the formation of governments - are established, and the relative
importance of these reserve domains to Finnish politics.
Article 33 of the Finnish constitution establishes unequivocally that "the
President shall determine the relations of Finland with foreign powers."
W hile treaties and declarations of war must be approved by parliament,
foreign policy initiative clearly rests with the president. Moreover, students
7 "Two French Leaders Take Delicate Duet to Tokyo," New York Times, May 5 , 1986, p. A7.
8 Finnish constitution, see Articles 2, 39.
62 Presidents and assemblies
and participants in Finnish politics agree that the overwhelming impor­
tance of foreign affairs in postwar Finland - in particular, of sustaining
amicable relations with the Soviet Union while retaining as much indepen­
dence as possible - has produced a system in which the president's role is
enormously important, even apart from the workings of the parliamentary
government (see Arter 1981:221 ; Arter 1985; Nousiainen 1988). As ex­
President J. K. Paasikivi (1946-55) said, on being asked what his job en­
tailed, "I sit, read and ponder over how to keep this nation indepen­
dent . . . one mistake might well prove fatal" (Arter 1981:222).
The centrality of foreign affairs to Finnish politics is one reason why
parties have delegated to the president, especially during crises, the ability
to act outside his reserve powers. The party-based electoral college (see
Chapter 10) is another reason; we also discuss the implications of the
method of presidential election in Finland further in Chapter 12. Examples
of presidential actions beyond those specifically authorized in the constitu­
tion include President Urho Kekkonen's intervention in 1970 in income
policy negotiations, which had bogged down between the government and
the nation's largest labor federation. W hen talks collapsed, the president
presented his own alternative settlement in a speech broadcast on televi­
sion and radio. His intervention prompted renewal of the negotiations, and
an eventual settlement. In another instance, Kekkonen applied pressure on
successive governments deadlocked on farm incomes policy and pensions
policy to overcome intracoalitional divisions, or resign. Following the
back-to-back resignations, Kekkonen then explicitly demanded conces­
sions on the part of the Centre Party, stating that its future claims to govern
would otherwise be jeopardized (Arter 1981). In this instance, Kekkonen
was extending his constitutional power to appoint the members of the
Council of State9 to influence policymaking outside the realm of presiden­
tial authority. It was Kekkonen's ability to take such actions without
prompting serious resistance that induced Arter (1981) to describe the
Finnish system as "president-dominant and president-reliant" in times of
consensualism in crisis.
Studies of the postwar Finnish presidency have until recently been tanta­
mount to studies of the Kekkonen presidency, 1955-81. The most vocal
critics of Kekkonen's exercise of powers beyond constitutional limits under­
standably were members of the Conservative Party, which Kekkonen ex­
cluded from every government he named after 1966. But both Arter (1985)
and Nogueira (1986a) argue that in a multipolar party system such as
Finland's, presidential prerogative as builder of governments will necessar­
ily be greater than in a more bipolar system, such as France's.
Finally, while Kekkonen wielded power to steer even specific domestic
policies under certain conditions, the end he consistently pursued was the

9 Finnish constitution, Article 36.

 Premier-presidentialism, president-parliamentarism 63
maintenance of relatively broad coalitions in government - which is, after
all, well within the realm of the president-as-arbiter described above.
While his exclusion of the Conservatives shows that Kekkonen was not
strictly a nonpartisan arbiter, his role in arbitration among a wide spectrum
of parties is consistent with his having been regularly endorsed by broad
multiparty coalitions in the Finnish electoral college. He pursued his
government-building actions without demonstrating outright bias toward
the party (Agrarian, later renamed Centre) for which he ran. His manipula­
tion of presidential prerogative as government builder aimed at maintain­
ing the Red-Green alliance between the urban left and center-left, and
rural interests, in Finnish coalition governments to the greatest degree
possible (Arter 1981:231; Anckar 1984:135; Arter 1985:480). In Chapter
12, we say more about the role of the Finnish electoral college in relations
between parties and the president.

Portugal
Although it was regularly described as premier-presidential (semipresiden­
tial) (Duverger 1980; Opello 1985; Bruneau and Macleod 1986; Nogueira
1986a), Portugal's 1976 constitution suffered from precisely the lack of
clarity between presidential and prime ministerial responsibilities de­
scribed in Chapter 2 as president-parliamentary. Competitive dyarchy
within the executive was exacerbated, moreover, by the existence of an
"outsider" president.
Portugal has historically had a somewhat confused balance of power be­
tween head of state and head of government. While the constitution of the
Estado Novo vested the nation's president with broad authority, Premier
Antonio de Oliveira Salazar in fact wielded authoritarian power from 1930
to 1968 (Bruneau and Macleod 1986:128). But the nondemocratic Estado
Novo is not of too much interest here. Portugal's experimentation with
democratically elected presidents began with the constitution promulgated
in 1976.
Under this constitution, a popularly elected president would serve along­
side a prime minister, named by the president and subject to both presi­
dential and parliamentary confidence, who directed the government. In
addition to his powers to form and dismiss governments and dissolve the
assembly, the president was given control over national defense, through
his chairmanship of the Council of the Revolution; and he was given a de
facto pocket veto over all legislation, since his signature was required for
promulgation of any law and could not be impelled by the government or
parliament. At the same time, the president was not given authority to
preside over the cabinet (Opello 1985:149). And while his chair on the
Council of Revolution assured him control over defense policy, the consti­
tution reserved initiative on all nonmilitary foreign affairs for the govern-
64 Presidents and assemblies
ment. 10 This arrangement seems to ensure that control over the executive
would not be secured by either the president or the prime minister, a
point supported by a theoretical discussion to be presented in Chapter 6.
Moreover, the specific mixture of authorities does not suggest any clear
division of authority, either by type of power or by policy area.
The ambiguous position of the president was exaggerated by his nonparti­
san status during Portugal's early years under the new democracy. Portu­
gal's first president under the 1976 constitution was General Ramalho
Eanes, who had been instrumental in thwarting a leftist military coup
during the transition from the Estado Novo to the new constitution. Al­
though Eanes enjoyed widespread popularity, he had been elected without
party affiliation. 1 1 His relationship to the government, then, could be de­
fined solely in institutional terms - by the constitution.
W ith an outsider president, a cross-partisan alliance among parliamen­
tarians opposed to presidential influence over the policy process seems
natural. After all, the common interests were institutionally defined: to
reduce the powers of the president and increase those of the government,
made up of assembly members and supported by an assembly majority. At
the same time, Eanes had no partisan defenders in the assembly. Thus,
when the center parties began to call for constitutional reform because of
objections to the socialist economic goals elaborated in the 1976 constitu­
tion, they were able to enlist the support of the Socialist Party by making
presidential power the central issue (Bruneau and Macleod 1986: 121-2).
W ith a two-thirds majority in the assembly supporting the revision, Presi­
dent Eanes was unable to block the adoption of a new constitution in 1982.
As promised, a number of presidential powers were severely curtailed, in­
cluding the president's power to dismiss cabinets or ministers, thus moving
the current Portuguese system down along the vertical axis in Figure 2. 1, into
the premier-presidential area in the lower left quadrant. While parliament
can still remove the government for political reasons, the president can do so
only when the functioning of the nation's democratic institutions is threat­
ened. 12 This restriction permits the president to invent "threats" to democ­
racy, but it forces the president to provide extraordinary justification, which
was not required under the 1976 constitution or in other cases that we
identify as president-parliamentary. Given the existence of a Constitutional
Court, a president's use of this power in dubious circumstances could be
challenged and possibly overturned. The new constitution also restricts the

10 Constitution of 1976, Article 138.

1 1 Portugal's only requirement for inclusion on the presidential ballot is the submission of
between 7,500 and 15,000 signatures to the Constitutional Court at least a month before
the national election (Art. 127). This minimal requirement, it would seem, should encour­
age the "political outsiders" described by Blonde! and Suarez (1981) to throw their hats in
the ring.
12 Constitution of 1982, Article 198.
 Premier-presidentialism, president-parliamentarism 65
president's power to dissolve parliament: not within six months of its elec­
tion, nor within six months of the end of the presidential term. Furthermore
the document suspends the president's pocket veto. It now requires promul­
gation or veto within twenty days of passage, but the veto may be overridden
by an absolute majority of parliament except for foreign policy and defense
legislation and laws pertaining to the Constitutional Court or elections.
Vetoes to the latter types of bills require a two-thirds vote to override
(Opello 1985:155). Moreover, the Council of the Revolution is eliminated,
and the president is given chairmanship of the Council of State. But the new
council is a purely advisory body; and the authority over defense policy that
had been vested in the Council of the Revolution is transferred to the govern­
ment and the new Superior Council for National Defense. Finally, the power
to suspend regional governments is also transferred from the president to the
government.
Eanes complained that the new document not only reduced his powers but
also made it more difficult to tell who was responsible for what (Bruneau and
Macleod 1986:129). In fact, while Eanes's dissatisfaction is understandable,
the 1982 constitution is more clear than that of 1976 on the relative powers of
president and cabinet. It unequivocally reduces the powers of the presi­
dent - virtually eliminating legislative powers - and correspondingly in­
creases those of the cabinet. This is not to suggest that the new presidency is
completely ineffectual. For example, even in 1985, Eanes was able to use a
threat to dissolve parliament to compel Prime Minister Mario Soares to
refrain from perpetual cabinet shuffling, and make concessions to the mem­
bers of his coalition government (Bruneau and Macleod 1986:141).
Nevertheless, the powers of the Portuguese president are significantly
less than they were previously. The difference since 1982 is reduced law­
making power and the denial of a reserve domain in policy initiative. The
change should not have been surprising, given the combination of a confus­
ing division of responsibilities under the 1976 constitution, and an outsider
president.

Sri Lanka
If Portugal from 1976 to 1982 does not fit our definition of premier­
presidentialism, the Sri Lankan system fits even less. The reason is not
dyarchy, however, but complete presidential dominance over the rest of the
executive, guaranteed in the constitution itself.
Article 30 states unambiguously that the president is the head of state, of
the executive, and of the government, without qualification. Moreover, in
Article 44, the presidents are given unlimited power to appoint all minis­
ters and make their assignments; and presidents may retain personal con­
trol over any ministry or ministries, as they see fit. As a result, even a vote
of no-confidence against the government in parliament would not affect the
66 Presidents and assemblies
president's standing in the executive (Zafrullah 1981:39). The government
would be removed, and the president would be required to name a prime
minister who enjoys parliamentary confidence, but even within an opposi­
tion government the prime minister would remain entirely subordinate to
the president.
Because of the presidential dominance inherent in the fundamental rules
of the game, there is no confidence among students of Sri Lankan politics
that the system could respond without a serious crisis to the election of a
legislative majority hostile to the president (see Warnapala, 1980; Wilson
1980, chs. 4, 5; and Warnapala and Hewagama, 1983). Three scenarios are
possible. First, the president might acquiesce completely to the parliamen­
tary opposition, naming an acceptable prime minister, consulting and defer­
ring to the prime minister on other ministry appointments, and directing
the government according to the demands of the opposition. Given that the
president has clear and viable alternatives to acquiescence, there is no
reason to expect such a result. The second scenario would be a hostile
confrontation with parliament, with the president retaining a minority cabi­
net. In the event that such a confrontation escalates to a crisis point, the
president can unilaterally declare a state of emergency (Perera 1979:37).
But the state of emergency must be approved by parliament after fourteen
days, so we can imagine that this scenario would ultimately yield a severe
crisis of legitimacies. In the third scenario, the president would name an
opposition prime minister, and possibly a cabinet with substantial opposi­
tion representation, but he would proceed to legislate in direct competition
with the government and parliament through the use of referenda.
The authority to invoke referenda granted to the president by the Sri
Lankan constitution is far broader than that in the Fifth Republic of
France. The president may submit at her or his discretion any bill that has
been rejected by parliament. 13 Coupled with the president's authority to
initiate legislation, it is clear that the power of the referendum can allow
even a minority president to bypass parliament on legislation of any type.
A wily chief executive will be able to appeal directly to the people on any
issue for which he or she can engineer one-time popular support. The effect
is to put in the hands of a majoritarian official the most majoritarian of
policymaking devices - at the expense of parliament's role in legislation,
and of the control of the government of the assembly.
President J. R. Jayewardene's use of the referendum to extend the life of
parliament in 1982 presents a fascinating case that illustrates at once the
president's fear of a hostile parliament and the danger of his unrestrained
recourse to direct democracy. The sequence of events is somewhat com­
plex. In the first place, the Sri Lankan parliament is elected for a full six
years, both under the constitution of 1972-7 and under the current constitu-

13 Constitution of Sri Lanka, Section 85(3).

 Premier-presidentialism, president-parliamentarism 67
tion of 1978. Thus, the parliament elected in 1977, which had approved the
new constitution and had originally named Jayewardene president before
the days of popular presidential elections, remained in place in 1982.
Moreover, Jayewardene's party, the United National Party (UNP) held
more than 80% of the seats in parliament despite having gained little more
than 50% of the aggregate 1977 vote, owing to the distortions inherent in
Sri Lanka's former one-seat district plurality electoral formula. The old
plurality formula had historically caused enormous swings in parliamentary
representation, and as part of the constitutional revision, the UNP had
implemented a proportional representation system. The party leadership
evidently believed that its broad support would ensure it a majority, or near
majority, consistently under PR; whereas under plurality, even minor fluc­
tuations in support could have drastic repercussions in terms of seats.
But in 1982, the UNP did not want to subject its majority to a test. In the
country's first presidential elections under its new constitution in October
of that very year, Jayewardene had won a full six-year term. But judging
from Jayewardene's mandate, the UNP did not expect to receive a two­
thirds parliamentary majority in PR elections. The president would likely
be constrained in pursuing his agenda of further amendments to the new
constitution. The near total domination the UNP had enjoyed since 1977
seemed preferable. So Jayewardene simply presented a referendum to the
people in December 1982, proposing the extension of the 1977 parliament
for another full term, beginning in 1983, when its current term ended. As
expected, the referendum received bare majority support, and Jayewar­
dene retained a docile parliament (see Warnapala and Hewagama 1983,
esp. p. xiv and ch. 7).
Sri Lanka's dramatic ethnic violence during the late 1980s and early
1990s has overshadowed the less than satisfactory performance of its gov­
ernment institutions in mediating political contestation. Nevertheless,
given the disincentives to compromise built into the institutions, and
Jayewardene's ability to avoid coalition building, it seems likely that the
1979 constitution has contributed to the transfer of conflict from the institu­
tions to the countryside. As with the Portuguese case before 1982, how­
ever, it is important to recognize the ways in which the constitutive rules of
the game have been contradictory from the outset. W here an assembly
majority supports the president, the president will dominate the executive,
much as in France. But where an assembly majority opposes the president,
the president's constitutional authorities both allow and encourage him or
her to ignore the assembly. One of the principal means by which the presi­
dent can end-run, or even overrun, potential assembly opposition is
through the exercise of unrestrained referendum power. The dominance of
the Sri Lankan president, based on formal powers, precludes the system
from meeting the description of premier-presidentialism established in
Chapter 2.
68 Presidents and assemblies

The Weimar Republic

No review of the performance of premier-presidential regimes would be
complete without consideration of Germany's Weimar Republic of 1919-
33, although a close examination of the Weimar experience demonstrates
that it, too, is a case of president-parliamentarism rather than premier­
presidentialism. The failure of the Weimar regime, of course, carried pro­
found implications. The specific regime type may be insufficient to account
for the totalitarian regime and world war that followed it. And we do not
claim that some other type of regime, such as true premier-presidentialism,
would necessarily have prevented the rise of Nazism, although the frequent
cabinet crises, dissolutions, and use of presidential emergency powers set
the stage for the appointment of Adolf Hitler as chancellor (premier) even
though he had failed three times in bids for the presidency. Our purpose
here is simply to examine the Weimar institutions and their relationship to
the destruction of democracy in Germany in the 1930s.
The Weimar president retained considerable legislative powers and the
right to dismiss cabinet ministers in such a way as to thwart the possibility
that ultimate authority over the direction of the government could rest with
the assembly majority. Thus, the principal problem with the Weimar consti­
tution, as we see it, was not so much any ambiguity over the relative status
of the president and chancellor (as in Portugal, and perhaps Finland) as the
constitutional dominance of the president, despite the requirement for
parliamentary confidence in ministers. This dominance rendered the Reich­
stag helpless to control the government, and ultimately made the govern­
ment and chancellor mere instruments of the president. 14
The Weimar president was elected for a term of seven years, by a two­
round majority-plurality system. In lieu of a first round majority, a runoff
was held among the top three candidates, at which a plurality would suffice.
The president alone named and removed the chancellor, and on the chan­
cellor's advice named the government. It was not required that the chancel­
lor be a member of the Reichstag. The Reichstag could remove the govern­
ment, or any minister, by a vote of no-confidence, although no vote of
investiture was necessary for the president to form a government. More
important, the president had virtually unrestrained authority to dissolve
the Reichstag.
The above provisions stacked the constitutional powers of both legislation
and government formation heavily in the president's favor. Although the
14 This description, of course, does not fit the situation after Adolf Hitler's appointment as
chancellor in January of 1933. But by that point, the entire political order in Germany
was rapidly disintegrating. We are concerned here with the function (or dysfunction) of
the Weimar regime leading up to the breakdown of formal democratic procedure. Hit­
ler's final destruction of the Weimar Republic is beyond the scope of this institutional
analysis.
 Premier-presidentialism, president-parliamentarism 69
Reichstag could reject his ministers and whole governments, the president
needed no parliamentary approval to replace censured cabinet positions.
Moreover, a troublesome Reichstag could simply be dissolved, subject to the
weak limitation that it not be dissolved more than once "for any one reason"
(Art. 25). Only the most unimaginative president could not conjure up
unlimited "reasons" for dissolution. Indeed, as conflict grew more intense
and chronic between president and Reichstag in the 1920s and early 1930s,
President Paul von Hindenburg relied repeatedly on dissolution.
In addition, the Weimar president had the authority to refer any law
passed by the Reichstag to popular referendum. While there was no presi­
dential veto, then, this provision prevented the assembly from securing
sovereignty over legislation, and made the president the gatekeeper over
such sovereignty. Finally, the president could issue decrees with the force of
law "in states of emergency to restore order, with the endorsement of the
chancellor" (Lepsius 1978). The endorsement of the chancellor really pro­
vided little constraint, given the chancellor's responsibility to the president.
And given the power of dissolution, even the Reichstag's formal authority
to revoke decree laws proved ineffective. The events of the early 1930s
demonstrated the dominance of the president in the Weimar system, and
the disincentives for compromise and coalition-building that this domi­
nance implied.
It is worthwhile to note here the theoretical foundations on which the
Weimar constitution was based. The document was written, during the
transition to elected government in 1918, by the legal and political theorist
Hugo Preuss, on the counsel of two other eminent social scientists: Robert
Redslob and Max Weber. Preuss accepted Weber's proposals for the appro­
priate term, means of election, and powers of the presidency, and Red­
slob's theory of the proper balance of powers between assembly and execu­
tive. Redslob sought to ensure that the Weimar system was not dominated
by the assembly, as he felt the French Third Republic was, and so wanted to
vest in the independent executive the ability to dissolve the Reichstag. His
search for "balance" of power between the branches, however, failed to
reckon with the absolute lack of reciprocal responsibility on the part of the
president before the Reichstag. 1 5
In the first years of the Weimar Republic, Presidents Friedrich Ebert and
Paul von Hindenburg tried to form governments with parliamentary sup­
port. The limitations on the use of presidential decree powers, however,
were never entirely clear, despite the constitutional provision that they be
used to "restore order" (Art. 48). Ebert used the decree both in the face of
revolts against the new government and to implement economic policies
(Lespius 1978:47). In the latter years of Weimar, however, the use of
15 For more background on the Weimar constitution, see Neumann, 1942; Needler, 1959;
Weber, 1968; Kircheimer, 1969.
70 Presidents and assemblies
decrees expanded tremendously, and efforts to base government formation
on parliamentary support ended.
The fragmentation of the Weimar party system made forming and sustain­
ing parliamentary coalitions exceedingly difficult. In the first place, the
Reichstag was elected by proportional representation out of a single, na­
tionwide district (see Taagepera and Shugart, 1989). Such an arrangement,
coupled with the ease of admission to the assembly, generated a prolifera­
tion of tiny, special-interest parties, and decreased incentives for coalition
building. In addition, the majority runoff system of electing the president
failed to provide any incentive for coalition building around the competi­
tion for control of the presidency. Together, these two institutional charac­
teristics complicated considerably the project of maintaining a government
with support in the Reichstag.
As the economic crisis escalated in Germany with the onset of the Great
Depression, and the threat to the political system grew from antidemo­
cratic movements - primarily the Nazis - President Hindenburg began to
rely more heavily on his formal powers to create governments, and to forgo
efforts at building parliamentary support. On numerous occasions, he pre­
empted or responded to opposition in the Reichstag by exercising his
power of dissolution. Beginning in March 1930, Hindenburg named Hein­
rich Bruning to head a "presidial" cabinet - one based only on the support
of the president. Hindenburg proceeded in July of that year to decree the
national budget. W hen the budget decree was revoked in the Reichstag,
the Reichstag was dissolved. This event was a benchmark in executive­
assembly relations during the Weimar era. Lepsius (1978) provides the
following data on the explosion of presidential decree activity, and the
corresponding decline of assembly activity in the years 1930-2:
Laws passed Presidential Days in session,
by parliament decrees Reichstag
1930 98 5 94
1931 34 44 41
1932 5 66 13
W hile such figures are by nature impressionistic, and do not illustrate the
content of the various laws, the pattern is nevertheless impressive.
By 1932, the Nazis had become the largest political party in Germany,
and the threat to institutional stability had grown correspondingly. In an
effort to capture support from Hitler on the authoritarian right, Hinden­
burg named a new government headed by Franz von Papen in June. To
preempt a certain parliamentary vote of no-confidence, the president imme­
diately dissolved the Reichstag (Lepsius 1978:48). The new Reichstag
elected at the end of July was also subsequently dissolved in September.
Following the November election of yet another Reichstag, Hindenburg
 Premier-presidentialism, president-parliamentarism 71
proceeded to form a new government in December, this headed by yet
another non-Nazi authoritarian, General Kurt von Schleicher. Finally, con­
fronted with the prospect of civil war and the explosive growth of Nazi
protest and activism, Hindenburg named Hitler chancellor in January 1933.
Two days later, the Reichstag was dissolved again, to preempt an expected
no-confidence vote. In the elections that followed in March, the Nazis
secured 44% of the vote, and in coalition with an antidemocratic partner,
put together a 52% parliamentary majority for Hitler's chancellorship. The
Reichstag subsequently suspended the constitution to allow Hitler absolute
authority to impose his program.
Of course, this account of the institutional maneuverings of Hindenburg
and Hitler cannot provide a full explanation for why Weimar could not
survive. But it should be clear from these events and from the institutional
qualities of the system, that the formal political rules in Weimar encour­
aged polarization and political conflict, rather than compromise. Even at a
time when the Nazi rise to power was far from inevitable, and while politics
conducted within the constitutional framework was predominant, the posi­
tion of the Weimar president rendered the Reichstag ineffectual. Needler
points out that it was formal powers, not the existence of procedural "loop­
holes" in the Weimar constitution, that "allowed men of ill will to fashion
presidia! cabinets" (Needler 1959:697). W hat is more, the chronic instabil­
ity of the early 1930s provided Hitler the perfect opportunity to blame the
current system for Germany's economic and political problems, and sub­
stantiated his claims that only through outright rejection of the democratic
process, and under his leadership, could Germany prosper. The Sri Lankan
and Weimar cases both demonstrate the important point that the formal
characteristic of government responsibility to the assembly must not be
rendered ineffectual in practice by the dominance of the president.

Other cases
Although we do not undertake further detailed case studies here, it is
worth noting that both Duverger and Nogueira identified three additional
countries as semipresidential: Austria, Ireland, and Iceland. 16 The most
notable characteristic common to these three is the failure of their presi­
dents to exercise political influence - either in the legislative process or as
arbiters among parties. In the case of Ireland, as we shall see, this is not
surprising, given that the president has no constitutional powers. In the
Austrian and especially Icelandic cases, however, the evident passivity of
the president is an enigma.
For the Austrian case, Nogueira offers a plausible explanation. The
leverage of the president as arbiter rests primarily upon the power to name
16 Neither identified Sri Lanka as a semipresidential system. Thus. each identified a total of
six extant cases of the regime type.
72 Presidents and assemblies
prime ministers, but where a party system is d ominated by tw o main parties
or bl ocs, one of which generally secures a maj ority in parliament, the
arbiter p ower of the president will likely be mitigated. The leaders of the
principal Austrian parties have n ot run for president, but have remained in
the assembly t o c ompete f or the prime ministership. M ore over, the use of
grand c oaliti ons in Austria f or much of the p ostwar peri od, reduced the
p otential f or presidential discreti on even further.
The Icelandic case seems t o challenge N ogueira's l ogic. Here, we have
b oth a multiparty system and strong f ormal presidential p owers over legisla­
tion and g overnment formati on. Yet n ot only are Icelandic presidents uni­
f ormly acquiescent t o g overnment directives, the presidency is not even
s ought by partisan candidates. The c onstituti onal pr ovisions may be in part
attributable t o b orr owing of the f ormal p owers of the Danish m onarch and
applying them t o the Icelandic head of state at the time of Iceland's inde­
pendence fr om Denmark. Just as the Danish monarch's p owers are never
exercised in the era of a c onstituti onal m onarchy, neither are th ose of the
Icelandic presidents. Still, the pr ovisi on of p opular election w ould seem to
imply that partisan candidates w ould c ompete f or the office. Perhaps the
main reas on that n o such candidate has emerged is the ease with which the
c onstituti on c ould be amended by parties that w ould oppose such a presi­
dent. Lijphart (1984:189) sh ows n o premier-presidential system besides
Iceland in which the c onstitution may be ref ormed by "pure maj ority rule"
within parliament. Thus the presidency (like a m onarchy) c ould be simply
amended away or have its p owers curtailed, as happened in Portugal,
sh ould a threatening president ever emerge. 17
In additi on to the cases identified by Duverger and Nogueira, and th ose
discussed ab ove, we shall discuss briefly tw o new regimes, th ose of Haiti
and Namibia, as well as one officially end orsed prop osal. In Argentina,
there has been a pr op osal f or premier-presidential g overnment by the
C omisi6n p or Ia C ons olidaci6n de la Dem ocracia established during the
Alf onsin administrati on. The Argentine prop osal f or a new f orm of g overn­
ment was a response t o the standard criticisms of presidentialism outlined
in Chapter 3. The members of the comisi6n outlined a premier-presidential
alternative f or Argentina, but the pr op osal has not been f ormally c onsid­
ered by the Argentine C ongress or a constituti onal c onventi on. F or the
purp oses of c omparison, we include the relevant elements of the reform
purposal presented by the comisi6n's report in our analysis in Chapter 8 of
presidential p owers.
The new Haitian c onstituti on of 1987 stands out f or at least three rea­
s ons. First, Haiti's was the first new dem ocratic c onstituti on of all th ose
17 Although the provision requires a three-fourths majority to carry forward, the Icelandic
parliament may call for a plebiscite on the president's removal from office , providing a
further check on presidents' exercise of powers that might threaten the institutional inde­
pendence of the assembly.
 Premier-presidentialism, president-parliamentarism 73
enacted in the Western Hemisphere in the 1980s in which the cabinet relies
exclusively on parliamentary confidence. Second, the constitution attempts
to avoid crises over appointment of the premier and relations between the
premier and president. The president must, according to Article 137 of the
constitution, propose as premier a member of the majority party in the
assembly. In the event that there is no such party, the president must first
"consult" with the presidents of both chambers of the assembly. While the
precise meaning of "consult" is left unclear, the intention is obviously to
make sure that whenever there is no majority party, the assembly retains
primacy in determining the makeup of the government. In this sense, the
Haitian president would seem to be virtually, if not completely, powerless,
as is the Irish president. The final noteworthy aspect of the Haitian constitu­
tion, and one that could be a source of friction, is its requirement for
simultaneous confidence in the cabinet by both chambers. Since the cham­
bers are constituted differently, it is possible that their majorities might
differ at times. 18 As Lijphart (1984) observed with respect to Australia,
such bicameral confidence can hold the potential for a constitutional crisis.
Indeed, the requirement of confidence before a bicameral assembly could
be construed as a form of dual democratic legitimacy, a factor that critics of
presidentialism attribute uniquely to that regime type.1 9
Namibia under the constitution of 1989 is another president-parliamen­
tary regime, but with an unusual variation that may provide a "safety valve"
for conflicts between president and assembly, although it is still too early to
tell. The president has the power to name the entire cabinet and may dismiss
it in whole or in part, provisions that on the surface imply presidential
dominance akin to that in Weimar and Sri Lanka. But Namibia may have
retained a means of ensuring that the president cedes at least some authority
to the assembly (or, more directly, to its agent, the prime minister) over the
direction of the cabinet by requiring that a president who dissolves the
assembly stand for reelection concurrently with new assembly elections.
That is, if the Namibian assembly repeatedly censures cabinets, presidents
have the alternative of calling for new elections, but only if they are willing to
face the electorate themselves. Rather than risk such a confrontation in
which they could be removed from office, we might expect Namibian presi-
18 The electoral system used for both chambers is majority runoff, diminishing somewhat the
likelihood of different majorities. On the other hand, the upper house is to be elected on
staggered six-year terms, with one-third of the seats at stake every two years, while the
lower house is to be renewed in full every four years.
19 It is indeed possible, that with the "arbiter" in cases of noncompatible majorities in the two
assemblies being a popularly elected president instead of an appointed governor-general
(as in Australia), the potential for conflict might be less. On the other hand, the Haitian
head of state lacks the authority to dissolve either chamber as a possible means of resolving
deadlock, as happened in Australia in 1975. If the president were from one party or an
"outsider" and each assembly were controlled by different coalitions, there could be "triple
democratic legitimacy"!
74 Presidents and assemblies
dents to be willing to accept independent direction of the cabinet by the
prime minister. At this point, with the first Namibian President Sam Nujoma
enjoying a secure majority of SWAPO partisans in the assembly, the system's
performance under other conditions is undetermined.
Finally, Chile's so-called parliamentary republic (1896-1925) and the
current regime in Peru have been described both as parliamentary and as
effectively premier-presidential, on account of the existence and frequent
use of the no-confidence vote for cabinet members (Needler 1965 ; Heise
1974) . Here too, however, we maintain that it is more apt to classify these
systems as president-parliamentary regimes, owning to the dual responsi­
bilities of the cabinet to both the president and assembly, as well as the fact
that while the assembly could reject cabinets, it had no say in how to
reconstitute the government. Although the effectiveness of censure in each
case significantly weakened the president's control over the cabinet, cen­
sure did not ensure cabinet responsiveness to the assembly majority. 20
Our definition of premier-presidentialism requires that a prime minister
direct the cabinet, which in turn requires that the cabinet depend only on
the confidence of the assembly. Those who would label Chile and Peru
premier-presidential would argue that the assembly (without a prime minis­
ter as agent in Chile's case) effectively directed the cabinet. But when the
president and congress were fundamentally at odds on the direction of the
cabinet, both Chile and Peru experienced chronic cabinet turnover -
effectively, cabinets with no clear direction at all. We reiterate that just as
Chile and Peru do not meet our requirements for premier-presidential
government, neither does Portugal from 1976 to 1982, nor Sri Lanka, nor
Namibia. In the Sri Lankan case, the assembly clearly has little recourse to
ensure that the prime minister can direct the cabinet. Indeed, given the
inordinate formal powers of the president, it is difficult to understand how
the existence of a prime minister and the requirement of cabinet responsibil­
ity to the assembly can moderate presidential dominance at all, short of
triggering some sort of constitutional crisis. For all these systems, we shall
be able to see more clearly why they fail to be genuine cases of premier­
presidentialism (or, for that matter, presidentialism) when we consider
powers of presidents more systematically in Chapter 8. For now, we simply
point out one of that chapter's conclusions, which already has been fore­
shadowed in the discussion here: Those regimes with dual responsibility of
the cabinet to both the assembly and the president have tended to be
unsatisfactory in their performance.

CONCLUSION
Part of the problem in assessing the viability of premier-presidentialism has
resulted from confusion over what the regime type is. In Chapter 2, we
20 The same observation is likely to prove true for the new constitution of Colombia.
 Premier-presidentialism, president-parliamentarism 75
proposed a number of characteristics that we understand to be essential to
the regime type originally described by Duverger. In this chapter we have
suggested that the performance of premier-presidential regimes, particu­
larly during the periods of cohabitation that are most likely to threaten
regime stability, will depend on the clarity of the division of executive
responsibilities between president and cabinet. The specific combinations
of responsibilities can vary. One key point is that ambiguity will increase
the danger that either a president or an opposition assembly majority will
reject the claims to executive authority of the other. This has proven espe­
cially true in the cases we have identified as president-parliamentary. We
suggest that, to minimize conflicts between premier and president, a "resid­
ual powers" clause should grant any unspecified or unclear powers to the
premier. Another matter that should be recognized by constitution design­
ers is that placing sufficient powers in the hands of the president that he or
she can ignore pressures from the assembly altogether is a dangerous for­
mula, and one that we would not recommend. Finally, as we shall develop
more fully in Chapter 12, the relative timing of presidential and assembly
elections will largely determine the likelihood both of compatibility within
the executive and that claims to executive authority will be challenged. We
now turn to a more systematic examination of the constitutional balance of
powers between presidents and assemblies. We shall begin by reviewing the
means by which executives and assemblies are elected or selected, proceed
to examine their mutual dependence or independence for survival, and
finally consider their relative legislative powers.
 5
The constitution al origins of chief executives

INTROD U CTION

As defined in Chapter 2, a principal aspect in which presidential regimes

are distinct from parliamentary, premier-presidential, and various hybrid
regime types is in the separate origin and survival of executive and assem­
bly. Recall that a requirement for presidentialism is that the chief executive
be elected by popular vote. However, in a number of regimes that have
generally been considered presidential or hybrid systems, the chief execu­
tive has been constituted according to various means, which sometimes
modify and sometimes preclude the criterion of popular election. In this
chapter, we shall examine in detail various means of constituting chief
executive power in presidential and hybrid systems, reviewing a number of
case studies to illustrate the possible variations. We undertake this exercise
for two main reasons. First, we demonstrate that there are any number of
ways to constitute executive power in regimes that are neither presidential
nor premier-presidential, nor parliamentary. Second, and more important,
the cases highlight the impact that various rules guiding the origin of the
executive have on the incentives that drive political behavior. As we shall
see, the nature of the party system and of coalition building, as well as the
means of replacing the chief executive, all depend largely on the constitu­
tional means by which the executive originates. In this chapter, we shall
discuss in turn:
1. selection of the chief executive by the assembly (instead of the electorate);
2. variations on presidential eligibility for reelection;
3. the existence and role of the vice-presidency; and
4. the prospect of a collegial presidency.

SELECTION OF THE CHIEF E X ECUTIVE BY THE ASSEMB LY

Of the variations we shall discuss, the selection of the executive by the

assembly is one of those most likely to constitute a serious breach of the
presidential principle that the origin of assembly power should be distinct
 Constitutional origins of chief executives 77
from the origin of executive power. Indeed, only if assembly selection of
the executive is in practice no more than ratification of the popular elec­
toral plurality would we consider the regime presidential. Chile from 1925
until 1973 was such a case. In other cases, where selection of the executive
involves the formation of coalitions among parties within the assembly, we
cannot call the regime presidential, even if there is an initial round of
voting for executive by the electorate.
In the United States, the so-called V irginia plan at the Constitutional
Convention would have provided for selection of the president by congress
instead of the separately elected electoral college that was adopted (see
Ketchman 1986). In the U.S. Constitution there is, nonetheless, a provision
for congressional selection in the event that the electoral college does not
produce a majority for one contender. In such cases, the House of Repre­
sentatives selects the chief executive from among the three (originally five)
top candidates (Art. II, Sec. 1). Of course, the conditions for congressional
selection have proven truly extraordinary; it has happened only twice, and
not since the 1824 election.
Interestingly, the provision for congressional selection even under extra­
ordinary circumstances contradicts Alexander Hamilton's argument in fa­
vor of an electoral college as offering the best combination of directness
and prudence in selecting an executive. Hamilton writes:
[The authors of the constitution] have not made the appointment of the President to
depend on any pre-existing bodies of men who might be tampered with beforehand
to prostitute their votes; but they have referred it in the first instance to an immedi­
ate act of the people of America, to be exerted in the choice of persons for the
temporary and sole purpose of making the selection. And they have excluded from
eligibility to this trust all those who from situation might be suspected of too great
devotion to the President in office. No senator, representative, or other person
holding a place of trust or profit under the United States can be of the number of
the electors.

And elsewhere, "Nothing was more to be desired than that every practi­
cable obstacle should be opposed to cabal, intrigue, and corruption" (Ham­
ilton, Federalist No. 68).
It may be that Hamilton was able to reconcile his determined opposition
to congressional selection with its existence in the constitution because it
seemed such an unlikely turn of events. At any rate, his is perhaps the most
eloquent rebuttal to current arguments in favor of backroom parliamentary
negotiations to select an executive. 1 Still, as we shall see, a regime with
congressional selection of the executive can offer certain conditional advan­
tages over either presidentialism or parliamentarism, especially for soci­
eties with deep divisions.

1 See Linz (1987), as well as the discussion of his argument in Chapter 2.

78 Presidents and assemblies

Congressional selection: not parliamentary, but

assembly-independent regimes
Where the executive emerges out of c oalitions in the assembly, the regime
takes on a key feature of parliamentarism. H owever, if the assembly can­
not pass a resolution of n o-c onfidence and replace the executive with a new
one before the end of the constituti onal term of office ( other than by an
extra ordinary pr ocedure, such as impeachment), the regime is not parlia­
mentary. In Chapter 2, we offered a general typ ol ogy of regime types in
which we label such a hybrid an assembly-independent regime, and we
shall offer a m ore extensive discussi on of this typ ol ogy in Chapter 8. In this
hybrid regime, the origin of executive p ower is n ot separate fr om assembly
p ower, but the survival of the tw o branches remains independent.
The oldest current case of this type of hybrid regime is the Swiss system.
In Switzerland, a c ollegial executive is ch osen f or a fixed term by the
bicameral elected assembly. The Swiss experience suggests what might be a
m ore general less on of this regime type: Unlike the assemblies of many
parliamentary regimes, the Swiss assembly remains an imp ortant source of
legislati on. Lijphart (1984) rep orts that, measured by the degree t o which
the assembly makes b oth general laws and m ore specific rule-making (regu­
lations), the Swiss assembly is the sec ond m ost p owerful legislature in his
sample of twenty-tw o dem ocracies. We partially disagree, h owever, with
his assessment of the reason. Lijphart attributes the p ower of the Swiss
assembly as a legislature t o its being "the sec ond clear case of separati on of
p owers" (Lijphart 1984:79) in additi on t o the United States, which Lijphart
rep orts as having had the m ost active legislature.
It is only partly accurate t o refer t o Switzerland as having a separati on of
p owers. The separati on of survival of p owers - owing t o the lack of a v ote
of n o-c onfidence ( or diss oluti on) - n o d oubt acc ounts f or the activism of
the assembly. H owever, the origin of executive p ower in the assembly
means that, t o be selected in the first place, the members of the executive
must be in general agreement with the p olicy wishes of their parties in the
assembly. They d o n ot have separate acc ountability before the elect orate,
as d oes a presidential executive. On the other hand, the assembly can
engage in active lawmaking without risking bringing down the government
if s ome pr op osal deviates fr om the principles upon which the g overnment
was originally f ormed. In parliamentary systems, all owing the assembly t o
remain activist in lawmaking may entail a cost: executive instability.
In order t o see if there is such a c onnecti on between assembly activism in
legislati on and sh ort-lived cabinets, we have prepared a simple table t o
capture a sample o f parliamentary democracies in these tw o dimensi ons
(Table 5.1). We rely on Strom's (1990) sc ores of " opp ositi on influence" and
P owell's (1989) sc ores of "government d omination" in legislative commit­
tees t o measure the extent t o which the assembly initiates legislati on of its
 Constitutional origins of chief executives 79

Table 5 . 1 . Opposition influence and cabinet durability in parliamentary democracies

Cabinet durability
High Low
(over 48 (less than 48
months) months)

High Austria Belgium Greece

Canada Denmark Iceland
Norway France IV Italy
Opposition Sweden Finland Spain
influence Germany

Low Australia Israel

France V Netherlands
Ireland
New Zealand
U.K.

Source: For cabinet durability, Lijphart (1984:83); fo r opposition influence, Strom

( 1990:73) and Powell ( 19891.

own or amends legislation presented by the cabinet. Lijphart (1984) sup­

plies the data on cabinet durability, which we have broken down into two
categories: greater or less than four years, which is the typical term of
parliaments, discounting early dissolutions. We should expect that active
assemblies and short-lived cabinets would tend to go together. As legisla­
tion is made by shifting coalitions, given a high degree of opposition influ­
ence, some of these coalitions would form for a larger purpose: replacing
the government.
Indeed, nine cases are located in the cell identified by high opposition
influence and short-lived cabinets. Israel and the Netherlands stand out as
exceptional, but given their very large district magnitudes and associated
extreme multipartism, their cabinets would be expected to be unstable
regardless of the degree of opposition influence. Dutch and Israeli cabinets
would probably be even shorter-lived if opposition influence were high,
since minority governments might then be more common and such cabinets
are typically short-lived (Lijphart 1984:81). 2 At the other corner of the
2 According to Strom ( 1990), minority governments are most common when opposition
influence is high. The reason is that, given opportunities for parties not in the cabinet to be
effective legislatively, the costs of being in opposition are not great. This may appear to
contradict our argument regarding a tendency of parties in active assemblies to initiate votes
of no confidence. However, our observation is independent of the frequency of minority
governments: the cost of being out of government is indeed low if one can influence
legislation from outside government, but is also low if one expects that changes in govern­
ment are relatively easy and will occur relatively frequently. Lijphart's data on the short life
of minority cabinets support this assessment.
80 Presidents and assemblies
table, we find Canada, which because it uses single-member districts usu­
ally has single-party (and therefore long-lived) cabinets. Likewise, Austria
has a two-party system that makes long-lived cabinets the norm. Sweden's
presence there is harder to account for, although even its minority cabinets
have often consisted of one party that was near a majority, making it (and
Norway, too) a case of a two-bloc (government-opposition) party system
(Strom 1990). Generally, then, there is a trade-off under parliamentarism
between having an assembly that is active in legislation and having cabinets
that endure in office long enough to provide a program of policy options for
the polity. Powell's (1989) findings of a low correlation between "represen­
tational" constitutional designs and governmental accountability and a
negative correlation between majoritarian constitutional design and his
"representative delegates" model of citizen control offer confirming evi­
dence for our assertions.
Because the Swiss hybrid regime provides for fused origin and separate
survival, executive stability can be assured while maintaining an assembly
independent enough to allow the enactment of legislation by shifting
multiparty coalitions. We should expect such a regime format to be espe­
cially favorable for divided societies. Indeed, Switzerland, with its multiple
cleavages and assembly-selected collegial executive, is Lijphart's clearest
case of a consensus democracy (Lijphart 1984).

Lebanon. A regime similar to that of Switzerland has existed in Lebanon.

While Lebanese democracy broke down in the mid 1970s, as Lijphart
(1977) observes, nearly all proposals for dealing with the Lebanese situa­
tion start from the same basic assumptions about power sharing that lay
behind the original creation of the Swiss and Lebanese constitutions. In
1989, a proposal known as the Taif Accord called for reinstituting a power­
sharing executive in which the Christian president would continue to be
selected by the parliament for a fixed term. Under the Taif Accord, how­
ever, the Muslim premier, who is responsible to parliament, would be given
nearly coequal powers (personal communication with Lijphart). Given that
part of the Lebanese executive has been dependent on parliamentary confi­
dence, the regime is of a different type than Switzerland. Still, with a fixed­
term president selected by the assembly and possessing political powers,
the regime is not parliamentary. See Chapter 8 for a further characteriza­
tion of this and similar regimes.

El Salvador. Three Latin American systems are also noteworthy for their
use of assembly selection of the president. One, El Salvador, used this
method only as a transitional device before the writing of a new constitu­
tion which provided for the direct election of the president. The brief
(1982-84) regime must be considered a success, given the extremely diffi­
cult circumstances under which it was implemented. A presidential election
 Constitutional origins ofchief executives 81
at that time would surely have pitted Roberto D' Aubuisson, the leader of
the extreme (perhaps one could say fascist) right, against the left-of-center
Christian Democratic leader, Jose Napoleon Duarte. The powers of the
office they were contesting would not yet have been determined'. Such an
election thus would have confirmed Linz's worst fears about winner-take­
all presidential elections. Instead, those two parties and several others
participated in an assembly election in the midst of a heated guerrilla war.
The assembly, in which the Christian Democratic party won a plurality but
various right-wing parties combined had a majority of the seats, selected a
Christian Democrat other than Duarte to be interim president. It also
created three vice-presidencies (each held by a different party) and a
power-sharing cabinet, thus resembling the Swiss model. By 1984, a new
presidential constitution had been produced that gave only limited powers
to the president, thus preserving the assembly as an arena of compromise
while still creating a popularly accountable chief executive.
Bolivia and Chile are the two other instances of congressional selection
of the president. However, the operation of these systems presents a stark
contrast, in which the Bolivian president has been selected out of assembly
bargaining, while in Chile, the electorate's plurality choice was consistently
simply ratified by congress. After discussing each case in turn, we shall
consider some reasons why Bolivia and Chile differed even though both
have had the same formal provision for the selection of the president in the
assembly in the event that the popular election does not produce a majority
winner.

Bolivia. The Bolivian case is different from the others so far discussed in
that there are presidential as well as assembly elections, although voters
must vote the same party ticket for both assembly and executive. If one
candidate obtains a majority of votes, that candidate becomes president. In
the event of no majority of popular votes, the assembly chooses the presi­
dent from the top three contenders. 3 As we shall see, the prevailing situa­
tion has been for presidents not to have been the choice of a plurality of the
electorate.
Given its seemingly endless history of political turmoil, it is difficult to
distinguish the exceptions from the rules in Bolivia. The presidency has
3 We are leaving aside here what Malloy and Gamarra (1987) call the salida option. By the
salida, Congress, in an unconstitutional procedure, selects a compromise candidate who
had not contested the popular election and is inoffensive to any of the competing parties.
During the tumultuous year 1979, as Bolivia began its halting transition back to civilian
government after fifteen years of military rule, Congress opted for the salida twice. Bo­
livia's salida presidents have proven ineffectual vis-a-vis the legislature (and vis-a-vis the
military as well); but of course, this was most likely the result intended by Congress in
selecting them. The salida represents an attempt to buy time, maintaining civilian control
over the executive while a stronger mandate can be sought, either through another election
or through further elite bargaining. See Gamarra (1987) for details.
82 Presidents and assemblies
been seized by extra-c onstituti onal means m ore frequently than by electi on
in the last half-century. 4 Still, of those presidents selected by c onstituti onal
means since the implementati on of the 1967 c onstituti on, congressional
selecti on by parliamentary-style neg otiati ons has been the rule. What's
more, plurality winners in the general electi on have fared remarkably
p o orly in the c ongressi onal selecti on pr ocess. 5
In 1982, tw o years after he had w on a plurality in the June 1980 electi on,
Hernan Siles Suaz o of the Nati onalist Rev oluti onary M ovement (MNR)
was awarded the presidency as the military returned t o the barracks. But
Siles's ascendance t o office was neg otiated between party leaders and the
outgoing generals rather than in c ongress. When electi ons were next held,
in 1985, V ict or Paz Estenss or o (Paz E.), one of the original leaders of the
MNR, finished sec ond in the p opular v ote. F ormer military dictator Hug o
Banzer, heading the Dem ocratic and Nati onalist Acti on Party (AON), led
by less than 2%. In the c ongress, Banzer was unable t o secure maj ority
support. The MNR and AON v oted t ogether t o supp ort the accession t o
the presidency b y Paz E . His g overnment pr oceeded t o enact a major
econ omic restructuring or "sh ock" t o stanch hyperinflati on (Gamarra
1990). Banzer ran again in 1989, and finished sec ond with 22.6% t o the
MNR's 23.0%. This time, in c ongress, Banzer threw his supp ort behind the
candidate of the center-left M ovement of the Rev oluti onary Left (MIR),
Jaime Paz Zam ora (Paz Z., Paz E.'s nephew), wh o had finished third with
just under 20% of the v ote. Like his uncle and predecess or, Paz Z. has
g overned with a c oaliti on cabinet c omp osed largely of members of Banzer's
AON and c ontinued with the pr ogram of structural adjustment.
The legislative effectiveness of the c oaliti on governments led by Paz E.
and Paz Z. since 1985 has been n otew orthy. F or better or w orse, they have
succeeded in implementing and maintaining a harsh package of austerity
measures. B olivia's public sect or has been cut c onsiderably and its currency
has stabilized, making B olivia a sh owcase f or ne oliberal devel opment eco­
nomics (Gamarra 1990). In these g overnments, the c onservatives have
f ormed the largest c onsistent supp ort bl ock in c ongress. The center and
center-left, on the other hand, have been awarded the presidency in ex­
change f or acquiescence t o the right's econ omic pr ogram. As l ong as
centrist or leftist presidents are willing t o trade contr ol over p olicy f or
congressi onal support, the coaliti on can survive. Ir onically, the very success
4 By our estimation of Bolivia's often complicated transitions and interim periods between
1952 and 1989, the presidency has been secured by overtly unconstitutional means eleven
times, and by either direct election or congressional selection eight times.
5 Since 1967, the president has been selected by Congress five times. Only in 1982, when
Hernan Siles Suazo was named president based on aborted 1980 election results, has the
plurality winner actually assumed office. Two times in 1979, members of congress who were
not presidential candidates at all were selected in extra-constitutional procedures. In 1985.
the second-place presidential contender was named; and in 1989 the third-place finisher
became president.
 Constitutional origins of chief executives 83
of these administrati ons may be attributable t o the dependence of their
presidents on c ongressi onal c oaliti ons. Str onger presidents - perhaps plu­
rality winners - may have been less pr one t o compr omise with and even
acquiesce t o c oaliti on partners. And it is this c ompr omise that is essential
t o breaking c ongressi onal deadl ock.
Thus, the system appears t o be executive-d ominated, leading C onaghan,
Mall oy, and Abugattas (1990) t o argue that the B olivian C ongress was
reduced t o n othing m ore than legitimating or ratifying, as Gamarra
(1987:411) puts it, p olicies designed by a small circle of techn ocrats insu­
lated within the executive. Gamarra (1987:412) g oes s o far as t o brand
B olivia an "auth oritarian dem ocracy." We maintain that the pr ocess is
subtler than that. As the pr oducts of c ongressi onal bargaining, these appar­
ently d ominant executives actually represent a r ough equilibrium between
parties in c ongress and the executive (see Ordesh o ok 1986; Rasmussen
1989). H owever, there may be severe l osses when auth ority is delegated
without a specific means of retracting it. Despite the origin of executive
auth ority in the assembly, the assembly cann ot v ote n o confidence and thus
enf orce its will on the executive, as in a parliamentary system. This ex­
plains why B olivian c oaliti on g overnments have functioned in what appears
t o have been a maj oritarian, if n ot d ownright auth oritarian, manner. G ov­
ernments have achieved c onsiderable insulati on from p olitical pressures,
but their own origin has depended on the willingness of the assembly t o
delegate the auth ority in the first place. Even where n o specific retracti on
of delegated auth ority is available immediately, we sh ould expect t o see
attempts by the parties t o equilibrate the situati on subsequently, given the
means of executive selection in B olivia.
This is, in fact, what we see in the case of intrac oaliti on conflicts during
the Paz Estenss or o g overnment, which prevented the c oaliti on from h old­
ing t o a prior agreement, signed by the AON and MNR in May 1988, t o
present Banzer as a c omm on candidate in 1989 (Gamarra 1992). Gamarra
indicates that, while MNR militants c omplained ab out the d ominati on of
Paz E.'s g overnment by the AON and the AON did not trust the MNR's
pr omises of support for Banzer, weekly meetings between the AON and
MNR leadership were instituted t o h old the coaliti on t ogether. When the
MNR indeed failed t o end orse Banzer, the path was again paved f or a new
p ostelecti on c oaliti on after the 1989 election. Banzer, in effect, was thus
denied the presidency because of the MNR's l oss of "c onfidence" in a
g overnment that had become overdependent on the AON f or its p olicies
and supp ort.
Had the parties stayed t ogether and presented a c ommon candidacy in
1989, the r ole of p ostelecti on neg otiations within the assembly might have
been reduced. Even if the Banzer c oaliti on candidacy had failed t o win a
maj ority of p opular v otes, the c ombined v otes of the AON and MNR could
have put it in such a c ommanding p ositi on that the opp ositi on in c ongress
84 Presidents and assemblies
would have found it nearly impossible to put together an alternative to
Banzer after the election. As it turned out, the parties ran separate candida­
cies, so a bargain afterward was necessary again. Now the MNR paid the
price for having broken the agreement with the AON, which voted for the
MIR candidate. The implication is that any one of the three largest parties
needs at least one of the other two to win: It needs either for the other not to
present a candidate or for the other to vote for it after the election. This basic
fact of the Bolivian "game" should enforce considerable compromise be­
tween parties - even when they may complain about executive dominance -
and thus helps account for early success in implementing the difficult and
politically risky austerity program. 6
A parliamentary system, on the other hand, could have been a victim of
considerable executive instability,7 given the large number of players and
the difficult policy choices needed. A pure presidential system would not
have been as capable of holding together or reconstituting coalitions as has
Bolivia's hybrid (assembly-independent) system, since executive power
would then be independent of cross-party bargaining in the assembly. The
Bolivian system combines the coalitions of multiparty parliamentarism and
the executive stability of single-party parliamentarism or (true) presiden­
tialism. W hile we would not be inclined to endorse the Bolivian system as a
model, its performance under difficult conditions and, even more notably,
in a country with a history of repeated coups earns it considerable respect
in our view.
One final set of observations about Bolivia concerns the comparison with
Switzerland, with which we started this section on assembly selection of a
fixed-term executive. Despite the similarity between Bolivia and Switzer­
land in the method of executive selection, obviously the Bolivian Congress
has not been an active legislature in the same sense that the Swiss assembly
has been. W hile we have argued that this does not mean that the Bolivian
Congress and the parties represented therein have been rendered meaning­
less or mere "rubber stamps," neither has Bolivian legislation been initi­
ated and debated in the congress to a great extent. Partly this is a result of
an executive in Bolivia that, unlike the Swiss, possesses a veto on legisla­
tion (subject to a two-thirds assembly override). Thus it takes a large
coalition to legislate over the president's head. The fact that some of the
legislation has been carried out under a state of siege , which requires a two­
thirds vote of congress to sustain (Gamarra 1992), has made it even more
unlikely that we would see legislative activism. Thus a very large coalition
was required to enable the implementation of policies suppressing labor,
6 We do not mean to endorse the austerity program itself; indeed, according to Gamarra
(1990), it dealt "a heavy blow to labor" and has caused considerable hardship. What we
regard as instructive is that other countries' governments found it considerably more diffi­
cult to implement similar programs, while Bolivia was able to hold to the program.
7 See Gamarra (1992) for a discussion of this point.
 Constitutional origins ofchief executives 85
for example, in the first place. Finally, the Swiss executive is collegial,
unlike Bolivia's, meaning that there is no single "program" of government
in Switzerland against which to compare legislation passed in the assembly.
In Bolivia, in a context of severe economic hardship, the congress has
delegated wide responsibility to the executive. The selection of the execu­
tive in congress has been crucial to the early (compared to some other Latin
America countries) success of the stabilization program. A lesson of the
Bolivian case, then, is that mechanisms that give incentives for cross-party
coalitions can make for effective policymaking even in as difficult an envi­
ronment as that afforded by Bolivia.

Congressional ratification: Chile

Before its interruption by the military coup in 1973, Chile's line of demo­
cratically elected presidents led back to the early 1930s, when a populist
and reformist junta ruled briefly. Beyond that, Chilean presidents had been
chosen by legitimate elections (albeit with restricted suffrage) back into the
mid-nineteenth century. Until the promulgation of the most recent constitu­
tion in 1989, the Chilean constitution of 1925 had provided for the selection
of the president by the Chamber of Deputies in the absence of a majority
winner in the popular election. However, throughout the 1925-73 period,
presidential elections operated under a de facto plurality rule. That is,
when no candidate won 50% or more of the vote, the chamber named as
president the candidate who had won the plurality, almost always without
rearranging party coalitions or striking cloakroom deals for the formation
of an executive.
Before the adoption of the constitutional provision for congressional
selection in 1925, there had been one case in which a Chilean president was
appointed by means other than direct vote. In this instance, the plurality
winner was bypassed for the second-place candidate, Arturo Alessandri
Palma, who finished second by 1 % of the popular vote. The constitution of
1891, which was in effect until 1925, did not call for congressional selection,
but because of disputes over election results, congress appointed a tribunal
of legislators to select the new chief executive (Drake 1978). At the time,
Alessandri was a political outsider whose platform called for redistributive
policies. The tribunal awarded him the presidency to quell protests by his
working-class supporters, but the plan was to block any attempts at socio­
economic reform in the conservative-dominated congress (Drake 1978).
Alessandri broke Chilean presidential tradition, however, by campaigning
hard for reformist candidates for congress in midterm elections, success­
fully strengthening his parliamentary support base. His reformist efforts
paid off, as in 1925 Chile ratified the new constitution which, among other
changes, strengthened the president's control over cabinets.
The Chilean Chamber of Deputies proceeded to name the pluraity win-
86 Presidents and assemblies
ner to the presidency without exception in every successive election in
which there was no majority winner: in 1946, 1952, 1958, and 1970. So
while the practice of congressional selection in mid-century Chile was
nonpresidential by strict constitutional standards, the system functioned
essentially as presidential. Given the regularity with which the chamber
named the plurality winner president, in fact, the Chilean case actually
illustrates the implications of plurality elections better than of congres­
sional selection.

Accounting for different practice in Bolivia and Chile

Before the Chilean coup, congressional and presidential executive elec­
tions were never concurrent. We suggest that this difference in the electoral
cycle can account for why Bolivian parties regularly did not coalesce before
presidential elections, while Chilean parties usually did. Additional factors
include ballot format and, crucially, the number of contenders from which
congress may choose.
At the time of presidential elections, the electoral mandate of the Chilean
Congress was always at least one year old. In contrast, in 1985 and 1989, the
Bolivian Congress has been elected concurrently with the direct balloting for
president. The timing of elections affects both the knowledge parties have of
their strengths before the selection of the president and the actual bargaining
power that they will have in postelection negotiations. Owing to concurrent
elections, Bolivian parties have no congressional election before the presi­
dential election to provide them knowledge of what their congressional
bargaining power will be in the postelection congressional negotiations that
lead up to the vote on presidential selection. Therefore, it is rational for each
of the three major parties to present its own candidate to establish bargain­
ing power rather than risk being in a lesser coalition and thus squandering
bargaining power. 8 Moreover, the separation of presidential and congres­
sional power in Bolivia is weakened by the ballot structure. Elections for
president and congress are "fused," meaning there are no split ballots. Thus,
unlike Chile's elections, which are separated by date as well as by ballot,
Bolivia's elections are indeed more like parliamentary elections: one voting
result determines the electoral resources of the parties. If, as we have sug­
gested, the very weakness of these Bolivian presidents has been the source of
congressional-executive cooperation since 1985, it is likely that the electoral
format has contributed to this effect.
In Chile, on the other hand, the congress that would have the ultimate
choice in the absence of a popular majority would be the same one that was
already sitting at the time that the presidential election would take place. An
8 Bolivian parties do have midterm municipal elections, which provide one element of the
information puzzle, tracking voter loyalties, but this information tells them far less about
bargaining resources than would a congressional election.
 Constitutional origins ofchief executives 87
election would already have taken place shortly before the presidential elec­
tion to track any possible trends in voter support. Other parties than the
strongest would find it rational to join either with the party that looked likely
to win or with an opponent that might be able to block the rising party.
Because parties would know the bargaining power they would have in con­
gress should a vote there be necessary, they could anticipate both the likely
electoral rewards of any preelection coalition and the ability to structure a
new postelection coalition, if necessary. For this reason, some parties gener­
ally had forged a coalition before the presidential election that either ob­
tained a majority or was sufficiently close to a majority that no new coalition
could be easily formed in congress to undo the preelection coalition.
One final matter needs to be addressed with regard to the Chilean case:
W hat about those cases when the candidate who received a plurality did so
very narrowly? W hy, when Jorge Alessandri won only 31 % in 1958 and,
even more significantly, when the Socialist Salvador Allende won only 36%
in 1970, did the parties in congress not make use of their formal power to
select a contender other than the plurality winner? In 1958 the answer is
obvious enough, since the congress had to select from among the top two
contenders. The runner-up, Allende, was certainly less desirable to a ma­
jority than was Alessandri. In 1973, the congress faced the same choice,
except that this time Alessandri was the runner-up. Why opt for Allende,
the plurality winner? The key was in the Christian Democrats' preferences.
The incumbent president, a Christian Democrat, had been elected in 1964
by a majority because he received the conservative parties' endorsement.
Because Alessandri's candidacy in 1970 represented a break with the
center-right coalition and the conservatives and Christian Democrats had
been fighting for hegemony over the nonsocialist electorate, ironically, the
vote to Allende may have appeared politically less risky. Allowing Ales­
sandri to become president would have been to risk conceding hegemony
over the nonsocialist forces to the Christian Democrats' convervative ri­
vals. Thus the restriction on congressional choice to the top two candidates
was as important as the electoral cycle: Unable to bargain for the selection
of the third candidate, as happens in Bolivia, the Christian Democrats,
whose candidate had placed third but who held the largest share of seats in
the previously elected congress, punished their erstwhile allies and voted
for the plurality winner.

V A RI A T I O N S O N PR E S I D E N T I A L R E E L ECTION

The modal rule among presidential systems prohibits immediate reelection.

In many, reelection is forbidden outright; in a number of others, a presi­
dent can be reelected after one interim term; and in Venezuela, two terms
must elapse before a former president may be reelected. In the United
States, which many scholars have assessed as the presidential archetype (an
88 Presidents and assemblies

Table 5.2. Reelection of presidents

No Two interim One interim Two-term No

Reelection terms term limit limit
Colombia 1991 Venezuela Argentina Bulgaria Dominican
Costa Rica *Bolivia Namibia Republic
Ecuador Colombia pre- Portugal Finland
El Salvador 199 1 Romania France
Guatemala Panama United States Nicaragua
Honduras Peru Paraguay
Mexico Philippines
Sri Lanka
* Although Bolivian presidents are usually elected by the assembly, the process begins with
a popular election. The term limit applies regardless of whether the president is ultimately
chosen in the assembly or by the electorate.

assessment that seems to have been based mostly on provincialism, and

one that we do not share), the Twenty-second Amendment to the constitu­
tion imposes a two term limit where there had been none. And in a handful
of countries, there remains no limit at all. Table 5.2 groups systems with
elected presidencies according to their rules on reelection.
What may account for these variations among systems on provisions for
presidential reelection? Bans or restrictions on reelection are typically per­
ceived as a check on presidential power, or, more specifically as a safeguard
against an executive's self-perpetuation in office. A second concern ha�
been articulated clearly by supporters of the Twenty-second Amendment to
the U.S. Constitution, who feared that the executive branch, which had
grown enormously during the thirteen years of President Franklin Roose­
velt's administration, could grow uniquely responsive to a multiple-term
president. This point is applicable to other countries as well, where presi­
dents have occasionally served even longer than Roosevelt. We might ex­
pect that variations would reflect the nature of those elites whose accep­
tance was necessary to put into effect the constitution.
In Mexico, "effective suffrage, no reelection" was a battle cry of the
revolution against the dictatorship of Porfirio Diaz. Given that no cohesive
party emerged out of the early revolutionary struggles, the adoption of
absolute no reelection not only signified the codification of revolutionary
rhetoric but also served the interest of the many competing factions within
the new leadership. The chances that all of them could one day take turns
in the presidential office would be enhanced by a provision that ensured
that any one caudillo's term in office would be his last.
Similarly, another source of absolute no reelection might be a situation of
fragmentation, in which several parties expected to have a chance at the
presidency but wanted to ensure that one would not, by virtue of winning
 Constitutional origins of chief executives 89
one election, be able to assume a dominant role. 9 Ecuador and Guatemala
fit this scenario.
At the other extreme are systems with no restrictions on reelection. Any
leader who by himself could prevail on the question of whether or not to
incorporate in a new constitution a provision on reelection would be ex­
pected to prefer constitutional silence on the issue. Thus, if one caudillo
dominates the process of forming the new regime, reelection may be permit­
ted with no restrictions. In the Dominican Republic, the dominant force at
the time of the constitution writing in 1962 was the Dominican Revolution­
ary Party, a tightly organized quasi-Leninist party headed by a caudillo,
Juan Bosch. Bosch's dominance at the time may explain the lack of restric­
tions on reelection in the Dominican constitution. If so, this is highly
ironic, as it is another caudillo, Joaquin Balaguer, a onetime figurehead
president under the dictatorship that Bosch fought against, who has bene­
fited from the ability to be elected again and again. Balaguer has been
elected five times since 1966.
Nicaragua's constitution of 1986 also provides no restrictions on the
number of terms a president may serve. Perhaps this case is also explicable
by the dominance of one leader within a party. While the Sandinista Front
was sometimes portrayed as having had a collegial leadership, the leading
role of Daniel Ortega (as "Coordinator" of the revolutionary junta) had
been established even before a formally presidential system was put in
place. A collegial party would be expected, as in Mexico, to opt for no
reelection, while Ortega's dominance since 1979 explains why other party
leaders would accept a constitution which would theoretically permit Or­
tega to remain president for many terms.
There are cases of Leninist or otherwise tightly organized parties not
imposing constitutions allowing perpetual reelection. In Venezuela, the
Acci6n Democratica (AD) by 1961 had accepted consensus among parties as
critical to the maintenance of democracy (Levine 1973) ; thus it did not insist
on a constitution that would permit long-run rule by its maximum leader,
R6mulo Betancourt, but would have allowed him to serve again after two
terms had elapsed, had he lived long enough. Compare the current Venezue­
lan constitution with that of 1947. The earlier constitution, adopted when
AD was hegemonic, allowed reelection only after a single intervening term ,
rather than two terms, as in the 1961 document. The reason for the differ­
ence may be found in AD's choice of a nonpartisan and famous novelist as its
candidate in the first democratic presidential election, in 1945. Preventing
this president, R6mulo Gallegos, from running for immediate reelection
would have made Betancourt the easy choice for AD in 1950 (and perhaps
1960) had a military coup not intervened. By 1961, with Betancourt already
having been elected in 1958, a rule requiring two intervening terms must

9 We are indebted to Barbara Geddes for spurring us to think of these issues.

90 Presidents and assemblies
have been desirable, not only to other parties but also to younger AD
activists who had fought the dictatorship at home while the older leadership,
including Betancourt, had been in exile (Levine 1973).
In Namibia, the majority party in elections to a constituent assembly in
1989 was a former guerrilla movement, the South-West Africa People's
Organization. SWAPO would be expected to be in the same category as the
Sandinistas and the Dominican Revolutionary Party, having tight central
leadership. It would thus be expected to opt for no provision prohibiting re­
election. However, the UN-imposed rules under which the constitution was
drafted required a two-thirds vote in the constituent assembly to approve the
constitution (Shugart 1992b ). SWAPO lacked such a large majority, and thus
had to defer to other parties, which feared SWAPO leader San Nujoma's
potential for perpetuating his rule. Thus Namibia has a two-term limit.
Limitations on reelection, however, have effects on presidential govern­
ment beyond simply forcing turnover in the executive. First, a no-reelection
rule prevents the electorate from returning to office a popular president.
This in itself is a loss for the public interest because an equally competent
successor may not always be available. Moreover, by removing the possibil­
ity that the president will face judgment at the hands of the voters, prohibi­
tions on reelection eliminate the only link by which the executive in a com­
petitive political system is directly accountable to the electorate. Voters may
make the initial decision as to who leads the executive - and in this sense
their impact on the nature of the branch will likely be greater than in most
parliamentary systems - but presidents barred from reelection know they
will not answer to the voters at the end of their term.
A final effect of the no-reelection rule relates not to the president's rela­
tionship with the electorate, but to that with congress. As many critics have
pointed out, presidentialism offers scarce incentives for congresspersons -
even those of the president's party - to support the executive. A president
who is immediately a lame duck can offer no coattails to potential supporters
in congress, and so will likely have even greater difficulties soliciting coopera­
tion among congresspersons than one who can be reelected. This difficulty
may exacerbate the problem of immobilism that leads to criticisms of presi­
dentialism (Coppedge 1988).
None of the arrangements regulating reelection violate the essential crite­
ria for presidentialism we have established. Prohibitions on reelection do
not prevent the executive from being shaped by popular vote, and certainly
do not threaten the independence of the executive from the congress for its
survival. Nevertheless, in our opinion, placing no limit on reelection is
most in keeping with the principles of presidentialism. Presidentialism re­
quires that the chief executive be elected by popular vote. A logical corol­
lary of this statement would be that there should be as few restrictions as
possible on the will to be expressed in that vote. In addition, we have
already argued that a contingent advantage of presidential democracy is the
 Constitutional origins of chief executives 91
increased accountability it provides of the executive to the electorate. To
prohibit reelection erases that characteristic of presidential government
and the advantage it provides. Finally, the concern about incumbent presi­
dents' misusing the office to electoral advantage - while often valid - is
diminished to the extent that a system has competitive elections. Although
fraud and intimidation are still serious problems during elections in some
presidential systems, in recent years, many more than ever before have
reliably neutral, multiparty agencies to conduct elections. Where electoral
fairness is less well institutionalized the presence of international observer
groups, United Nations oversight, and the computerization of ballot counts
have all contributed tremendously to deterring or exposing fraud by incum­
bents. W ithout implying that presidential elections are uniformly beyond
reproach, we only suggest that elections are more likely to become progres­
sively more fair rather than more corrupt. Thus, prohibiting reelection to
prevent fraud may be an unnecessarily large sacrifice.

THE VICE - PRESIDENCY , OR NOT

The question of whether or not to have a vice-presidency, and if so, how to

fill the post, is important to the question of the origin of chief executive
power in the event of a vacancy in the presidential office. The office of the
vice-presidency is subject to perpetual criticism - in the United States at
least - for a couple of reasons. In the first place, pundits claim to struggle
in vain to discern exactly what it is that vice-presidents do, once in office.
Second, vice-presidential candidates are widely thought to be chosen as
presidential running mates for reasons unrelated to their most important
function: serving as president if the office is vacated before the end of the
term. Certainly a presidential contender chooses his or her running mate
(or, in some cases, running mates) to maximize the ticket's strength at the
polls, and the value of a vice-president in this regard does not necessarily
correspond to qualifications to serve as president.
Nevertheless, our project here is to demonstrate that the office of vice­
president, and the means of filling vacated presidencies more generally, can
be arranged in various ways. W hat problems, if any, a vacated presidency
causes depends on the path to succession. In addition, the vice-presidency
may be organized in such a way as to encourage compromise among con­
tenders for control of the executive, to allow presidential contenders to
signal to the electorate their political inclinations, or to ensure electoral
legitimacy for any presidential replacement. Other systems simply have
opted not to have a vice-presidency at all - in the event of a vacancy, a new
presidential election is held, in most cases.
The first decision in organizing the path to succession is simply whether
the vice-presidency should exist or not. When there is a prime minister, the
nearly unanimous choice is no, although the 1991 constitution of Bulgaria
92 Presidents and assemblies
provides for both a (constitutionally powerless) president and a vice­
president. Peru likewise has all three posts. Some strictly presidential sys­
tems, including Venezuela, Mexico, Paraguay, and the Chilean constitu­
tions of 1925 and 1989, also do not provide for the office. The absence of
the vice-presidency obviously means that voters electing a president can
have no idea of who will fill that office if it is vacated during the regular
constitutional term. However, holding a new election means that, in the
event of a vacancy, voters again enjoy the advantages of choice and identifi­
ability inherent in any presidential election. Ironically, in a few cases that
are otherwise presidential the path to succession is often more parliamen­
tary (or, more accurately, akin to the assembly-independent regimes dis­
cussed earlier), in that voters are excluded from influencing the choice of
the interim chief executive, who is chosen by the congress.
In Mexico, if more than four years remain in the presidential term when
the office is vacated, then an interim is chosen by a joint session of congress
to serve for fourteen to eighteen months, at which time a popular election
is held to elect a president to serve the rest of the original term. If less than
four years remains when the office is vacated, however, the interim chosen
by congress serves out the remainder of the term, and no special election is
held. Colombia's constitution of 1886-1991 had a similar provision. 10 Like­
wise in Paraguay, if more than half of the original term remains, new
elections are held; but if less than half remains, congress selects the succes­
sor. In Chile 1989, the successor is chosen by the senate to fill out the
remainder of the term, regardless of when the presidency is vacated. In
Venezuela, a similar choice is made by a joint session of congress. And in
unicameral Sri Lanka, the assembly chooses the successor as well. 1 1
Among the systems without vice-presidents, all currently provide that
newly elected or appointed presidents shall serve only the remainder of the
original term - not a full new term. However, the 1925 constitution of
Chile provided that the replacement serve a new six-year term. We regard
as a desirable characteristic the provision for simply filling out the original
term in order not to disrupt the electoral cycles of the systems. We discuss
the importance of electoral cycles at length in Chapter 11. For now it
suffices to say that even though a successor elected to fill a vacated presi-
10 In Colombia, the Congress selected a designado, or designated successor, every two years.
while the presidential term was four years. This unusual arrangement allowed the choice of
successor to respond to shifting congressional coalitions, although not as readily as if the
designado were chosen after the presidency is vacated. On the other hand, the Colombian
system provided that there would be a successor prepared to fill the presidency immedi­
ately, if need be.
11 Finally, there is the consistently unique Bolivian formula, in which presidential and vice­
presidential candidates run on a team ticket during the popular election, but where none
receives a majority the offices are filled separately by Congress from among the top three
candidates for each. Thus, the Congress may mix and match from among the tickets, so as
to construct a coalition executive.
 Constitutional origins ofchief executives 93
dency might win a strong mandate at the polls, such a successor should
serve no longer than the duration of the original term.
If the office of vice-president is to exist, the next decision is whether it
should be filled jointly with the presidency or separately. Most systems
elect both president and vice-president on a joint ticket; voters cast one
vote for a team of candidates. W here this is the case, and where the
president chooses her or his running mate, the choice can serve as a signal
to voters as to the political inclinations of the candidate for chief executive.
Voters can assess both the sagacity and the coalitional instincts of the
presidential candidate by his or her choice for vice-president. On the other
hand, separate election of the vice-president provides for maximum influ­
ence of the electorate over the line of succession to the presidency, and so
prevents the possibility of a successor who cannot claim electoral legiti­
macy. This is currently the case in Peru and was so in the Brazilian constitu­
tion of 1946 and the Philippine constitution of 1935.
If the elections for president and vice-president are on separate ballots, it
is possible (perhaps likely) that the two officials will be of different parties,
perhaps even one the other's opposition. Such a situation occurred most
prominently in Brazil in 1960 when Janio Quadros was elected president,
while the far more radical Joao Goulart was elected vice-president. When
Goulart succeeded Quadros in 1962, a regime crisis resulted that ultimately
led to the 1964 coup. Despite that experience, is there any reason why the
president and vice-president must be politically compatible? As long as the
vice-president assumes no executive functions until the permanent absence
of the president, the answer is no. Providing separate elections at least
allows voters directly to choose not only the president but also the
president-in-waiting, should a successor be necessary. 12 The primary advan­
tage of using a fused ticket for president and vice-president is in multiparty
systems, in which parties can form coalitions for two offices more easily
than for one. The ticket arrangement does, however, lessen somewhat the
connection between voter choice and chief executive in the event that the
vice-president must assume the presidency. Perhaps the best solution is no
vice-presidency at all, with a special election being held to fill the remain­
der of the term when there is a vacancy.

D I V I D E D PO L I T I E S A N D C O L L E G I A L PRESIDENCIES
Chapter 3 noted that one of the criticisms of presidential systems has been
their tendency toward majoritarianism in the executive. W hile the most
12 Going a step further, the elections need not even occur at the same time. Colombia's
constitution of 1832 provided for nonsynchronized terms, with the vice-president elected
at the president's midterm (Gibson 1948). This arrangement does strike us as undesirable,
as it is preferable for the two executive-level offices at least to emerge out of a common
campaign and common voter concerns.
94 Presidents and assemblies
majoritarian democratic political system model is surely a form of parlia­
mentarism - the Westminister model in which one party ordinarily cap­
tures a majority of seats and thus forms its own government - in polities
using proportional representation, the divide between parliamentarism
and presidentialism is more stark. With PR, most parliamentary execu­
tives will be coalitions, or else minority cabinets dependent on support of
other parties in parliament. However, even with PR and in the absence of
a majority party in the assembly, most presidential executives are com­
posed only of those ministers whom the president chooses. In many cases
this means members of only the president's party or closely associated
parties. For this reason presidentialism is criticized as majoritarian, de­
spite varying degrees of institutional checks placed on presidents and their
cabinets. Of course, the greater the checks, the more the risk of
executive-assembly stalemates or even crisis and the less the chance that
the popularly endorsed executive program will be enacted.
In the last two chapters we have discussed one type of constitutional
design, premier-presidentialism, that seeks to address the problems of
(pure) presidentialism while still maintaining the perceived advantages of a
directly elected president to serve the various roles of head of state 13 and
arbiter, and as an expression of the broader electorate in a plural political
context. However, what if a political system is so divided that placing a
presidency in the hands of any one political tendency might itself be
destabilizing? The assembly-independent regime is one solution, but one
that does not allow direct voter choice of the executive. If the response is to
design either a weak presidency or an electoral format that minimizes the
chances of a presidential majority in the assembly, we risk losing one of our
conditional advantages of having a directly elected presidency in the first
place, which was to allow voters to identify alternative possible govern­
ments, and then choose one and hold it accountable. Can these goals be
maintained without risking a majoritarian executive? We think that the
answer is yes, as a variant of presidentialism can be designed that makes for
an executive both collegial and accountable.

The collegial executive under presidentialism

As we discussed in Chapter 2, some definitions of presidentialism exclude
the possibility of a multi-person presidency. The question may be framed
in terms of whether or not it might be contradictory to have a system
based on checks and balances but having a unitary executive. Lijphart
(1989) suggests that it is contradictory, but that this internal contradiction
13 Of course, a system could be a pure presidential regime, by our definition. even if the
directly elected head of government were not also the head of state. It is almost inconceiv­
able, however, that a premier-presidential regime would have some figure other than the
elected president as head of state.
 Constitutional origins ofchief executives 95
of majoritarianism versus consensualism is inherent in presidentialism. By
this logic, a collegial "presidency" must be an entirely separate type of
regime. Aside from the taxonomical debate, our idea that a regime can
incorporate both separation of powers and a multi-person executive is
most forcefully challenged by Alexander Hamilton. Thus we now turn to
Hamilton's words expressed in the debate over ratification of the U.S.
Constitution.
In Federalist No. 70, we see quite clearly the tension between Hamil­
tonian and Madisonian democratic theory - a tension manifest in this in­
stance in the issue of whether a plural executive is compatible with presiden­
tial government. Hamilton's answer is an emphatic no. He makes the case
that the principle of checks and balances should not apply within the execu­
tive branch. Hamilton thus distinguishes the internal organization of the
executive from that of Congress - with its bicameral structure and multiple
opportunities for minority vetoes - and of the federal system as a whole.
He justifies this distinction arguing that while caution is desirable in the
legislative process, the execution of the laws requires singlemindedness and
an "energy" that can only be supplied by a unitary presidency (The Federal­
ist Papers 1961:423-35).
This distinction between the executive and the rest of the American
constitutional structure is based, ultimately, on the presumption that the
power of the executive in shaping policy is small relative to that of
Congress - that the executive is an executor of laws with little discretion
over their content. Insofar as the executive is a mere executor, Hamilton's
logic is compelling. To the extent that the executive role transcends this
limited sphere, however, the logic of checks and balances would seem to
apply to the executive equally as it applies to Congress. In practice, the
executive in some presidential systems is accorded powers of decree, at
least for purposes of rule making (see Chapter 7) - and almost universally
is active in drafting and initiating actual legislation. Given the uniformity
with which presidential executives exceed the role as mere executor of
legislation, however - even while remaining well within their constitu­
tional powers - Hamilton's vision of the presidential executive is somewhat
limited. In the context of modern presidencies it is not at all clear that
Hamilton's argument for the unitary executive agrees with the larger princi­
ples of presidentialism elaborated elsewhere in The Federalist.
There is a second area in which Hamilton's condemnation of the plural
executive seems unfounded. He argues that such an arrangement would
"split the community into the most violent and irreconcilable factions,
adhering differently to the different individuals who composed the ma­
gistry" (The Federalist Papers 1961:426). The logic here runs counter to the
consociational argument that a plural executive can be created to guaran­
tee the representation in that branch of minority groups which would be
excluded by a majoritarian executive. The consociational logic holds that a
96 Presidents and assemblies
plural executive can foster compromise and cooperation (Lijphart 1977),
bridging cleavages rather than fostering them.
Before dismissing Hamilton, we want to acknowledge his most persua­
sive argument in favor of the unitary executive: the case for accountability.
We argued earlier that the popular election of the executive provides supe­
rior accountability of the elected to the electors in presidential over parlia­
mentary systems, although this accountability is undercut if a president is
constitutionally barred from reelection. Hamilton's point extends our argu­
ment considerably - in a way that is valid under some conditions but not
under others. Hamilton contends that when responsibility is shared, it is
more difficult to determine where the credit or blame for executive action
lies. By this logic, even where popular election may enhance accountabil­
ity, a plural executive might undermine it. But we can imagine cases in
which power sharing is desirable in the executive, where multiple or deep
political cleavages in a society mean that a unitary presidency will necessar­
ily be unrepresentative of a large proportion of the population. These are
the conditions Lijphart identifies as requiring consociational government,
although Lijphart concludes that the collegial executive is viable only under
parliamentarism. We, on the other hand, maintain that, where power shar­
ing is desirable, a collegial presidency need not be so constructed as to
reduce accountability. Indeed, where a unitary but highly unrepresentative
executive can win office, its accountability must also be suspect.
In sum, there may be presidentialism with a plural executive. Such an
executive maximizes checks and balances, and may provide superior repre­
sentation to minority groups in deeply divided societies. The level of ac­
countability to which the plural executive can be held will depend on the
cleavage structure of the society. Collegial executives may make the attribu­
tion of credit and blame more complicated, but where a unitary executive is
not representative of its society, neither can it be held accountable. Finally,
where checks internal to the executive are unrestrained, the plural execu­
tive may increase a system's tendency toward immobilism. We shall return
to this problem after considering the performance of some experiments
with collegial presidencies.

Actual popularly elected collegial executives

Perhaps the most famous example of a collegial executive is the Swiss
Federal Council, which was discussed at the beginning of this chapter. The
Swiss system is not, however, presidential, since the executive is chosen by
the assembly. Two countries have used popuarly elected collegial execu­
tives: Uruguay from 1952 to 1966, and Cyprus from 1960 to 1963. The
Uruguayan experience was specifically inspired by the Swiss example, even
though the origin of executive power in the Uruguayan case and its
"model" differ fundamentally.
 Constitutional origins of chief executives 97
Our first step in this section is to provide a brief overview of the Uru­
guayan and Cypriot collegial presidencies. We wish to indicate at the out­
set, however, that our reason for discussing these systems is not that we
consider them to be models to be applied elsewhere. In fact, in many ways
they represent models of what is to be avoided in designing collegial presi­
dencies. After reviewing the Uruguayan and Cypriot experiences, we shall
discuss alternative ways of constituting plural executives under presiden­
tialism that will stand in contrast to the institutional designs employed in
those two countries.

Uruguay. Uruguay's experiment with its colegiado was the belated result
of the Jose Batlle y Ordofiez's long effort in the early twentieth century to
establish a collegial executive based on the Swiss formula, which he ad­
mired. Troubled by the personalistic and caudillo nature of Uruguayan
politics, Batlle's idea was to let all the caudillos be president at the same
time (Taylor 1960:153), in much the same manner that the Swiss formula
ensures that none of the major players is shut out of the executive at any
time. The 1918-33 Uruguayan constitution provided for a dual executive,
with a unitary presidency existing beside a nine-member National Admin­
istrative Council. This was the closest Batlle was able to come to achiev­
ing a colegiado in his lifetime. But even this arrangement was not a truly
plural presidency, since executive responsibilities were sharply divided
between the two arms of the executive, rather than shared (Gillespie
1989:5-6).
In the early 1950s, however, debate over the collegial executive once
again dominated Uruguayan politics. This time, the idea was championed
by Jose Batlle's sons, Lorenzo and Cesar, leading a faction of the Colorado
Party known as List 14. In the 1950 election, List 14 finished behind Colo­
rado List 15, headed by Luis Batlle Berres, who assumed the single-person
presidency. Jose Batlle's sons turned to Blanco Party leader Luis Alberto
de Herrera for support. Herrera had made his sixth unsuccessful run at the
presidency in 1950, and had won ninety-two thousand more votes than Luis
Batlle Berres, but lost the office under Uruguay's electoral rules because
his party did not win a plurality. In an effort to prevent Batlle Berres from
controlling a powerful presidency, Herrera threw his support behind the
efforts of Colorado List 14 to establish a fully collegial executive. A special
constitutional convention ratified the new arrangement in 1951, the follow­
ing year forming a popularly elected executive consisting of six members of
the party winning the presidential election and three members of the
second-place party. For the heirs of Jose Batlle y Ordonez, the colegiado
represented an effort to fulfill the vision of their father, who distrusted
caudillo rule. But the colegiado could not have been established without
the support of Herrera and the Blancos, who originally opposed it (Marotta
1985:36). For them, the plural executive was an opportunity to reduce the
98 Presidents and assemblies
power of Batlle Berres and to lay claim to a guaranteed share of executive
power. 1 4
If the conditions under which the colegiado was established were com­
plex, so must be any assessment of its performance. Early in its tenure, the
colegiado seemed to be serving the goal of fostering interparty coopera­
tion. At the end of 1952, Dr. Andres Martinez Trueba, a Colorado member
of the executive and former President of Uruguay, praised the colegiado as
more efficient than the single-member presidency. In particular, Martinez
pointed out that the opposition's presence in the executive prevented it
from attacking all executive actions in order to make political capital from
any controversial decisions. Martinez contended that the colegiado had
been able to "act firmly" in the face of illegal strikes by government employ­
ees "keeping control in the hands of the Government where the unions
were unreasonable" - all without prompting accusations of dictatorial ten­
dencies from a disenfranchised opposition . 15 Because the secondary party
was not represented with parity in the executive, it could, of course, have
been outvoted by the majority party. Therefore its participation in an
executive council in which one party held a majority of seats could not
guarantee that its views would be taken into consideration. On the other
hand, the ability of the secondary party to distance itself publicly from the
majority party is diminished by its being part - even a junior part - of the
executive. Notice that in Martinez's account, he praises the collegial execu­
tive not for having moderated the actions of the majority party, but rather
for having restrained the criticisms of the minority. This feature of the
Uruguayan system is not one we should want to endorse - to be truly a
power-sharing presidency, no one party should hold a majority of the
seats - but Martinez's comments certainly contradict the argument that a
plural executive must necessarily be indecisive, or "feeble" in Hamiltonian
terms. By Martinez's account, energy in the executive and interparty com­
promise are not at all incompatible.
Other accounts with the advantage of greater hindsight, however, have
concluded that the nine-person executive committee was unwieldly. In
1960, Taylor intended an unflatterinng comparison when he likened the
colegiado to a third legislative chamber, where extended debate slowed
down decision making, ultimately impeding the administration of govern­
ment programs (Taylor:103-4). While one similarly could call any unitary
presidency that possesses a veto a "third legislative chamber" that impedes
14 The static 6-to-3 division of seats in the colegiado was insisted on by the Blancos. This plan
was a move away from proportional representation in the executive . as it made rigid the
distribution of partisan power regardless of who controlled the majority of seats, and
excluded all but the top two parties from any share of executive power. This latter condi­
tion came under attack from the leftist segment of the electorate that was growing in the
1960s.
15 "Uruguay's System of Ruling Praised," New York Times, December 18, 1952.
 Constitutional origins of chief executives 99
(or immobilizes) the governing process, Taylor's comments provide an
echo of Hamilton's warnings: Applying checks and balances within the
executive might sap the energy of day-to-day administration. More specifi­
cally, Needler pointed out that the difficulty in reaching decisions within
the plural executive forced Uruguay twice in the early 1960s to send delega­
tions to inter-American conferences uninstructed on national policy (Need­
ler 1965: 156).
In addition to these political difficulties, Uruguay suffered an extended
economic decline beginning in the 1950s. By the mid-1960s, both the need
for government austerity and the inability of the now Blanco-dominated
executive to adhere to its economic program were attributed at least in part
to the country's unorthodox executive design, although such claims appear
to be more a result of scapegoating than of careful analysis. The List 15
Colorados remained determined opponents of the colegiado, and they were
joined by the Blancos whose support had earlier been essential to its adop­
tion. By 1966, a constitutional reform that included the return to a unitary
executive was on the national ballot and was supported by 62% of the
Uruguayan electorate (Gillespie 1989:5-6).

Cyprus. Cyprus's experiment with a plural executive was even more ephem­
eral, and considerably less successful, than Uruguay's. On the other hand,
the problems that Cyprus's dual presidency sought to resolve were far more
daunting. Cyprus gained independence from British colonial rule in 1960,
under a constitution that sought to guarantee shared political power be­
tween the Greek majority composing 78% of the population, and the Turk­
ish minority composing 18%. The 1960 constitution established a dual
executive structure with a Greek president and a Turkish vice-president,
holding almost equal powers and elected separately by the Greek and
Turkish Cypriot communities. Similarly, in the cabinet and the national
assembly, seats were to be divided on a 7:3 Greek-Turkish ratio, as was
membership in the police, army, and civil service. Further, the constitution
established separate assembly chambers by which the two ethnic groups
could govern religious, educational, cultural, and personal status matters.
Finally, both the president and vice-president had ultimate veto power over
all legislation concerning foreign affairs, defense, or security; and any tax
legislation required concurrent majorities of both the Greek and Turkish
caucuses in the assembly (Kyriakides 1968). 16 In short, the 1960 constitu­
tion implemented a host of consociational measures in addition to the
plural executive, including disproportionally high representation of minor­
ity Turks in government offices, communal autonomy, and minority veto
(Lijphart 1977: 158).
16 For a detailed description of the provisions of the 1960 constitution, see Kyriakides (1968),
ch. 3.
100 Presidents and assemblies
The Cypriot dual presidency and the other consociational features of the
constitution, however, failed to accommodate the level of political conflict
between the island's ethnic adversaries. Lijphart suggests a number of
reasons why this was so. In the first place, consociational arrangements
with extensive power-sharing and minority vetoes are most feasible where
no single group constitutes a clear majority of the population. Such a
majority is likely to see consociationalism as unnecessary and discrimina­
tory, according minorities excessive negative powers. This attitude was
prevalent among Greek Cypriots, and the status of the Turkish vice­
president as coequal to the president was a particular target of criticism
(Polyviou 1980:20). Second, Lijphart notes that the Greek and Turkish
populations were homogeneously interspersed among Cyprus's six dis­
tricts, complicating the construction of any federal structure that might
have allowed for more complete self-rule according to geographic divi­
sions. Third, there existed at the time of indepdendence virtually no con­
ception of or loyalty to Cypriot national identity; instead, Greek Cypriots
understood themselves to be Greek, and Turkish Cypriots saw themselves
as Turkish (Lijphart 1977:159-60).
A final point is that the conditions for Cypriot independence were
negotiated not primarily by Cypriot leaders themselves but by the Greek,
Turkish, and British governments, during talks in London and Zurich in
1959. Thus the legal frameowrk of the new republic was constructed by
outside forces, and presented to the Cypriot leaders, who accepted it as a
package in 1960 (Polyviou 1980: 13). This process did nothing to instill a
commitment to the institutional arrangements among Cypriot leaders.
And in fact, the consociational spirit of the 1960 constitution was never
reflected in the political process itself. The Turkish minority in govern­
ment used its veto on legislation to press demands outside the legislative
arena (Lijphart 1977:160). And the Greek president, for his part, had
presented by November 1963 a set of thirteen constitutional amendments
that would have eliminated outright most of the consociational provisions
protecting the Turkish minority (Ertekun 1984:11). Civil war broke out on
the island within days, destroying the Cypriot experiment with consocia­
tionalism and a dual presidency.

The experiences of Uruguay and Cyprus are not especially encouraging if

one wanted to make a case for a collegial presidency. However, the specific
institutions employed in these cases may have been well short of optimal,
and the conditions they were expected to deal with may have overwhelmed
any system, at least in Cyprus. The failure in Cyprus can be attributed less
to its effect on assembly-executive relations (although its performance
here was by no means satisfactory) than with vehement opposition by
ethnic groups to an entire government system arranged primarily by out­
side forces. In Uruguay, the colegiado was blamed for stalemate in both the
 Constitutional origins of chief executives 101
legislative and administrative process. However, chronic standoff between
Colorados, Blancas, and leftist parties was a problem in the late 1960s,
even after the abandonment of the colegiado; so it is difficult to see how
this institutional arrangement was uniquely responsible. Next, let us con­
sider some problems of collegial presidencies that are suggested by the
Uruguayan and Cypriot cases.

Pitfalls of actual collegial presidencies

Having reviewed the actual experiences of two countries with plural presi­
dencies, we now turn to alternative arrangements. We identify two pitfalls
of these collegial executives:
1 . restricted access, in the case of Uruguay; and
2. deficiencies in accountability and identifiability, in both cases.

Restricted access. In Uruguay, even though there were nine seats available,
access to the executive was restricted to the two largest parties. This restric­
tion became an issue as a third party gained in strength in the 1960s.
Ironically, the method chosen to redress this issue was not to open up
access to the third party, but rather to abolish the council altogether in
favor of a unitary executive elected by plurality. 17 In this manner, the two
largest parties ensured that executive power would remain in their own
hands.
The restriction on access to the collegial presidency runs counter to the
ostensible consociational design of the system. It would be preferable to
have access restricted only by the threshold set by the magnitude of the
district and the formula employed, as with assembly elections in democra­
cies. For instance, with M = 9, in a hypothetical distribution of vote shares
of 40-35-15-10, the four parties would split the nine seats 4-3-1-1 if the
common procedure of simple quota and largest remainders were used, 18
instead of Uruguay's 6-3-0-0.
Besides shutting out all parties other than the leading two, having a fixed
ratio of seats in the council, regardless of the votes, meant that representa­
tion in the council could not be reflective of shifts in the support of the two
parties (except when one surpassed the other), any more than it could
reflect the growing support of a third party (again, unless that party sup­
planted one of the others). 19 This deficiency gets us to the next set of
concerns, those of accountability and identifiability.
17 More precisely, the presidency was once again elected by double simultaneous vote.
18 And, of course, the possibility of such small parties as in the example being in the execu­
tive could be avoided, if desired, by simply lowering the magnitude.
19 Had the latter occurred, we suspect that the two traditional parties would have had
considerably greater interest in opening up representation on the council to parties other
than the two largest.
102 Presidents and assemblies
Deficiencies in accountability and identifiability. Both elected collegial ex­
ecutives suffered from deficient accountability and identifiability. Voters in
Uruguay could not easily express any policy direction in their executive,
partly because of the fixed ratio of representation. To make matters worse,
the electoral rules used allowed internal factions of a party to be the real
competitors. The large spoils available to each party (6 seats to the larger
and 3 to the smaller) virtually guaranteed fragmentation while making
turnover difficult.
In Cyprus, the president and vice-president were effectively co-presi­
dents, as noted. Even though each had almost equal powers, each was
accountable to only a portion of the electorate. Greeks and Turks formed
separate electoral rolls and separate constituencies for the election of an
executive with powers over all citizens. W hile it might be desirable, ac­
cording to the theory of consociational democracy, to ensure that only
Greeks could designate the Greek member of the executive and only
Turks the Turk, this format also made it difficult for voters to express any
other policy preference for the nation as a whole, other than protection of
their ethnic identity.
We do not agree with Hamilton that collegial executives necessarily en­
courage divisiveness in the polity. However, by enhancing ethnicity as the
source of electoral support for "co-presidents" elected in virtual isolation
from one another, the Cypriot format may indeed have encouraged divisive­
ness. From the standpoint of diminishing the salience of this cleavage in the
national election for the executive (as distinct from assembly elections), as
well as for the sake of accountability, it would be preferable to require the
prospective Greek and Turk co-presidents to stand together before the
whole electorate.

Alternative designs
The deficiencies of the Uruguayan and Cypriot cases suggest that, where a
collegial presidency is desirable as a conflict-regulative device, there might
be alternative designs that would function better.

PR in the executive. If we want to ensure minority representation in the

executive, one method is simply to use proportional representation to elect
the multi-seat council. This possibility was suggested in the context of our
discussion of the problem of restricted access in Uruguay. There are empiri­
cal examples of such collegial presidencies, although at the national level
there never have been any that we are aware of. The actual cases are in
some of the Swiss cantons. All Swiss cantons have collegial executives, but,
modifying the Executive Council that exists at the federal level, some
cantons directly elect their executives. Thus, unlike Switzerland as a whole,
they have, by our definition, presidential systems.
 Constitutional origins of chief executives 103
W hile the use of PR in the executive redresses the problem of excluding
minorities, it does so at a cost: undercutting accountability and, especially,
identifiability. As we have made a theme of this book, one purpose of an
executive is to articulate policy alternatives for the nation as a whole. If this
executive role is desirable, then an executive elected by PR will most likely
be undesirable.

Co-presidential tickets. A possibility that we prefer over the use of PR in

the executive is to create a co-presidency, as in Cyprus, but to have both (or
perhaps three) co-presidents elected on a common ticket by the whole
electorate. If a ticket of co-presidential candidates were elected by a
method that encourages the formation of broad preelectoral coalitions (see
Chapter 10), there would be a powerful incentive for parties to seek tickets
that spanned across the blocs that divide the party system. Since the pro­
spective co-presidents would have to stand or fall electorally as a ticket, it
would be feasible for them to present a program before the electorate. On
the other hand, using an electoral formula whereby the winners would be
the two or three most-voted-for candidates (or slates), running indepen­
dently, would make the task of presenting such a program inherently more
difficult. A format without tickets would also run the risk of having a co­
presidency made up of leaders of the more extreme parties, rather than a
coalition tending toward moderation.
An objection might be that the format we propose would not be faith­
fully representative. Critics might make such a claim on the grounds that if
an extreme is one of the two or three most popular options, it should win
one of the two or three executive seats. Our response is that we seek to
preserve the advantages of presidentialism stemming from the regime
type's ability to give voters a choice between alternative executives and
then give that executive ability to lead (subject, of course, to whatever
institutional checks might also be part of the constitutional design). This
proposed form of collegial presidency is meant to make the coalitions that
tend to emerge under plurality rule (and its variants) explicit. To illustrate
the point let us consider the example of Chile's 1964 election.
In Chile in 1964 the Christian Democrat Eduardo Frei was elected presi­
dent with 56% of the vote. Frei interpreted his victory as a mandate for
Christian Democracy, pursuing a single party administration despite the
fact that Chile's was a deeply divided political system the customary prac­
tices of which were coalition rule. More to the point, Frei's own victory had
been made possible only because the right-wing parties had sought to
protect themselves against what had nearly happened in 1958: a victory in a
three-way race by a socialist. Thus the right endorsed Frei rather than
present its own candidate.
Had Chile been using a co-presidential system at the time , the center­
right coalition would have been explicit. Frei and a conservative (assuming
104 Presidents and assemblies
a two-person presidency) would both have been running together and a
single-party cabinet and its policies would not have occurred. It is true that
this design might not have allowed voters a center-left coalitional possibil­
ity, since there would probably not have been any nonleftist forming a
ticket with the socialist Salvador Allende; but then, neither did the actual
unitary presidential election of 1964 offer a center-left option to Chilean
voters. In either format, voters would have been offered a choice between
a Socialist-Communist coalition or a center-right coalition. The difference
with the co-presidency is that this executive would have been more likely to
continue functioning like the (preelection) coalition it really was.
Besides preventing one party from assuming full executive power despite
a preelectoral coalition, a co-presidential format would make far less likely
the three-way races that preceded and followed the 1964 Chilean election.
With the winner-take-all stakes reduced, coalitions might well have been a
more attractive option in those campaigns. Indeed, by 1970, Chile's center
had been fully taken over by the Christian Democrats and the declining
Radicals moved leftward. W ith the different coalition-building incentives
of a co-presidency rather than a unitary presidency, a center-left alternative
ticket versus a continuing center-right co-presidency might have been of­
fered voters.

Size of the co-presidency. We mentioned earlier in this chapter that defining

an optimal size for a co-presidency was not a clear-cut task. We did, how­
ever, stress that a colegiado as large as that in Uruguay (nine) did not seem
desirable and that it should be as small as possible. A two-member execu­
tive eliminates - within the context of a collegial presidency - the problem
of unwieldiness that dogged Uruguay's experiment.
A two-member presidency does, of course, raise the potential for dead­
lock on some issues. The co-presidential option could be therefore modi­
fied to consist of a three-person rather than a two-person executive. With
three, presumably the executives would still seek consensus whenever possi­
ble, but in the event they could not achieve it, a 2-to-1 vote would prevent
deadlock. However, it should be stressed, in a system of great interparty
polarization, deadlock is not inherently bad. W ithout periodic deadlock,
there would be no incentive to make compromises. One other potential
plus for a three-person presidency is that it opens up more room for coali­
tion building, as a clearly stronger party would not have to make a smaller
party an absolutely equal partner. We shall not take a stand on which size is
best, but simply say that we suspect a collegial presidency with more than
three members would not ordinarily be desirable.

As we have suggested, a collegial presidency might be applicable to societies

divided by deep partisan or ideological divisions as well as societies that
conform more generally to what are denoted by the term "plural societies,"
 Constitutional origins of chief executives 105
those with ethnic or other subcultural divisions. For some multiethnic soci­
eties there may remain an appeal of electing a unitary president, but using a
method that guarantees the election of a moderate (see Chapter 10).

CONCLUSION

In this chapter, we have considered a series of variations on the means of

constituting chief executive power. Our survey has examined diverse meth­
ods of choosing executives, as well as variations on the structure of the
executive itself; and we have spanned both pure presidential and hybrid
regimes. Many of the cases represent deviations from the principle of
separation of the origin of assembly and executive powers, in particular
those in which selection of the chief executive is left to negotiation in the
assembly.
We have argued that fixed-term executives who are selected by assem­
blies sacrifice the ability of the electorate to determine directly the nature
of the executive; but nevertheless we acknowledge that such an arrange­
ment can reduce the dangers of winner-take-all presidential elections, and
foster coalition building under difficult political conditions. We have re­
viewed variations on eligibility for reelection among presidencies, conclud­
ing that to place no restrictions on reelection is most in keeping with the
principles of presidentialism that the electorate should choose its executive
free of constraint, and that the executive should in turn be accountable to
the electorate. We have also reviewed diverse means of organizing the
office of the vice-presidency, or, in its absence, paths to succession to a
vacated presidency. Here we pointed out the trade-offs between providing
for a successor at the outset of a presidential term versus waiting until the
presidency is vacated by some unforeseen event. Finally, we undertook a
more extensive analysis of collegial executives under presidential govern­
ment. Differing from other students of democratic institutions, and basing
our case on our understanding of The Federalist, we maintain that the
collegial presidency is not incompatible with the principles of presiden­
tialism. Indeed, we suggest that under certain conditions, collegial presiden­
tialism may offer the best possibiity for resolving conflicts peacefully in
deeply divided societies.
 6
Constitution al limits on sep ar ate
origin and surviv al

We turn now exclusively to regimes with single-person, separately elected

presidencies. While separation of the origin and survival of the executive
from the legislative branch was seen by the authors of The Federalist as
imperative, the notion of checks and balances implied to them some over­
lap of functions, even to the extent that the Federalists did not see Senate
"advice and consent" in the president's cabinet appointments as a violation
of the principle of separate powers. Among presidential systems, however,
this provision is rare: Complete separation of the origin of executive office
from the assembly is the norm. There are many other regimes, such as the
president-parliamentary systems, however, in which the principle of sepa­
rate survival is obviated.
In this chapter, we shall consider three means by which separation of
origin and survival has been limited in some systems:
1 . assembly involvement in the appointment of cabinet ministers ,
2. censure of cabinet ministers by the assembly, and
3. dissolution of the assembly by the president.
The latter two attributes, in particular, can entail sharp deviations from the
Madisonian ideal that each branch, secure in its independence, will without
hesitation check the other in the lawmaking process. Yet it can be argued
that the possibility of both censure and dissolution encourages cooperation
between branches, as each anticipates the likely reaction of the other to its
actions. We shall see that whether censure or dissolution produce such
cooperation and anticipation depends largely on the specific rules by which
each authority is exercised, and the incentives these rules create for politi­
cians. In particular, incorporating a theme from the preceding chapter, we
shall demonstrate that the impact of assembly censure authority on the
nature and stability of cabinets hinges on the process by which cabinet
members are originally appointed.

A S S E M B L Y I N V O L V E M E N T I N T H E A PP O I N T M E N T O F
C A B I N E T M E M B ERS
In archetypal presidentialism, the origin of the entire executive is divorced
from congressional influence. In a few presidential systems, nevertheless,
 Limits on separate origin and survival 107
although the president is chosen by popular election, appointees to cabinet
positions are subject to congressional approval. Congressional involvement
in the appointment of cabinet ministers appears at first to represent a threat
to the separation-of-powers principle as we have defined it. If an assembly
were in fact to make cabinet nominations, then we would no longer con­
sider the system presidential. In the cases we shall consider, however,
congressional powers in the appointment process are strictly negative. That
is, they are powers of approval or disapproval, but do not give the assembly
power of initiative in cabinet formation. Because the assemblies we discuss
cannot actively name the members of the executive, we do not regard these
cases as violations of the third criterion, and so as departures from presiden­
tialism. On the other hand, a requirement of advice and consent clearly
imposes some congressional influence on the construction of the executive.
It is worth reiterating that the modal arrangement of presidential systems
does not include congressional approval. In the vast majority of such sys­
tems, presidents appoint their cabinets with no congressional input. In this
regard, Korea 1987, Nigeria, the Philippines, and the United States deviate
from the presidential norm.
The power of initiative in the "appointment game" gives presidents a
definite advantage over congressional opposition in filling cabinet positions
with nominees who are acceptable to them. From observation of the ap­
pointment process in the United States one might easily induce that the
advantage is manifest simply in the fact that so few cabinet nominees are
actually rejected. We contend that the dynamic of congressional appoint­
ment is more subtle than that.
Consider the "appointment game" represented by Figure 6. 1, in which
we posit that the preferences of the president and the assembly (in the
United States, the Senate) for candidates for positions as cabinet ministers
(secretaries) can be mapped in one dimension. Point P represents the
president's ideal candidate for a given post, while point S represents the
Senate's ideal. At point Ip, the president is indifferent between filling the
post and leaving it vacant. The same is true for the Senate at point ls. That
is, the president would rather leave the position vacant and accept what­
ever political repercussions there are than to fill the post with a nominee
further from her ideal point, P, that is lp, And the same is true of the Senate
for any nominee beyond ls. If point Is lies to the left of point Ip, as depicted
in Figure 6.1, by definition, there would then be no section of the line
segment S-P along which a nominee could be acceptable to both players.
The president would never nominate a candidate acceptable to the Senate,
and the Senate would never accept a nominee chosen by the president. The
post would remain unfilled.
Now consider Figure 6.2, in which point Ip lies to the left of ls. All of the
points along line segment Ip-ls represent potentially acceptable nominees.
The president's goal is to nominate a candidate as close to her or his ideal
point as possible and still acceptable to the Senate. That candidate would
108 Presidents and assemblies

s p
Figure 6. 1 . Hypothetical indifference points for senate (S) and president (P): no
overlap, thus position would remain unfilled

s p
Figure 6.2. Hypothetical indifference points for senate (S) and president (P): presi­
dent nominates candidate at point ls

be located at the point ls. The Senate's goal, of course , is to secure an

appointment at point Ip. The president's advantage lies in the fact that the
power of initiative is the power to make a take-it-or-leave-it offer. 1 To the
extent that the president has accurate information about the location of the
Senate's indifference point, he or she will be able to nominate a candidate
approaching point ls. The Senate is powerless to secure a nomination at Ip
unless it can convince the president that Ip is the same as ls.
Despite the just-described power of initiative in the hands of the chief
executive, one might feel that the assembly still has both the right and the
ability to reject several nominees until the president finds one more in turn
with the preferences of the assembly majority. W hy should the assembly
(Senate in the U.S. case) settle for a nominee who is merely not objection­
able? The question becomes all the more important when we consider
those premier-presidential systems in which the president also names a
prime minister, who must be approved by the assembly. The appointment
game in those regimes suggests a critical contrast with the appointment
game in the United States. For example, there is the case of the French
Fifth Republic in 1986, when President Frarn;ois Mitterrand faced a just­
elected assembly in which his Socialist Party had lost its majority.
The French constitution gives the president the power to initiate a nomi­
nation to the post of prime minister. Mitterrand appointed the leader of the
major opposition party, Jacques Chirac, as prime minister, who was, of
course, confirmed by the assembly. W hy did Mitterrand appoint his princi­
pal opponent, when, given his powers of initiative in this matter, he could
have demanded that the assembly accept a moderate prime minister more
to the president's liking? Or, asked from the other perspective, why do
Republican presidents facing Democrat-controlled Senates in the United
1 For more on the importance of initiative in policy games between president and congress.
see Kiewiet and McCubbins (1985).
 Limits on separate origin and survival 109
States not act as Mitterrand did, nominating cabinet officials whose prefer­
ences conform to those of the assembly majority? The answer to this ques­
tion rests in the next phase of the game: W ho has power of dismissal. After
discussing censure in a number of cases, we shall see, in a section on
building and maintaining cabinets, that presidential initiative in the appoint­
ment game is merely a necessary, but not a sufficient, condition for the
president's ability to obtain ministers generally in accord with her or his
preferences. First, however, it will be helpful to consider the appointment
process in some presidential systems in more detail.

Experience in the United States

Clearly, the figures above are only heuristics; appointments cannot be
mapped neatly on two dimensions, and candidates would not be available
at every conceivable point on the plane even if they could be. But the
figures demonstrate nicely the strategic advantage that the power of initia­
tive provides the president in the appointment game, and the degree to
which incomplete or inaccurate information about the other player's prefer­
ences limits this advantage. Moreover, the figures demonstrate that there is
more than one possible explanation for why cabinet nominees in the United
States are rejected so infrequently. It has been suggested that the U.S.
Senate has forgotten or abandoned its power of "advice and consent"
(Hess 1988; Moe 1987, 1989). 2 We see no reason to believe that the Senate
would willingly abandon this formidable power and adopt a norm of con­
sent for appointment votes. Moreover, the periodic rejection of a cabinet
nominee indicates that senators have not forgotten their authority alto­
gether. To us, it is more plausible that the regularity with which nominees
are accepted merely demonstrates that U.S. presidents generally have
pretty good information regarding Senate indifference, or else generally
are cautious in naming candidates closer to the Senate's ideal point than to
its indifference point, in order to avoid political battles.
Nevertheless, there is some evidence that reforms in the nomination
process for presidential candidates - most notably the dominance of prima­
ries and the corresponding reduction of influence for party leaders - have
forced acknowledgment (in words, if not by vote) of a president's preroga­
tive in assembling a cabinet that is his personal political resource (see
Polsby 1983, ch. 3). Here, the tension is quite apparent between the presi­
dential ideal of separation of powers and the involvement of congressper­
sons in the construction of the executive. To the extent that Senators ac­
knowledge the president's right to fill cabinet positions with candidates
politically compatible with the president, the only room left in public dis-
2 This theme is by no means new among students of the United States. Ford (1898) proposed
the same idea of an acquiescent Senate in the nineteenth century.
1 10 Presidents and assemblies
c ourse for Senate objections is f or objections based on pers onality and
character (as opp osed t o p olitical) deficiencies.
President Ge orge Bush's n ominati on of J ohn Tower for the p ositi on of
Defense Secretary in 1989 illustrates this p oint nicely. Tower was unaccept­
able t o many senat ors f or p olitical reas ons; he was far m ore hawkish than
the Senate median. Yet sc ores of senators wh o certainly f ound Tower p oliti­
cally objecti onable justified their nay v ote on the n ominati on acc ording t o
Tower's alleged s ocial excesses, drinking and w omanizing. These are hardly
habits that ordinarily raise eyebr ows ar ound Capit ol Hill. Tower's case
c ontrasts sharply with the Senate's rejection of R obert B ork as a Supreme
C ourt n ominee less than a year bef ore. Objecti ons t o B ork were purely
p olitical, and they were justified as s o publicly.
The difference in Senate disc ourse between the tw o cases seems particu­
larly odd given that Tower's cabinet p ost was an explicitly p olitical p osition,
while B ork's seat was ostensibly n onp olitical. But the difference highlights
an imp ortant quality of app ointments in presidential systems. The cabinet
is underst o od t o be the p olitical resource of the president al one, especially
s o because the president alone may dismiss a member of the cabinet (while
neither the president n or the Senate may dismiss a Justice of the Supreme
C ourt). Alth ough the Senate can reject candidates - and does in fact reject
them f or p olitical reasons on occasi on - rejecti on on p olitical gr ounds is
difficult t o defend and often n ot w orth the tr ouble, since a president c ould
always dismiss a cabinet member wh om he or she f ound t o be t o o receptive
t o Senate preferences. F or the occasi onal rejecti on, then, other justifica­
ti ons are s ought, perhaps fabricated. The difficulty f or senat ors in objecting
t o cabinet n ominees sh ould c ontribute t o an explanati on of why only nine
cabinet n ominees have been rejected in U.S. hist ory, c ompared with
twenty-eight Supreme C ourt n ominati ons. We have m ore t o say ab out this
c omparis on of cabinet app ointments t o Supreme C ourt app ointments in a
later secti on of this chapter.

Other cases
Apart fr om the United States, only K orea, Nigeria, and the Philippines
require any f orm of c ongressi onal approval f or presidential n ominees t o the
cabinet. The K orean and Nigerian experiences are sh ort as of this writing,
s o we cann ot offer discussi ons of actual perf ormance. W hat we can do is
c onsider the pr ocesses themselves, and h ow they might resemble or differ
fr om the experience in the United States. K orea's requirement f or confir­
mati on dates only t o 1987 and is m ore restricted than in the other cases
c onsidered here anyway. Only the prime minister is subject t o confirmati on
by the unicameral assembly. Subsequently the prime minister app oints the
other cabinet ministers. Neither the prime minister n or any other members
of the cabinet are subject t o ong oing c onfidence of the assembly; thus, as is
 Limits on separate origin and survival 111
so by definition in presidential (but not premier-presidential) systems, only
the president may remove members of the cabinet. In the Korean context,
the ability of the president to get cabinet ministers acceptable to him or her
depends entirely upon how good an agent the prime minister proves to be.
If presidents are forced by the assembly to compromise considerably in the
choice of this official, then they could endure significant agency loss in the
composition of "their" entire cabinet. However, since the process of filling
the prime minister's seat is identical to the appointment game of Figure 6.2,
we should expect Korean cabinets to reflect closely the preferences of the
president, although not necessarily the president's ideal.
In Nigeria, only two presidents have been elected under the constitution
of 1979, which was suspended in a military coup in 1983 but was scheduled
to be reestablished in 1992. Article 135 of the constitution restricts the
president's appointments to ministries in two ways. First, as in the United
States, Senate confirmation is required. Second, there must be at least one
minister native to each state. Both provisions are attempts to ensure repre­
sentation of as many of Nigeria's diverse ethnic groups as possible, as state
boundaries were based on ethnic communities and each state is accorded
equal Senate representation.
Cabinet nominations in the Philippines are subject to approval by the
congressional Commission on Appointments, consisting of the president of
the Senate as well as twelve senators and twelve representatives. The com­
mission is selected according to the proportional representation of parties
in each chamber. 3 The president of the Senate presides over the commis­
sion, but only casts a vote on appointments in the case of a tie. The
provision for a form of congressional approval of cabinet nominations in
the Filipino constitution of 1935 may represent an emulation of one of the
unusual characteristics of United States presidentialism. A number of dis­
cussions of Filipino political history note that congressional challenges to
presidential dominance in policymaking grew more frequent in the 1950s
and early 1960s than they had been in the 1930s and late 1940s (Veloso and
Guzman 1969; Stauffer 1975), and that congressional challenges to presi­
dential discretion affected the appointment process. The Commission on
Appointments exercised its authority to reject nominees during the 1960s
to the extent that it drew criticisms that it was harassing the president
(Stauffer 1970).

C E N S U R E OF C A B I N E T M E M B E R S
As with congressional selection of the executive, provision for the censure
of cabinet members by the assembly can entail a dramatic break from
separation of powers, provided that censure entails meaningful ministerial
responsibility before the assembly: the right of the assembly majority to
3 Philippine constitution, Article VI, Section 18, and Article VII, Section 16.
112 Presidents and assemblies
dismiss individual ministers or the entire cabinet. In some systems, censure
has not always implied the automatic dismissal of a minister or cabinet, as
the term implies under parliamentary government. In some presidential
systems, censure is an official political rebuke of a member of the executive
by congress, but the decision of whether or not to retain the censured
official is ultimately the president's. When censure (sometimes called "in­
terpellation") is nonbinding, we might refer to it as "toothless." Next, there
are cases where the actual practice of censure has been rare or simply not
politically threatening to the executive, despite its provision in the constitu­
tion. Among the cases in this group, we shall consider the specific condi­
tions that reduce the frequency or effectiveness of the practice. Where the
term "censure" is meaningful and may be equated with parliamentary confi­
dence is where assemblies have relied on censure as a potent and standard
means of exerting pressure on a presidential executive by requiring the
dismissal of censured ministers. Here, of course, the impact of censure on
the balance of power between the executive and assembly has been most
profound: A regime with such a provision is not presidential.

Potent censure: Peru, the Chilean

"parliamentary republic, " and Ecuador
Two Latin American cases deserve special attention because their potent
censure has represented such a contrast to the norm of pure presiden­
tialism in the region. These two cases are Peru and Chile (before 1925). A
third case, Ecuador, has also employed censure, but there are troubling
ambiguities in the constitutional language that may exacerbate inter­
branch confrontations.
In both Peru and Chile, presidents have remained the formal heads of
government ( even if ineffective) despite cabinet turnover. Nevertheless, in
both cases, the dispatch with which majority congressional oppositions
have disciplined cabinets has meant that ministers have been both directly
answerable to congress for their actions and dependent on congress for
their survival.
Although congressional censure was included in the Peruvian constitu­
tion of 1933, the provision was exercised only six times in the next thirty
years (Needler 1965:158). During much of this time, of course, Peru was
ruled by governments that could not be called democratic. During the first
administration of President Fernando Belaunde Terry (1963-8), the Ameri­
can Popular Revolutionary Alliance, (APRA), in opposition to the presi­
dent, wielded the censure weapon freely. As APRA party leader Raul
Haya de la Torre explained, "We did not win control of the [presidential]
Palace, but we did win Congress, and from there we have begun to teach
the conception of a new democracy, to endow Parliament with true institu­
tional hierarchy. It is Parliament which makes the laws that are the norms
 Limits on separate origin and survival 1 13
and the base of an organized democracy" (McCoy 1971:352). The first
Belaunde cabinet was censured within six months of inauguration. Blaming
cabinet responsibility in part for the instability that preceded Peru's 1968
coup, McCoy wrote that "cabinet instability was primarily due to pressure
from Congress" and that "many ministers and even entire cabinets resigned
under threat of censure" (McCoy 1971:351-2). By 1968, the New York
Times reported that during his five year term, President Belaunde Terry
named 178 cabinet ministers. 4
Peru's current constitution of 1979 retains the provision for parliamentary
censure (Art. 226). Under the first two administrations following redemocra­
tization in 1980, Presidents Belaunde Terry and Alan Garcia each enjoyed
congressional majorities (a reliable coalition in the former case, a single­
party majority in the latter). W ith the election of Alberto Fujimori to the
presidency in 1990, however, the Peruvian system once again confronts the
problem of a president lacking a stable majority in congress. Moreover, in
this instance, the president is a self-proclaimed outsider to partisan politics.
While it is early to assess the performance of the system under these condi­
tions, we are circumspect that the provision for censure, combined with
absolute presidential authority to fill vacated cabinet posts, can be a recipe
for extreme cabinet instability, and possible standoff between president and
congress. We discuss the theoretical problems with such a formula further in
Chapter 8. For now, let us consider the performance of another such regime:
Chile's so-called Parliamentary Republic.

Chile. In his two-volume history of Chile through 1925, Julio Heise Gonzalez
(1974) argues that the country was far more parliamentary (and conversely,
less presidential) than has been generally acknowledged by political scien­
tists and historians, even before what is commonly called the "Parliamentary
Republic" of 1891-1925 (Heise 1974, 1:61-3). One of the key factors noted
by Heise as contributing to the parliamentary quality of the system was
congressional censure of cabinet ministers. Provision for censure existed as
far back as Chile's constitution of 1833, although it was not invoked until
1849. Throughout the latter half of the nineteenth century, censure was the
standard means by which congressional majorities enforced political compli­
ance on the executive. Censure - either of an individual minister or of an
entire cabinet - was voted many times between 1862 and 1864, and in 1870,
1875, 1878, 1881, and 1890. In addition, many more ministerial changes in
the 1880s were prompted by the threat of censure. Throughout this period,
censure votes were defended by congressmen on the basis that the removed
ministers "merely represented the policy of the President," and that their
maintenance in office contradicted the fundamentals of a parliamentary
republic (Heise 1974, 1:63-7).

4 New York Times, October 6, 1968, p. 4E.

1 14 Presidents and assemblies
Opposition to the principle of cabinet responsibility was central to Presi­
dent Jose Manuel Balmaceda's objections to the current political system in
1890. With Balmaceda's defeat and the promulgation of the constitution of
1891, however, Heise contends that commitment to the principle of respon­
sibility was virtually unanimous among the leaders of Chile's significant
political parties (Heise 1974, 1:121). Finally, although he presents no com­
prehensive data, Heise argues that censure was commonplace during the
1891-1925 period, providing an "escape valve" for the relief of political
tension between the branches. Moreover, he contends that the control the
legislature exerted over the ministries created a situation in which the
president was not necessarily even head of government. Rather, the cabi­
net reflected the will of the parliamentary majority, while the president
"played the role of regulator between government and opposition" (Heise
1974, 2:118-19). Although the system Heise describes seems a bit more
unwieldy than pure parliamentarism, its provision for censure, like Peru's,
clearly violated the principles of presidentialism.

Ecuador. Ecuador's 1929 constitution allowed for a congressional "vote of

confidence" on cabinet ministers. The constitutions of 1945, 1946, 1967,
and 1978, however, have allowed instead for votes of "censure." Although
students of parliamentary procedure may not recognize a distinction be­
tween the two, and we have used them interchangeably, the difference in
Ecuador often has been between dismissal and symbolic rebuke. Censure
has been treated as no more than a recommendation of dismissal from
congress, with the president making the ultimate decision on whether to
retain the minister. The practice has, however, been of some political
significance. Pinoargote argues that motions of censure have been used by
the Ecuadorian Congress to extract concessions from the executive since
the nineteenth century. Although he provides no comprehensive list of
censures, Pinoargote (1982:41-2) points to specific cases in 1961, 1970, and
1979 in which the procedure has been used to challenge the legitimacy of
the cabinet. Yet the 1978 constitution remains ambiguous.
The most troublesome aspect of Ecuador's censure mechanism for stu­
dents of political institutions is that the constitution of 1978 itself is unclear
on whether censure is meant to be implemented for political reasons, or
only in cases of criminal wrongdoing, as with impeachment. As Pinoargote
points out, Article 59, Section (f) authorizes congress to make "political
judgment" on members of the executive "for infractions committed in the
exercise of official functions." The implication is that censure can be a legal
procedure, like impeachment. Moreover, the constitution uses different
language with reference to censure of different officials. For example, the
president may be censured, but only for taking bribes, treason, or the like.
No such restrictions are placed upon censure of cabinet ministers. We have
entered into this discussion of censure in Ecuador not to attempt to resolve
 Limits on separate origin and survival 1 15
the question of whether or not censure has been effective in the particular
case in making cabinets approximate the will of the assembly majority, but
to indicate that provisions for censure should be explicit as to whether they
are intended to be used for political reasons or not. Ecuador's constitution
is fraught with ambiguity that only serves to heighten the pitch of confronta­
tion when the president and assembly majority are of different political
tendencies.

Restricted censure: Uruguay, Guatemala, and Cuba

In another group of systems with constitutional provision for censure, the
mechanism has likewise not proven a threat to the independence of the
executive. It is not that censure in these systems is toothless, however, but
rather that it has been implemented so infrequently or ineffectively. The
reasons lie in constitutional provisions that are worth brief consideration.

Uruguay. In Uruguay, the congress can invoke censure only at the risk of
beginning a process under which the president can dissolve the assembly
itself.5 Between 1934 and 1952, moreover, if congressional censure had led
to dissolution, the next step for the assembly that was returned could have
been removal of the president. This process is discussed later, in a separate
section on dissolution. Under any of these conditions, then, for the two
houses of congress, acting jointly as the General Assembly, to invoke
censure has been to begin a game of political chicken with the president.
Because of this deterrent, the Uruguayan assembly has barely made use of
its power to remove cabinet ministers. In 1969, the assembly censured one
of President Pacheco's ministers, but by less than a two-thirds majority.
The constitution permitted the president in such cases to veto the censure,
which Pacheco did. Then members of Senator Jorge Batlle Bere's faction of
Pacheco's own Colorado Party defected to uphold the censure and ensure
that the president could not invoke dissolution (Gillespie 1989:22; Gonza­
lez 1989:5-6). In doing so, Batlle Bere's faction stated explicitly that the
defection represented only a political maneuver and not a defense of the
principle of cabinet responsibility. This is the only case of censure in Uru­
guay outside the period of the plural executive, or colegiado. But even
under the conditions of the colegiado, censure did not carry the political
implications it does under a unitary executive. The heads of ministries who
were subject to censure were primarily administrators, whereas members
of the Executive Council - the collegial presidency that was the colegiado's
approximation of a cabinet - could not be removed by the assembly.
Censure is a potentially powerful provision in Uruguay's institutional
arrangement. But for the assembly to invoke it risks a lopsided version of
5 This was not the case during the period of the colegiado (1952-66), nor is it when the
censure vote passes by a majority greater than two-thirds.
116 Presidents and assemblies
mutual assured destruction. This deterrent has created a sometimes uneasy
standoff between congress and executive, preventing censure from being
used to enforce cabinet responsibility.

Guatemala. The Guatemalan constitution also provides for congressional

interpellation and censure of ministers. A majority of deputies in the uni­
cameral congress can vote no-confidence in up to four ministers at a time.
The president, however, can veto the censure, which requires a two-thirds
majority for override. 6 Thus censure cannot be used to invoke cabinet
responsibility to the legislative majority. We know of no cases when the
process has even been initiated (Garcia Laguardia and V asquez Martinez
1984:76-7). 7

Cuba. The Cuban political system in the 1940s has been called "semi­
parliamentary," primarily because the constitution of 1940 provided for
political responsibility of cabinet members to Congress (Stokes 1949). By
our estimation the system still functioned essentially as presidential. Presi­
dents were elected directly, and for absolutely fixed terms. Moreover, as
Stokes himself points out, the Cuban president remained the effective head
of government even while the new constitution created the post of prime
minister (Stokes 1949:342-3). Nevertheless, the vote of censure during this
period remains worthy of attention, largely because a principal intention of
the framers of the 1940 constitution was to create cabinet responsibility.
In this intent they were only minimally successful, mainly because there
were some procedural impediments to the enforcement of responsibility.
A new cabinet member could not be censured for six months after assum­
ing the post, and no motions of censure were allowed in the last six
months of a president's term. Both Presidents Fulgencio Batista and Ra­
mon Grau San Martin named their ministers without regard for the parti­
san balance in the congress, even though both originally faced majority
opposition in Congress.
In May 1942, Congress interpellated the entire cabinet for questioning to
assess the need for a congressional committee on executive oversight. Al­
though Congress voted confidence in the entire cabinet, every member
proceeded to tender resignations to President Batista, who accepted all but
one (Stokes 1949:348-9). The congressional challenge to Batista's adminis­
tration certainly appears to have prompted this turnover, although the
provision for censure was not specifically invoked.
A second incident demonstrates how particular constitutional provisions
for censure in what otherwise would be a presidential system may encourage
6 Constitution of Guatemala, Articles 166-7.
7 Moreover, given Guatemala's extended periods of nondemocratic rule, with close ties
between the military and the presidency, it is doubtful that Congress has been in a strong
enough position vis-a-vis the executive to challenge cabinet ministers.
 Limits on separate origin and survival 1 17
presidents to seek to circumvent the intended congressional confidence. In
October 1945, congress voted to censure the minister of commerce, Dr.
Inocente Alvarez, but did so on charges of fraud, not for acknowledged
political motives. More importantly, President Grau responded to the cen­
sure of his long-time friend by immediately naming Alvarez as prime minis­
ter (Stokes 1949:360-1). Grau's move was clearly intended as provocation,
but congress failed to respond to the challenge. So while the formal measure
of censure was not disputed, the principle of cabinet responsibility was
defeated. It was defeated within the limits of the constitution, however.
Presidents (as in the several president-parliamentary regimes) were not re­
quired to obtain prior approval to appoint a cabinet minister. Thus, although
congress could respond with another censure, no provision exists in such
regimes to prevent this minister from initially assuming his post upon nomina­
tion by the president. Further complicating matters in Cuba, once ap­
pointed, a minister could not be censured for six months. Thus presidents
were afforded considerable leeway in filling cabinet posts or shuffling their
cabinets, despite the formal provision for censure. 8

Censure and the appointment game

We have seen that in both Peru in the 1960s and the Chilean so-called
Parliamentary Republic, cabinet responsibility to the assembly coexisted
with directly elected presidencies. In this sense, both appear to have fore­
shadowed the regime type we have identifed as premier-presidential. How­
ever, we should like to make clear here why we regard these two cases as
having been president-parliamentary rather than premier-presidential (for
reasons more profound than simply the lack of a post of premier in the
Chilean case), and why the difference between these two regime types is
important to the prospects for democratic success.
In the previous discussion of assembly involvement in the appointment
of cabinet members, we saw that initiative matters a great deal in structur­
ing the appointment game. Because they make the initial appointment, and
because the cabinet is recognized as the personal domain of the president,
U.S. presidents are able to obtain cabinet members that accord well with
their policy preferences, even when the Senate is controlled by the opposi­
tion. Additionally, U.S. presidents, and not the Senate, have the power to
remove a cabinet member at will. In France, on the other hand, the presi­
dent cannot remove a cabinet member or the premier, but the assembly
can; thus, the French president facing an opposition-controlled National
Assembly is weaker in the appointment game than is his U.S. counterpart.
In France, the asymmetry in powers of removal compensates for the
asymmetry in nomination: W hile the president retains initiative in propos-
8 In October 1947, Grau's minister of education was censured for explicitly political reasons.
and promptly resigned. We know of no other cases of censure under that constitution.
1 18 Presidents and assemblies
ing a premier, the assembly has the initiative in removing him or her. A
president who did not want to nominate someone who would be approved
and retained by the assembly would have to be willing to leave the post
unfilled. Such a threat is rarely credible. In the United States, asymmetry
works in favor of the president only: An assembly unwilling to accept any
nominee would have to be willing to leave the post unfilled, which can be
politically costly, especially at the cabinet level. The president, on the other
hand, does not bear the political costs of leaving posts unfilled, as long as
she or he is at least willing to submit nominations to the Senate. 9
How does the appointment game look in cases such as Chile before 1925,
and Peru? In neither case is it required that a president's nominee be
confirmed by the assembly. W hile the assembly can censure and remove
cabinet members, as in premier-presidential systems, so too can the presi­
dent remove cabinet members at will. The structure of the game by no
means compels the president to retain a cabinet acceptable to the assembly.
When the assembly exercises censure, the president can simply continue to
fill cabinet posts apace. In Chile and Peru, then, there is symmetry in the
game after appointment in that either president or assembly may remove
the cabinet or members of it. But because only the president can propose
appointees, the assembly lacks the ability to make its own preferences
stick. Thus censure becomes a strictly negative power, one with which to
harass the president, but not a power with which to reconstitute the govern­
ment. As we saw in Chapter 4, symmetry in removal powers contributed to
interbranch conflict in Portugal (1976-82), and to chronic cabinet instabil­
ity in Germany's Weimar Republic as well. We maintain that giving the
assembly the power to censure but effectively depriving it of the ability to
construct a new government by letting the president both propose appoin­
tees and remove sitting members of the cabinet is a dangerous prescription
for conflict and instability, of the sort that led to the coup in Peru in 1968.

B U I L D I N G A N D M A I N T A I N I N G C A B I N E TS

Our point is more clearly illustrated in Table 6.1, which situates all of the
regimes discussed in this book according to how cabinet positions are filled,
and how ministers or entire cabinets are removed. In the rows of the table,
regimes are grouped according to who has the authority to fill cabinet
positions: the president, the assembly, or both working in consensus. 10 In
9 Of course, in some cases, and especially at the lower levels of the appointed bureaucracy.
even the task of submitting nominations for the hundreds of executive posts to be filled can
overwhelm an administration - as it did that of George Bush for the first two years.
10 The regimes are located in this figure according to the scores given for cabinet formation
and cabinet dissolution in Table 8.2, below. Even in Chapter 8, our scoring system is
intended to approximate and to order presidential powers, not to quantify them. More­
over, for Table 6. 1 , we have collapsed some scores further, to provide a broad, but rough,
illustration of the origin and survival of cabinets. For example, those presidential systems
 Limits on separate origin and survival 1 19

Table 6. 1 . Filling and dismissing cabinets

Dismiss
Pres1'dent A ssembl1v E'1ther
President Presidential Finland President-
(e.g. Costa Rica, Iceland parliamentary
Venezuela) (premier- (e.g. Chile pre-
presidential) 1925, Namibia,
Peru, Sri Lanka)
Fill Assembly Parliamentary
systems
Both umer presidential Most premier- Portugal 1976
systems: presidential
Korea 1987 systems
Nigeria (e.g. France,
Philippines Portugal 1982)
United States

the columns, regimes are grouped according to who possesses the power to
dismiss ministers: the president, the assembly, or either on its own. Two of
the nine boxes in the figure are empty. We know of no regimes in which the
assembly alone fills cabinet positions, but in which a president - either
exclusively or in addition to the assembly - has the power to remove minis­
ters. The other boxes all contain at least one existing or historical regime.
At the top left lie most of the systems that we have identified as pure
presidentialism, where the cabinet is literally the personal domain of the
president. At the bottom left are systems that we also regard as presiden­
tial, but in which the president's dominance over the cabinet is compro­
mised slightly by the requirement that the assembly (or some delegation
thereof) approve of cabinet nominations. In our discussion and modeling
of the appointment game, we suggested that the president's initiative in
appointments, and the ability to make take-it-or-leave-it offers provides
him or her with a distinct advantage over the assembly in securing desired
appointments.
In the middle column of the figure are the pure parliamentary regimes
and the more successful premier-presidential systems. The differences
among these regimes (on the question of government formation and dis­
missal powers) are over who names the cabinet. Simply put, if a popularly
elected president has any substantive authority to fill cabinet posts, the
regime can be premier-presidential. If the cabinet is filled through assem-
in which presidents name all cabinet members, subject to assembly approval, are grouped
with those premier-presidential systems in which the president names the premier, who
subsequently names the ministers. Despite the differences, we argue that in all these
regimes, "both" the president and the assembly have significant powers in filling the
cabinet. For the purposes of Table 6.1, these differences are ignored.
120 Presidents and assemblies
bly initiative, or by parties in the assembly, then the regime is closer to
parliamentary - even if a ceremonial head of state formally appoints a
premier. 1 1 The premier-presidential systems shown differ over the particu­
lars of how cabinets are named. In Finland and Iceland, the president
names all the cabinet positions, which are, of course, subsequently subject
to parliamentary censure. In France and Portugal, the president names
only the premier, who proceeds to fill cabinet posts after being confirmed
by the assembly. Under both formats, however, the president makes the
critical first move in cabinet formation, after which the cabinet is depen­
dent exclusively on the assembly for its survival.
As suggested, this format encourages negotiation and compromise be­
tween president and assembly over the appointment of cabinet members at
the outset of the appointment game. The president's resource is his or her
initiative in the game, and the ability to make the assembly bear a dispro­
portionate political cost for leaving the cabinet unfilled if it rejects all
nominees. The assembly's primary resource is its exclusive ability to dis­
miss the cabinet; although it may also be able to impose a political cost on a
president whose nominees it routinely rejects, if the president is perceived
as making unreasonable nominations. Thus, presidents have an incentive
to name the acceptable candidate with whom they are most politically
compatible. The assembly's incentive is to accept the first nominee for
whom the political costs of rejection (and thus, leaving the cabinet empty)
are greater than the costs of confirmation.
Moving to the right in Table 6.1, we encounter the more problematic
regimes among those in which either the president or the assembly may
remove cabinet ministers. We suggested above that the interaction of cen­
sure and the appointment game becomes more complex, and potentially
dangerous, when there are both asymmetry in appointment and symmetry
in dismissal. If this is true, then Portugal (1976-82), while unique, might
not have proven problematic, because although either president or assem­
bly could remove ministers, both were also required to construct a cabinet.
Indeed, the conflict between governmental branches that produced the
constitutional reform of 1982 was based more on differences between a
nonpartisan president versus parties in the assembly, and did not threaten
to generate regime crisis or political disorder. We do not, then, see such an
arrangement as necessarily unworkable, since the same incentives for com­
promise exist here as under other systems in which presidential nomina­
tions require the consent of the assembly. Nevertheless, the responsibility
of cabinet members to both the president and the assembly can generate a
"confusion" of loyalties in the executive, as we shall see below. In short,
the Portugal 1976 formula is somewhat more clumsy than the premier­
presidential regimes in the center column of Table 6.1, but certainly supe-

1 1 It could still be premier-presidential, if the president holds the powers of dissolution.

 Limits on separate origin and survival 121
rior to the other president-parliamentary regimes in the upper-right-hand
box.
The problematic systems, then, are those in which the president alone
fills cabinet posts, but in which either the president or the assembly can
dismiss ministers. These regimes might be labeled "confused," in that the
responsibility of cabinet members is unclear and quite possibly contradic­
tory. They can also be characterized as "asymmetrical," since the president
alone makes appointments while either branch can disassemble the cabi­
net. The problem here is that there is no incentive for negotiation and
compromise on the part of any player, neither at the stage of filling cabinet
posts, nor in dismissing ministers. The president fills positions without
constraint. Thus, the assembly's only means of affecting the constitution of
the cabinet is to dismiss it. But this is a purely destructive tool, which
provides no manner of ensuring that the cabinet or portfolio that follows
the dismissal will be acceptable. Because cabinet positions need not remain
indefinitely vacant, moreover, neither the president nor the assembly can
be held directly responsible for a prolonged standoff. Instead, the domi­
nant strategy for each player is to make the next move, creating an
appointment-dismissal game in which there is no stable equilibrium. 12
Thus, when the president and assembly are at odds, the arrangement en­
courages the unstable pattern of appointment and dismissal that we see in
Weimar, Cuba, Peru, and Chile before 1925. Two newer systems, Colombia
1991 and Namibia, may be expected to encounter similar difficulties.

THE APPOINTMENT-DISMISSAL GAME

Having discussed the processes of building and maintaining cabinets under

various institutional formats and their potential for compromise or conflict,
we return now to spatial reasoning. The appointment game we have just
presented can now be expanded to accommodate not only the initial play -
nomination - but also the next move, taking into account who has the
authority to dismiss an appointee.
Table 6.2 represents the importance of the power of initiative in the
appointment-dismissal game. W here a party has initiative, either by mak­
ing a nomination or by dismissing an officeholder, we give a value of two
points. Since we wish to place the players (president and assembly) at
opposite ends of a unidimensional scale in order to map preferences and
predict outcomes in a two-player game, the president's nomination or dis­
missal is a + 2, while the assembly's dismissal (via censure) is a -2, from
the standpoint of its effect on the president's ability to retain ideal appoin­
tees. Assembly confirmation, being a power that is reactive, rather than
initiative, we score as -1 from the president's perspective; the president
12 For an extensive theoretical discussion of dominant strategies, see Ordeshook (1986), ch.
3.
122 Presidents and assemblies

Table 6.2. A schematization of presidential powers in the appointment-dismissal

game

President Assembly
Nominate Dismiss Confirm Dismiss Total Exam Jes
2 2 0 4 Most pres. systems
2 2 -1 0 3 U.S., Philippines
2 2 0 -2 3 Peru, Weimar
2 2 -1 -2 1 Portugal 1976
2 0 -1 0 1 U.S. Supreme Ct.
2 0 0 -2 0 Finland, Iceland
2 0 -1 -2 -1 France

loses power relative to the assembly when the latter must confirm her or his
nominations. If the assembly, or an agent thereof, could initiate appoint­
ments by presenting a nominee, it would gain two points, meaning a -2 on
presidential power. The scale that results is depicted in Figure 6.3 and runs
from -4 (which represents the assembly's ideal appointee) to +4 (which
represents the president's ideal point). We should caution that these scores
do not represent specific quantification; rather, this exercise is a heuristic in
which the scores represent relative placement in a unidimensional spatial
model. The several institutional arrangements are displayed in Table 6.2 in
numerical order, but we deal with them in the text below in a different
order, proceeding by regime type, as in previous discussions. The discus­
sion that follows allows us to refer back to Figure 6.3 and see how different
institutional arrangements affect likely outcomes in the appointment­
dismissal game.
The spatial model suggests that the U.S. or Philippine cabinet members
would be at about point 3 in Figure 6.3. They would thus be close to, but
somewhat removed from, the president's ideal point. U.S. presidents could
hold out for a 4, but to do so they would have to be willing to risk an empty
seat, for they do not have the unrestricted appointment power of presi­
dents in most other presidential systems, such as Costa Rica or Venezuela.
For the assembly (the Senate in the United States), it must prefer either an
empty seat or constant turnover not to confirm a nomination at 3, since the
president can always dismiss a minister who is too far from presidential
preferences. In other words, as we saw in Figure 6.2, the indifference
curves overlap closer to the ideal point of the president than to that of the
assembly, assuming that there is a divergence between the preferences of
the president and the majority of the assembly.
If the president initiates the appointment process, subject to confirma­
tion by the assembly, but neither may dismiss, we have a case such as the
U.S. Supreme Court. The score, according to the method of Table 6.2, is
thus + 1. The model thus implies that presidents typically would have to
 Limits on separate origin and survival 123
A ---+---1-----1----+---+----+---+---- p
-4 -3 -2 -1 0 1 2 3 4
Austria, Finland, Portugal 76, Chile pre-25, Korea 87, Arg., Brazil,
France, Iceland U.S. Sup. CL Colombia 91, Nigeria, Chile 25/69,
Portugal 82 Ecuador, Peru, Philippines, Colombia pre-9 1 ,
Ger. (Weimar) U.S. cabinet Costa Rica, Dom.
(No stable Rep., El Salvador,
equilibrium) Guat., Korea 62,
Mex.• Nie.• Pan.,
Ven.

Figure 6.3. Schematized outcomes of equilibria in appointment-dismissal game,

relative to ideal points of president (P) and assembly (A)

settle for appointees who are less close to their ideal than is the case for
cabinets, but still closer to presidential than to assembly preferences. This
turns out to be another way of saying that because these appointees are not
subject to having their career advancement curtailed if they deviate from
presidential (or assembly) preferences, they are relatively free of political
constraints. Knowing this, Senate majorities in the United States should be
more capable of demanding appointees who are farther from the presi­
dent's ideal (at the time of appointment) than is the case for cabinet posi­
tions. The frequency of rejection, relative to cabinet positions, bears this
out, as do the occasional occurrence of "mistakes," justices who prove in
office to deviate from the preferences of the presidents who nominated
them. Still, the president retains the initiative in the appointment game ,
thereby encouraging the Senate to be cautious in rejecting a nominee,
realizing that, in the event of rejection, a replacement nominee may be no
better for them, and could be worse. 13
Now let us consider the case of cabinet appointments in a premier­
presidential system that has the appointment-dismissal game characteristic
of France: The president nominates the premier, subject to confirmation,
but only the assembly may dismiss. The model predicts an equilibrium at
-1. In other words, we find that the post of premier, and, by extension,
other cabinet seats, should be filled by someone closer to the assembly
majority's ideal point than to the president's ideal point, but that the assem­
bly would not obtain its ideal. So what about the French case in 1986, when
President Frarn;ois Mitterrand nominated as premier the leader of the
assembly majority coalition, Jacques Chirac, thus representing the assem­
bly's ideal point ( -4 in the graph)? Why did he not try to fill the post with
some other rightist who was not as unpalatable to him as Chirac, his princi­
pal political opponent? The answer, again, lies in the dismissal phase
13 There are some presidential constitutions (El Salvador, for example) in which only the
assembly fills Supreme Court vacancies. This sort of arrangement leads to a prediction of
appointees at -2 (the score for assembly nomination, not. subject to confirmation or
dismissal), or responsive to the assembly more than to the president, but straying from the
assembly's ideal. Systematic research on court decisions in cases with this sort of appoint•
ment process would be worthwhile.
124 Presidents and assemblies
of the game, as well as in Mitterrand's lack of interest in provoking confron­
tation (see Chapter 4 and also Pierce 1990). If the president in such a
system would prefer cabinet instability or empty cabinet posts to accepting
a - 4, she or he can appoint a - 1, despite knowing that the assembly has
the next move and can topple the government. If the president prefers an
assuredly stable cabinet (-4) to the marginal utility to be gained by a -1
(who, after all, would still be from across the partisan divide in a bipolar
party system like that of France), he or she might as well appoint a - 4. This
is a fair characterization of the calculation that Mitterrand made in nominat­
ing Chirac. However, the potential for a president to risk cabinet crises by
refusing to appoint at or very near the assembly's ideal point is real and, as
indicated in Chapter 4, the French experience from 1986 until 1988 by no
means proves that cohabitation must go as smoothly as it did in that one
example. 14
Even if the appointment-dismissal game is exactly as in France, out­
comes might be different if there is a multipolar party system. In such a
case, with parties divided across several dimensions and therefore no clear,
stable majority, the president's initiative might permit a -1, since precisely
what candidate would represent the theoretical ideal point of the assembly
would be less clear. Presidents could exploit this uncertainty with their
power of initiative. They are even freer to do so if they have the power to
appoint all members of the cabinet, rather than just the premier, without a
formal vote of confirmation (usually called investiture in the parliamentary
context). In such a case, the model predicts appointments at 0. Thus,
theoretically, presidents operating in this kind of an institutional environ­
ment should be able to get appointees closer to their ideal than they might
in systems structured like France. However, the ability of the assembly to
invoke censure again, as in France, implies that presidents would have to
be willing to tolerate considerable turnover in cabinets in order to maxi­
mize the proximity of cabinets to their ideal point. They might opt instead
for appointments at points farther left in Figure 6.3.
This modeling of the game just described allows us to make some in­
formed speculation about Finland and Iceland, two premier-presidential
systems in which the president appoints all cabinet members. The frequent
portrayal of Finnish presidents as playing a stronger-than-legal role in cabi­
net formation (e.g., Duverger 1980) as well as the extremely high turnover
in cabinets (Strom 1990) both can be accounted for by our model. If Fin­
nish presidents (or at least the long-reigning Urho Kekkonen) have regu­
larly sought appointees at the midpoint of the scale, frequent censure
would be the predicted result. On the other hand, in Iceland, where presi­
dents are observed not to have exercised powers they are entitled to exer-

14 See Chapter 12 for further discussion of cohabitation and ways to design premier­
presidential constitutions so that it is less likely or less problematic.
 Limits on separate origin and survival 125
cise (Duverger 1980) and where turnover of cabinets is much lower, despite
a multiparty system (Lijphart 1984), it is arguably because presidents have
regularly avoided provoking censure by making appointments closer to the
assembly's ideal point.
Next we must consider instances in which the president makes appoint­
ments wihout needing confirmation, but either the president or the assembly
may dismiss ministers. Such regimes, which we have dubbed "president­
parliamentary," have been found in the German Weimar Republic, Peru,
and other cases we have identified. The scheme of Table 6.2 suggests an
equilibrium around 2, implying a consensus on cabinet appointments only
slightly farther from the president's ideal than in the presidential systems
that require confirmation. However, as indicated earlier, the dominant strat­
egy for each player in this game is likely often to be to make its move rather
than to negotiate a compromise. Presidents can appoint a 4 and dare the
assembly to censure. Presidents can appoint an official more acceptable to
the assembly, confident that they also can remove this official at any time.
Chronic cabinet instability is the likely result under this design when presi­
dent and assembly are of opposing political tendencies.
Perhaps somewhat less conflictual, as already discussed, is a game like
that of Portugal's 1976 constitution, in which both the appointment and
dismissal phases of the game involve shared powers. This design may thus
be the one most likely to generate consensus cabinets and thus true joint
confidence. The reason is that the president cannot make an appointment
at 4 and simply dare a censure as in cases like the Weimar Republic.
Moreover, as in the United States or the Philippines, any attempt to obtain
a 4 against the wishes of the assembly would require a willingness on the
part of the president to leave the post vacant, which is presumably worse
than turnover. This design builds the mutual confidence into the appoint­
ment phase, a clear improvement over designs such as those of the Weimar
Republic and Peru. However, we still caution against such an arrangement,
in part because its very limited use does not allow for meaningful empirical
assessments. Besides, the problem remains of a lack of ultimate authority
(and therefore of ultimate accountability before the electorate) for the
actions of the cabinet. Either presidential or premier-presidential systems
give one or the other player in the game ultimate authority through exer­
cise of asymmetrical dismissal powers.
The arrangements most conducive to clear accountability in cabinets are
those that provide for asymmetry in the authority to dismiss governments
or ministers. W here only the president may dismiss, cabinets will tend to
approximate the president's preferences. W here only the assembly may
dismiss, the assembly majority's preferences will be approximated. When
the branch that has sole authority to dismiss is not alone free to fill the
posts, as is the case variously in France and the United States, there is a
potentially desirable check by one branch on the preferences of the other,
126 Presidents and assemblies
compared to either the pure presidential cases (Costa Rica, for example) or
straight parliamentarism. However, when powers overlap to the extent that
the dismissal game is symmetrical - and especially if the appointment
phase of the game is asymmetrical - conflict and instability are severe
risks.

DISSOLUTION OF THE ASSEMBLY

Dissolution o f the assembly by the president is a significant break with the

principle of separation of powers and the final one we shall discuss. How­
ever, in the first two systems considered here, dissolution has been con­
fined to extraordinary situations; in the third, the provision for dissolution
is brand-new and unused. Another, Paraguay, has no record of democratic
government before 1990 on which to base any assessment. 15 Paraguay and
the 1989 constitution of Chile are the only cases that we know of in which
the president had both the power to direct the cabinet and the power to
dissolve the assembly purely on her or his own initiative. 16 In both the
Uruguayan and Peruvian cases, dissolution can be invoked only after con­
gress has censured cabinet ministers - or even the entire cabinet - for po­
litical reasons.
The provision for dissolution would seem to provide an institutional re­
source to resolve the problem of congressional deadlock decried by the
critics of presidentialism. Indeed, in a parliamentary context, dissolution
does just that. On the one hand, the parliamentary executive can use the
threat of dissolution as leverage to secure legislative support for cabinet
proposals. On the other hand, dissolution is not without risks for the parlia­
mentary executive and therefore cannot be used indiscriminately; the voters
may very well return a parliament which will replace the current govern­
ment. W here the authority exercising the power of dissolution is one who
stands outside of parliament, as with a president, there is thereby an exter­
nal check on parliament and its cabinets. Indeed, a hypothetical form of
premier-presidentialism would be one in which the power of dissolution is
the only political power that the president wields. If dissolution resulted in
the voters' returning parliamentary majorities more to the president's lik­
ing, the president would be able to obtain a cabinet (and therefore policy)
likewise closer to his preferences. If, however, the voters return parliaments
opposed to him, he must accept a cabinet not to his liking. Where dissolu­
tion takes on a very different characteristic is where cabinets are either
jointly responsible to president and parliament, or where they are presiden­
tial cabinets. In the first instance, often dissolution is allowed only after the
censure of ministers. The president then cannot use the threat of dissolution
15 At the time of writing, a constituent assembly was planned to revise the constitution.
16 The current Chilean provision for dissolution of the Chamber of Deputies, which, of
course, remains as yet untested, is described below.
 Limits on separate origin and survival 127
selectively as leverage to secure majorities for government policies. In the
case of presidential cabinets, the president remains head of government
even if a hostile congress is returned after dissolution. In such a situation,
the problem of deadlock is likely to remain or even to be antagonized.
Let us now turn to some systems that are not premier-presidential accord­
ing to our definition, but in which some constitutional provision for dissolu­
tion nonetheless has been made.

Uruguay
During the periods of Uruguay's history in which dissolution has been
allowed (1934-52, 1967-73, and 1983-present), it has always been contin­
gent on the censure of cabinet ministers by the assembly. Under the consti­
tutions of 1934 and 1942, the two houses of congress, acting together as the
General Assembly, could censure cabinet ministers. Provided that the cen­
sure passed by less than a two-thirds majority, the president then had the
right to dissolve congress and call new elections. However, if the new
General Assembly upheld the original censure, the entire government,
including the president, would fall (Gonzalez 1989:4). During the period of
the collegial executive (1952-66), congress retained the right to censure
ministers, but not members of the Executive Council. In turn, the execu­
tive did not have the authority to dissolve the legislature (Gillespie
1989:12). Under the 1967 constitution, with the return to the unitary execu­
tive, the president was once again accorded the right to dissolve congress,
although under somewhat more restrictive conditions. If a censure vote in
the General Assembly passed by less than a two-thirds majority the presi­
dent could refuse to dismiss the minister in question. The vote then had to
be repeated, and the president could dissolve congress only if less than 60%
of legislators upheld the censure. W hat's more, the president could do this
only once during his term, and not during the last year of an assembly's
term. The 1967 constitution has been retained largely intact since the re­
turn to democracy in 1983, including the provisions for dissolution.
Despite the many and extended periods in which dissolution has been
possible in Uruguay, it has never been invoked. For the 1934-52 period,
Gonzalez (1989) attributes this to consistent Colorado control of the execu­
tive and sufficient congressional support. Even since 1967, however, when
the provision for dissolution was reinstituted, the process has never been
carried out, and has only been initiated once. In the incident in 1969 that
we discussed in the section on censure, the legislators were able to protect
themselves from the threat to their positions that the dissolution provision
implies by a politically calculated defection of members of the president's
party. Since the return to democracy, no movement toward dissolution has
progressed as far. At one point, with his minister of the interior embattled
during a controversy over police brutality, President Julio Sanguinetti
128 Presidents and assemblies
threatened dissolution, successfully preempting a censure vote (Gillespie
1989:36).
The variations in Uruguay's rules and the episodes related illustrate
many of the points we wish to make regarding dissolution. First, although
dissolution was never invoked during that period, the 1934 and 1942 consti­
tutions allowed for the possibility of a congressional vote of no-confidence
in the executive following any dissolution of congress. Although the pro­
cess could only be initiated following a censure vote, the requirement of
mutual confidence was a distinctly nonpresidential characteristic of those
constitutions. Second, the fact that the president cannot be removed as
head of government under the current arrangement regardless of the re­
sults of new congressional elections suggests that deadlock is likely even
after dissolution, should it occur. Third, the provision for dissolution only
after censure does not significantly strengthen the president vis-a-vis con­
gress, since dissoiution cannot be used as an executive initiative. A presi­
dent can invoke dissolution only when under attack, to save an embattled
cabinet. The fact that dissolution has not been invoked despite conditions
of immobilism underscores this point. Finally, a president might possibly
try to provoke the censure of ministers in order to prompt the conditions
for dissolution. This strategy would require extraordinary political loyalty
on the part of "kamikaze" ministers, however, and would be inordinately
risky, given the potential for antagonizing congress. We suggest, then, that
dissolution under Uruguayan conditions will most likely remain a defensive
weapon, if it is used at all.

Peru
According to the Peruvian constitution of 1979, there is the possibility of
dissolution in response to censure of the cabinet, but the conditions are
more restrictive than in Uruguay. The Peruvian government has a pseudo­
parliamentary structure in that the directly elected president names a prime
minister to head the cabinet. The president and prime minister then concur­
rently name the cabinet members. 17 We refer to the arrangement as
pseudo-parliamentary, however, since the president remains head of gov­
ernment throughout, while the prime minister effectively serves as chief of
staff. The cabinet is subject to the confidence of the Chamber of Deputies.
The chamber can censure an individual minister, or the entire cabinet, by a
simple majority vote. 18 Only after the chamber has voted no-confidence in
three cabinets, however, can the president dissolve the chamber - and
even then may do so only once per term, and not during the last year of a
term. 1 9
17 Peruvian constitution (Art. 216).
18 Ibid. (Art. 226).
19 Ibid. (Arts. 227, 228, 229).
 Limits on separate origin and survival 129
We know of no instance in which dissolution has been invoked in Peru,
or even when the sufficient condition of three no-confidence votes has been
established. Under these conditions, even more than in Uruguay, it seems
that dissolution will remain the recourse of an embattled and highly con­
frontational president, and probably not much of a threat to the congress.

Chile: presidential cabinets, yet dissolution

In its provision for dissolution, the Chilean constitution of 1989 marks a
dramatically greater departure from the principle of separation of powers
than either Uruguay or Peru. In the latter two, the conditions for dissolu­
tion must be created (and can be avoided, as we have seen) by congres­
sional action. In Chile, this is not the cse. The Chilean president may
dissolve the Chamber of Deputies once during a term, and not in the last
year of the term; but there are no other conditions stipulated. 20 The lack of
preconditions seems to entail an enormous shift in power toward the execu­
tive in Chile relative to the pre-1973 congressional-executive balance. In­
deed, the power of the Chilean president to dissolve the chamber may be
even more significant than the power of a parliamentary executive to dis­
solve the assembly, since for the Chilean executive to do so entails no risk
to its own survival.
At this point, it is difficult to determine to what degree this institutional
arrangement will actually alter the traditional balance; the dissolution
power, of course, has not yet been used. Nevertheless, the degree to which
it compromises the principle of separation of powers is sufficient that Chile
should not be considered a true presidential regime by our criteria. Indeed,
it might be more appropriate to call the regime "super-presidential." On
the other hand, neither Uruguay nor Peru would be regarded as non­
presidential due to dissolution alone (Peru's potent censure, however, does
make the regime nonpresidential). In neither case has dissolution been
invoked, of course; but more importantly, in neither case may executives
take the initiative to dissolve congress except under a set of extraordinary
conditions in which cabinets appointed by the president have been censured
by congress either excessively, or by less than overwhelming majorities.

CONCLUSION

This chapter has emphasized three central points. First, we have further
evidence for the argument, made at the outset of the book, that there is a
huge variety of institutional arrangements for assembly-executive relations
in systems generally grouped together as presidential. To students of politi­
cal institutions, this statement should seem self-evident. Nevertheless, it is
20 Chilean constitution (Art. 32, Sec. 5).
130 Presidents and assemblies
worth emphasizing the point, particularly since much of the current litera­
ture critical of presidentialism leans heavily on blanket portrayals of
assembly-executive relations. Among its critics, presidentialism tends to
be dismissed as a monolithic regime type. We intend here to emphasize the
very diversity of institutional arrangements in presidential, and premier­
presidential systems, so that we may better understand what specific provi­
sions generate conflict between the branches, and whether these provisions
can be modified, short of eliminating the popularly elected chief executive
altogether.
Second, much of our discussion here has highlighted the ambiguities in
the concept of the separation of powers. We have suggested that merely a
provision for censure or for dissolution does not necessarily represent a
jump to a parliamentary executive. So when the authority to threaten the
survival of the other branch is contingent on that other branch's actions,
there are serious constraints on the use and impact of censure or dissolu­
tion. For example, dissolution power does not strengthen presidents who
can use the authority only when their cabinet ( or successive cabinets) has
been dismissed nearly so much as dissolution power that may be used at
will. The array of arrangements, and sometimes contingent combinations,
of the powers reviewed here suggest that many of the regimes widely
considered presidential (or parliamentary, like Chile 1891-1925 or Cuba in
the 1940s) were in fact hybrids.
Third, we regard as important the point that "confusion" over to whom
the cabinet is responsible is a recipe for dangerous cabinet instability. This
is especially true where one branch alone names the cabinet to begin with.
We know of no regime in which the assembly alone names a cabinet that
can be removed by either president or assembly. But we have discussed a
number of problematic cases in which presidents alone fill cabinets, which
are then responsible to both the president and the assembly. We have
reviewed the incentives for each player (president and assembly) in a
"game" of appointment and dismissal, and have concluded that asymmetry
of power in such a game is disadvantageous for cabinet - and, quite possi­
bly, regime - stability.
We have now completed our review of the constitutional origin and
survival of assembly and executive in presidential and premier-presidential
systems. We have provided an overview of how executives are constructed
and how they coexist with assemblies that may represent a different aggre­
gation of interests from their own. Our focus has been on the role and
strength of popularly elected presidents in these areas. We shall move in
Chapter 7 to discuss the legislative powers of presidents, particularly veto
and decree powers.
 7
Legisl ative powers of presidents

I N TRODUCTION

Regarding the subject o f presidential p ower in the legislative pr ocess, we

are interested in tw o main themes: the extent of constituti onal (en­
trenched) presidential p ower, and the extent of legislative p ower delegated
t o the president by c ongress. These tw o types of legislative p ower are n ot
unrelated. In particular, we argue that the presence of the f ormer may
enc ourage the delegati on of the latter, suggesting als o that students of
presidential systems have regularly mistaken delegated auth ority for the
usurpati on of p olitical p ower by presidents. We rec ognize that any delega­
ti on of auth ority implies the p otential f or agency l oss but intend t o dem on­
strate that the c onditi ons under which delegati on takes place can greatly
c onstrain presidents fr om defying c ongressi onal will.
The chapter begins with a brief discussion of the distincti on between
entrenched and delegated legislative p owers in parliamentary and presiden­
tial systems. We then pr oceed t o examine the sc ope and c onfigurati ons of
presidential p owers am ong systems with elected presidencies, f ocusing on
the veto, the auth ority t o intr oduce legislati on, decree p ower, and emer­
gency p owers. It is in the areas of decree and emergency p owers that the
questi on of delegati on versus entrenched auth ority is m ost pertinent.
Where relevant, we include in the discussion the legislative p owers of
presidents in premier-presidential systems. H owever, m ost such presidents
are acc orded n o such c onstituti onal p owers, and where delegation of legisla­
tive authority takes place in such systems, it generally inv olves delegati on
from the assembly t o the premier or the cabinet. Therefore, the premier­
presidential systems are n ot pr ominent in this discussi on.

DELEGATION VERSUS CONSTITUTIONAL POWERS

A s n oted i n Chapter 2 , one o f the defining criteria o f presidentialism i s that

the president p ossesses s ome legislative p ower. While the executive in any
parliamentary system als o p ossesses legislative p ower, there are tw o impor­
tant distincti ons between parliamentary and presidential systems in this re-
132 Presidents and assemblies
spect. First, a parliamentary executive holds only those legislative powers
that have been delegated to it by the assembly, while a presidential executive
typically has entrenched legislative powers in addition to powers delegated
by the assembly. The second distinction is that because some legislative
powers are constitutionally given to the president, there is wide variation
among presidential systems in terms of how much legislative power their
executives hold.
The importance of delegation in distinguishing parliamentary from presi­
dential executives (and some presidential executives from others) can hardly
be overstated. A parliamentary executive with a secure basis of support
among a majority in the assembly can legislate in a virtually unimpeded
manner. That is, nearly all proposed laws may be initiated by the cabinet and
supported without amendment after their official reading before parliament.
It would be folly, however, to infer from this situation that the parliament is
powerless and that the regime is therefore a "facade democracy." Yet such
assumptions are regularly made with regard to presidential systems where a
congress is observed not to play a significant role in legislating and rather
appears to be bypassed or, at best, acting as a rubber stamp. 1
W hat is going on in such situations? In the hypothetical majority-party
parliamentary case just sketched, a clear equilibrium stems from the consti­
tutional balance of powers between executive and assembly. To the extent
that the majority party can come to agreement internally on policy matters,
cabinet proposals will reflect the general position of the party at large.
W hen the cabinet is a minority or coalition government, finding such a
basic agreement may be more complicated and the scope of agreement
more limited, but still whatever powers the cabinet has to make and imple­
ment law reflect a delegation by the parliament to the executive. In having
been selected as the executive agents of the parliamentary majority, mem­
bers of the cabinet have been charged with the authority to carry out the
will of that majority.
From time to time, almost inevitably, the executive may deviate from the
intended policy direction of the mainstream of the party or parties that
constitute the assembly majority. In the language of principal-agent ap­
proaches, the principals - in this case, the party holding the parliamentary
majority - are said to be suffering agency losses. What happens if these
agency losses become severe but the agent - here, the cabinet - does not
change its ways? The requirement of parliamentary confidence means that
the principals can wield their ultimate sanction, voting for censure and
thereby bringing the downfall of the government.
In a presidential system, it may be more difficult to see what represents
delegation of authority and what represents an executive simply circum­
venting the assembly while the assembly abdicates its own authority. Much
literature on the U.S. system argues that the Congress has abdicated its
1 For a typical example, see Conaghan, Malloy, and Abugattas (1990).
 Legislative powers ofpresidents 133
formal powers to make law, which is instead made by executive regulatory
agencies (Lowi 1979; Fiorina 1982; Moe 1987). Yet, as is argued in
McCubbins and Schwartz (1984), and Kiewiet and McCubbins (1991), Con­
gress retains a number of means for minimizing its agency losses.
In many Latin American presidential systems, the emission of decrees
plays an important role in policymaking. The widespread use of decrees
raises question about the extent to which congresses in such systems main­
tain a role in the legislative process. Unfortunately, in our view, too little
research has been done on the legislative processes of individual presiden­
tial democracies other than the United States. All too often, studies of
particular countries have argued that congresses are rubber stamps or regu­
larly have their powers usurped by the executive (Taylor 1960; Scott 1973;
Levine 1973; Kelley 1977; V azquez Carrizosa 1979; Suarez 1982; Linz
1987; Archer and Chernick 1988). We argue, however, that, as in the
United States, many of these powers have been delegated by congress. As
such, they can also be revoked.
There is another element to presidents' legislative powers beyond delega­
tion: presidents' constitutional powers over legislation vary quite widely
among presidential democracies. In Chapter 8 we present a ranking of
presidential constitutions on the basis of how much legislative authority
their presidents are constitutionally granted. If a given executive is constitu­
tionally powerful relative to the assembly in the legislative process, then we
have a case in which more than delegation is taking place, but not a case of
circumvention, since presidents in such cases are acting within their consti­
tutional authority.
W ith regard to the question of presidential usurpation - or congressional
abdication - of legislative powers, we intend to make the following points
in this chapter. First, delegation of power from the assembly to the execu­
tive is neither usurpation nor abdication if authority is delegated carefully,
such that agency losses can be detected and authority easily revoked. Sec­
ond, to the extent that the constitution grants presidents wide powers, it is
inaccurate to speak of congressional abdication or presidential usurpation.
That is, congress cannot abdicate what it does not have and the executive
cannot usurp what it legally has. Third it is important to examine the
legislative powers that are constitutionally granted to presidents, both to
appreciate the vast differences in the powers of the office across systems,
and to determine whether some configurations of legislative power are less
conducive to smooth assembly-executive relations, and perhaps even more
prone to regime crises.

COMPARING LEGI SLAT I V E POWERS OF PRES IDENCIES

We now consider basic legislative powers that presidents may be constitu­

tionally allocated. First is the veto, but there are many variations on this
134 Presidents and assemblies
power. Override provisions vary, as do provisions such as whether the
president may veto parts of a bill (a partial veto) or must accept or veto the
entire bill (a package veto). If an executive wields a partial veto, this is a
powerful weapon against congress and can render the congress much less
effective in passing, for example, patronage-type legislation than if the
president has only a package veto. There is also considerable variation
among presidential systems in the extent to which lawmaking and rule
making may be undertaken via decrees or under states of emergency.

V A R I A T I O N S O N T H E V E TO

Across the range of presidential systems, the veto is the president's most
consistent and direct connection with the legislative process. Theoretically,
it is considered the primary check of the president on congressional power.
Nevertheless, the nature of the veto power provided in presidential constitu­
tions varies enormously. Highly constrained, the veto can allow the presi­
dent no more than an opportunity to express disapproval for a law, with no
ability to block it. At the other end of the spectrum, it may endow the
president with effective power to designate the specifics of large legislative
packages. Or, of course, the effective power of the veto can fall anywhere
between these extremes. Clearly, variations on the veto can tip the balance
of power between congress and the executive strongly toward one or the
other, and can do so without affecting the principle central to presidentialism
that the origin and survival of the two branches remain independent.
In assessing the impacts of alternate veto formats, it will be helpful to
break the variations into three categories, which we shall address in order:
1 . partial vetoes;
2. pocket vetoes;
3. condition for override.

Partial veto
Partial vetoes (also known as item vetoes) increase presidential power
dramatically. W ith a partial veto, congress cannot present large legislative
packages to the president, which effectively force the latter to accept pork
projects or other objectionable amendments or "riders" in order to pass
desired legislation. The partial veto allows the president to target specific
elements of legislation for veto while promulgating the rest. While it is still
technically a negative power, the partial veto allows the president to pull
legislation apart and so to craft final packages that are more acceptable to
the executive. As a result, presidential power becomes more flexible and
potent than is possible with the cumbersome package veto. Conversely,
one of congress's primary bargaining chips - and the ability to "logroll" -
is lost or weakened.
 Legislative powers ofpresidents 135
Partial vetoes are common in Latin American presidential systems. Bra­
zil, Chile, Colombia, and Ecuador are among the countries that have had
for long periods of time constitutions granting the president a partial veto.
In the Philippines, as in several U.S. states, the partial veto applies only to
the budget or appropriations legislation, while in many other countries it
applies to any legislation to which the president objects in part.
As with package vetoes, the conditions for override vary somewhat
across systems. Overrides of vetoes are discussed below.

Pocket veto
In all presidential systems, the president is required to take some action on
legislation passed by congress - either to promulgate or to veto it - within
a prescribed time period. The period is generally ten or fourteen days, but
in some instances can be shortened to five days by a congressional decree
that action is "urgent." The pocket veto, then, is never unrestricted in the
sense that a president could simply refuse to promulgate legislation, with­
out vetoing it, and thereby kill it without risking override. If a bill is passed
within the last week of a congress, however, the president might wait until
the recess period before vetoing. If the president alone is authorized to call
special sessions of congress, then any action on an override will necessarily
be delayed until the next congress. The pocket veto considerably strength­
ens the president relative to congress by denying congress the last move in
the legislative process - its opportunity to respond to the veto.
More than simply a negative power, the pocket veto is a passive power.
Restrictions are placed on its use in strikingly few presidential systems,
perhaps because the imperative to regulate an official's prerogative to do
nothing is less compelling than the need to regulate presidential actions.
Nevertheless, in both Guatemala and Honduras, the pocket veto is explic­
itly prohibited in the constitution. 2 Congress does not recess until the presi­
dent has either promulgated or vetoed all pending bills, and congress has
had the opportunity to override.

Conditions for override

Among presidential systems, the modal requirement for congressional over­
ride of a veto is a two-thirds majority in both chambers for bicameral
systems, or in the single chamber for unicameral assemblies. It is also
important to point out that among premier-presidential systems, effective
vetoes (those requiring an extraordinary majority for override) are almost
entirely absent. The ubiquity of the two-thirds rule among presidential
constitutions may be testament to the influence of the U.S. Constitution on
2 Guatemala, Article 178 ; Honduras, Article 217.
136 Presidents and assemblies
later framers of presidential systems. Or perhaps it would be less parochial
to suggest that this cutoff for extraordinary majorities has become a
Schelling-esque focal point for constitution builders (Schelling, 1960). At
any rate, the point is that there is no logic inherent to presidential govern­
ment requiring the two-thirds cutoff. The president's power could be re­
duced or increased by setting the cutoff at three-fifths or three-quarters,
respectively, while maintaining a similar executive check on congress. It
has been rare for overrides to require more than two-thirds, but overrides
of less than two-thirds are quite common.
In considering cases with lower thresholds for overrides, we must distin­
guish absolute from simple majority. An absolute majority is 50 percent plus
one of the membership of the assembly (or of each chamber, where applica­
ble), while a simple majority is 50 percent plus one of a quorum. Because
legislation passes in nearly all systems by a simple majority and frequently
the full membership is not present or some members or parties abstain,
raising the requirement for a veto override to an absolute majority implies
that some legislation may fail to clear this hurdle when vetoed by the presi­
dent. In Venezuela, the president may veto a bill, and immediate override
requires a two-thirds majority. A simple majority, however, can return the
bill to the president's desk. The president again may veto the bill, but the
second time around a simple majority in congress can promulgate it (Art.
173). In such a case, the president's "veto" does not ultimately require any
greater a majority to override the president's objections than to pass the bill
in the first place. Although the constitution may refer to the review process
as a veto, then, the process does not constitute an effective executive veto as
it is commonly understood. Rather, it constitutes a formal process by which
presidents can "go public" (Kernell 1986) with their objections to particular
legislation, forcing congress to reemphasize its support for a bill. But a
persistent assembly in the end is sovereign on legislation. This type of "veto"
is found as well in many premier-presidential systems, including France.
The executive check entailed in an effective veto is most important when
the president faces an opposition majority in congress. Where party disci­
pline is tight, therefore, and the requirement for override is a simple major­
ity, congress's immunity from such a check is underscored. This indeed is
the case in Venezuela, where the closed-list ballot structure ensures nearly
airtight party discipline. Coppedge (1992) has furnished examples, from
several administrations, where minority presidents in Venezuela have been
powerless to prevent narrow congressional majorities from legislating
"over their heads." According to Coppedge, since the Venezuelan presi­
dent can be circumvented by congress in the legislative process, there is the
danger that such a president will simply fail to execute laws which are
passed by narrow majorities over executive objections.
In Brazil and Nicaragua, both the package and partial vetoes require
only an absolute, rather than extraordinary majority to override. At first
 Legislative powers ofpresidents 137
blush, the fact that override requires only an absolute majority may seem
to indicate that the provision for a partial veto would be minimal. Even so
restrained, however, the partial veto could be a powerful tool for these
presidents. Consider the classical legislative logroll, in which assembly
members bundle diverse pork barrel spending items together, attaching the
items to legislation which the president regards as critical. Each individual
pork item may benefit a minority of assembly members - perhaps only
one - and could not survive an up or down vote on its own. Through the
logroll, however, assembly members agree to support each other's items in
order to secure a majority for their own. If presidents with only a package
veto find the rest of the legislation more desirable than the pork is objec­
tionable, they will sign it. Now consider the response of a president with a
partial veto, even with a simple majority override. If an override vote must
be held on each vetoed part of the bill, the logroll strategy falls apart. Item
by item, the pork is unpopular across all districts or parties.
In addition to these examples of low thresholds for override, there are
instances when the requirement of an extraordinary majority for override is
required or relaxed only for specific types of legislation. In Colombia be­
fore 1991, override of a veto on appropriations legislation required two­
thirds majorities, whereas override for all other legislation required only an
absolute majority (Art. 88). Although the Colombian budget process was
complicated by other institutional arrangements besides the veto, this ar­
rangement strengthens the president's hand in budget matters relative to
nonfiscal issues. In Paraguay, where the actual experience with nonauthori­
tarian government has been minimal, the constitution requires a two-thirds
majority for override of a package veto, but only a simple majority to
override a partial veto (Art. 158, 159). In Ecuador, on the other hand, a
partial veto can be overridden by two-thirds majority, whereas congress by
itself cannot override a package veto at all. Congress can, however, submit
vetoed legislation to a referendum (Art. 69, 70). In Ecuador, the president
is thus faced with a subtle calculation in deciding whether to exercise the
partial veto. The partial veto can be overridden by a two-thirds vote, while
a package veto cannot be directly overridden by congress at all. However,
Congress can move to have legislation vetoed as a package submitted to a
referendum. Therefore, use of the partial veto may reduce the obstacles to
override on that particular point, since a referendum is not necessary; but it
also could preempt congress from going "over the president's head" to the
people to pass the entire package.
Bolivia's constitution requires a two-thirds majority in a joint session of
congress for override (Art. 77). If the veto were to take place in the two
chambers separately, the power of the less populated departments would
be amplified in the senate, as is usual on all legislative matters in bicameral
systems where representation in the upper house is geographically deter­
mined. Because the vote takes place in joint session, the power of the less
138 Presidents and assemblies
populous departments is dampened somewhat during direct confrontations
between congress and the president. Brazil's votes on overrides similarly
take place in joint session.
As we noted above, the power of the president in premier-presidential
systems usually is relegated to the sphere of government formation and
possibly dissolution of the assembly. As the principle of separation of powers
is compromised, the primary check that Madison saw as being protected by
separate branches is generally absent; premier-presidential constitutions
rarely provide the president an effective veto. The proposed Argentine
premier-presidential model would be an exception, requiring a two-thirds
override. Poland's transitional premier-presidential arrangement after 1989
similarly provides for such a veto. The 1976 Portuguese constitution pro­
vided an absolute veto for the president, by stipulating that no legislation
could become law without the president's signature. This provision, how­
ever, was scrapped in the constitution reform of 1982, leaving the current
president with weak veto power on legislation on most matters other than
foreign policy (where there is a veto and a two-thirds override). In addition,
a number of president-parliamentary regimes have provided for a two-thirds
override on vetoes. These cases included Chile's "Parliamentary Republic,"
Cuba in the 1940s, and Namibia.
Of the areas we have identified in which presidential systems vary, the
veto formats offer the greatest range for potential manipulation. The other
types of variation we review can have tremendous impact on congressional­
executive relations, but the diverse blueprints for the exercise of the veto
can be adapted to adjust the relative strengths of the branches, in large or
small increments, and for specifiable situtations and types of legislation.

OTHER L E G I S L A T I V E POWERS OF PRESIDENTS

We now move o n to the other legislative powers of presidents, including

special authority to introduce legislation, decree powers, and emergency
powers, of executives in presidential and premier-presidential systems. The
preceding section illustrates that across the range of presidential systems
the veto is the most prevalent, flexible, and routine tool by which presi­
dents can affect the lawmaking process. We also saw, however, that some
chief executives in presidential systems, and most in premier-presidential
regimes, do not have an effective veto. Where this is the case, and where
the president faces an opposition majority in congress, we might expect
that an independently elected president should be irrelevant to policy out­
comes. Congress cannot terminate such an executive as under parliamen­
tary government, but even a bare majority can legislate "over the presi­
dent's head," as we saw in the Venezuelan case.
The first question that arises, then, is: Does a president without a veto
necessarily imply outright congressional sovereignty? We shall see that an
 Legislative powers ofpresidents 139
array of non-veto legislative powers exists in various combinations among
regimes with elected presidents. All constitute real lawmaking power. In
most presidential systems, moreover, at least one of these powers exists in
addition to an effective veto. The second question, then, is: In what ways do
non-veto lawmaking powers strengthen presidents relative to assemblies?

Legislative introduction and budgetary powers

The most common non-veto presidential power is the authority to intro­
duce legislation in congress. Most presidents have the authority to intro­
duce legislative proposals, but even where they do not, there is no meaning­
ful restriction on presidential power. Surely any president can find some
member of the assembly to introduce a desired bill. Where provisions
relating to the introduction of legislation become important is in those
cases in which only the president may introduce certain types of legislative
proposals. In particular, many constitutions grant powers of initiative to the
president in budgetary matters.
The importance of legislative introduction, especially where it is the
exclusive power of the executive, is fourfold. In the first place, it takes
presidents out of the position in which they can respond only to take-it-or­
leave-it legislative offerings from congress. In the discussion of congres­
sional approval for presidential appointments in Chapter 6, we used a
simple spatial model to demonstrate the advantage that such asymmetric
initiative power provides to the president. Thus, any president with the
authority to introduce legislation has the potential to counter unsatisfactory
legislative offerings from congress by attempting to engineer a congres­
sional majority for alternative legislation more to his or her liking.
Second, the advantage of initiative is increased when the president is
provided exclusive authority to introduce certain legislation. W here the
executive is given exclusive authority to initiate legislation in certain key
policy areas, as in Brazil, Chile, Colombia, and Uruguay, the president
necessarily becomes the most important legislative actor in those domains.
Even if presidential legislation can be amended in congress, congress­
persons are limited to a reactive position, able only to modify. If the presi­
dent wants no action, then no action can be taken. The bottom line is that
when presidents, or ministers who are exclusively accountable to presi­
dents, are allowed to initiate legislation on their own, they are generally
among the primary forces in the legislative process.
Third, legislative power is greater still when the president can make take­
it-or-leave-it proposals - those that cannot be amended. Although such
restrictions on amendment are rare, they exist in Uruguay, where congress
cannot increase tax exemptions proposed by the president or lower pro­
posed maximum prices. Similar restraints existed in Chile before 1973.
Congress could not increase the president's proposed ceiling for spending
140 Presidents and assemblies
on particular items without simultaneously creating new appropriations or
reducing other items. The current Chilean constitution goes a step farther:
The president's budget bill imposes a spending cap that cannot be amended
(Art. 64), meaning that congress can shift the allocation of appropriations
but only within a fixed budget. Some other constitutions have gone farther
still in granting initiative to the president in budgetary matters. For exam­
ple, in the Korean constitutions of 1962 and 1987, the congress may not
even move expenditures around from item to item, let alone increase items
of expenditure; it may only reduce amounts proposed for given items.
Fourth, presidents are commonly able to influence the legislative agenda
by combining introduction power with dramatic advantages in information.
The centralized nature of executives provides them an advantage over
assemblies in collecting and managing information. Much of the most im­
portant information for putting together viable legislative packages is avail­
able primarily in the ministries and bureaucracy of the executive branch. In
many presidential countries, moreover, congress is notoriously poorly
funded and understaffed. W here presidents have extraordinary powers in
the budget process, they may even be able to cultivate this informational
advantage by keeping congress underfunded.

Decree powers
Presidential decree power is related to the power to introduce legislation in
that it reverses the standard relationship by which congress proposes ac­
tions while the president serves as a veto gate. But it goes farther. A bill
introduced by the executive still must pass through the ordinary legislative
process in congress and then be submitted back to the president for ap­
proval or, if congress has modified beyond what the president will accept,
veto. A decree law is not merely a bill introduced by the executive, but
rather a law or regulation issued by the executive. In some cases, congress
must specifically act to accept or "veto" a decree. In other cases, a decree
stands unless specifically rejected by congress. In still others, and more
significantly for the balance of powers, a decree is simply law made by the
executive and not subject to congressional review.
We distinguish between two variants of decree authority granted to presi­
dents. The first is the constitutional authority to legislate by decree in
specified policy areas, whereby the decree is law unless it is overturned by
congress. The second is the ability to have the authority to legislate by
decree delegated by congress to president.
The first form, constitutional decree authority, is found, currently with
various limitations and permutations, in a few countries, including Brazil,
Chile, and Iceland. 3 Under Brazil's constitution of 1988, the president may
3 Historically, the president in Weimar Germany had extensive decree powers, as discussed in
Chapter 4.
 Legislative powers ofpresidents 141
decree "urgent" laws that are immediately effective; but if they are not
passed by congress within forty-five days, they lose effect. Since the presi­
dent may decree the same matter once again, effectively this "urgency"
clause grants the president decree power that is not subject to rescission. In
Chile under the 1989 constitution, the president, with the consent of the
cabinet, may decree expenditures that cannot be ignored without "causing
serious detriment to the country" (Art. 32, Sec. 22). In Iceland as well, the
president may issue decrees with the force of law when parliament is not in
session. Such laws are, however, subject to approval or rejection as soon as
parliament reconvenes, which it may do under its own initiative at any
time.
This first type of decree power allows the president to dominate the
agenda by which applicable legislation is considered. However, it would be
misleading to argue that any of these presidents - assuming competitive
electoral processes - are able to use their decree powers rampantly and
without any congressional checks. In each case, decree powers are embed­
ded within a constitional structure that imposes many restraints on presi­
dents in other aspects of the legislative process. For example, although the
Brazilian president theoretically may declare a matter urgent, wait for
congress to vote to rescind it, then re-decree it repeatedly, such a deter­
mined president might find many obstacles placed in his or her path by
congress. Foremost among these would be the ability of the congress to
override with relative ease (an absolute, as opposed to a two-thirds major­
ity) any presidential veto. Additionally, of course, the congress can use its
budgetary powers, even though they are somewhat restricted, to cut appro­
priations desired by the president. Thus the urgency powers give the presi­
dent considerable leeway to shape the legislative agenda, but an all-out
confrontation is one that the president could not win, given the consider­
able legislative weapons retained by the congress.
In Colombia before 1991 the president with the consent of all cabinet
ministers could declare a State of Siege or a State of Economic Emer­
gency without congressional sanction and could issue decrees with the
force of law. Given this authority, and the existence of one or the other
form of state of exception for almost three-fourths of the time since 1958
(Archer and Chernick 1988), the Colombian president may be said to
have had significant powers to enact laws on his own. On a careful
reading of the constitution, however, what at first appears as a virtual
absence of institutional constraint on these supposedly exceptional pow­
ers reveals itself as institutionalized delegation. In a 1968 amendment,
Congress assumed the power to reject or amend a State of Emergency
declaration, although not a State of Siege. The constitution itself could be
amended by absolute majorities in successive congresses, so one can imag­
ine that even the president's constitutional decree power could be rela­
tively easily revoked if it were abused to the point where it no longer
142 Presidents and assemblies
served congressional interests. 4 As evidence that this is so, consider that
the constitution provided that any laws issued during such periods lapse
when the period of exception ends. That many of these laws were subse­
quently converted into ordinary legislation by congress (Archer and
Chernick 1988) simply suggests that the "emergency" decrees passed by
the president did not deviate substantially from congressional majority
preferences. If so, then the Colombian legislative process, often por­
trayed as having been abdicated by the congress, was in equilibrium after
all, albeit an equilibrium based in an institutional context that was
skewed in the short term toward an executive that can always call yet
another state of exception. 5 As such, the Colombian president's "extraor­
dinary" and "temporary" decree powers were de facto ordinary legislative
powers that served to overcome the slowness and logrolling that charac­
terize ordinary congressional legislation, but not to run roughshod over
congressional preferences and prerogatives.
In most constitutions that permit presidential decree legislation, there
are countervailing powers in the hands of the assembly. In Brazil and
Colombia, the relative ease of veto overrides is one such countervailing
power. In France, the German Weimar Republic, Iceland, in the Korean
constitution of 1948 and the new Colombian constitution (as well as in
Brazil after a constitution amendment in 1961), the assembly holds the
power of censure over cabinet ministers. In all cases, the assembly holds at
least restricted powers over the budget. That assemblies retain means to
keep presidents from using their decree authority rampantly, however,
does not imply that these powers do not grant presidents a formidable
advantage in the legislative process. W hile a president must consider costs
that may be incurred when a decree is issued that goes beyond known
congressional majority sentiment, the existence of the authority allows
presidents to bring legislative outcomes closer to their ideal point than
would be feasible if only congressionally sanctioned measures could be­
come law. We return to this later on in the form of a spatial model when we
consider decree powers delegated by congress.

4 In addition, and as a more short-term and piecemeal solution, Archer and Chernick (1988)
report that the Supreme Court of Colombia has increasingly reviewed, and in some cases
overturned, legislation issued during states of exception. Under the 1991 constitution, the
emergency powers are restricted in time, subject to senate approval before extension, and
more closely guarded by congress owing to cabinet members' dependence on the confidence
of the chambers.
5 That this latter observation about the ever present threat of a new state of exception is not
the full explanation of congressional acceptance of these decrees is suggested by the discus­
sion in Archer and Chernick (1988). They indicate that presidents cannot hope to undertake
any fundamental reform under states of exception, presumably because the congress would
not convert such decrees into ordinary legislation. This, then, is an argument that congress
accepts only those that are not much at variance with its own preferences.
 Legislative powers ofpresidents 143

Distinguishing decree legislation from rule making

Any student of Latin American politics will immediately recognize that
"decrees" are issued far more commonly and in far more countries than is
indicated here. Are we therefore overlooking some factor of presidential
power that cannot be explained by reference to "formal" constitutional
powers? The problem is largely semantic. In Latin America the term
decreto (decree) is used more widely than we use it in this book. Almost
any executive act, from a regulatory decision through a pardon to constitu­
tionally sanctioned lawmaking, is called a decree. We use the term more
narrowly; only those laws which the president can initiate and which main­
tain the force of law unless specifically rescinded (vetoed) by congress are
meaningfully called decree laws. Although one might not realize it by
reading the literature on Latin America, few systems grant this form of
decree authority. Moreover, most allow it only in constitutionally specified
policy areas, as already discussed, and embed the authority within constitu­
tional rules that significantly restrict presidential legislative powers in other
ways.
Most other cases of decrees in Latin America or elsewhere are regulatory
matters rather than new legislation. As such, the range of options that may
be "decreed" by the president (or, in some cases, a minister) is limited by
legislation already passed by congress and, where necessary, approved by
the president through ordinary legislative procedures. These regulatory
decrees are thus not much different from the rule-making authority granted
to independent agencies in the United States, where the term "decree" is
never used. If the parameters within which the executives of some systems
issue regulations are much more vaguely circumscribed than in the United
States, one reason is the greater legislative powers of such presidents in the
ordinary legislative process compared to the United States. For example,
where the president may veto part of any bill, congressional ability to fine­
tune legislation is sharply limited and delegation of specific authority to the
executive is an obvious solution.

Distinguishing decree legislative authority from emergency powers

Decree legislation, as we have defined it, must be distinguished not only
from regulatory decrees (rule making) but also from states of emergency.
In nearly all presidential systems other than the United States, and in some
premier-presidential systems as well, emergency powers allow the presi­
dent to suspend civil liberties and take direct command of local agencies in
times of unrest. Constitutions differ in the degree to which congressional
ratification or oversight of states of exception is required. In Costa Rica
and Namibia, for example, a state of emergency may be declared only with
144 Presidents and assemblies
the consent of two-thirds of the assembly. Several require congressional
action, but only a vote of a simple majority to continue the state of emer­
gency. Article 16 of the French constitution provides broad and rather
vaguely defined authority to the president when the "independence" or
"integrity" of the nation is threatened. Requiring only consultation with
the premier, the presidents of the two chambers of the assembly, and the
Constitutional Council, the French provision offers potentially the most
extensive emergency powers of any presidency. If the premier is of a differ­
ent party or coalition from the president, the requirement of "consultation"
with the premier may imply a check on the president. However, it is poten­
tially significant that the constitutional language is not stronger, perhaps
requiring the "consent" of the premier and the assembly, rather than mere
consultation. Emergency powers are untested in France in situations in
which the president and premier are politically divided. We can say only
that language such as that in the French constitution carries the potential
for a constitutional crisis.
All things considered, only in very few regimes does a state of emergency
imply legislative powers. The case of Colombia has already been ad­
dressed. In El Salvador, the president may authorize the disbursement of
unappropriated funds to meet needs created by the emergency. And in
Ecuador and Bolivia, presidents may advance the collection of taxes and
other revenues ahead of schedule (but not authorize any new ones), in
order to meet the budgetary needs generated by the emergency. In Chile
before 1973 the president was allowed an amount equivalent to 2 percent of
the budget to meet "emergency" needs, at his discretion. Other than these
examples, presidents do not enjoy special legislative decree powers under
states of siege or emergency.

Delegation of decree powers

Under another variant, decree power may be delegated to the president
by congress for a specified time and in certain policy areas. The constitu­
tions of Chile, Colombia, Peru, and Venezuela specifically allow for such
delegations, although we expect that unless such delegations are explicitly
prohibited, they remain an option for congress. Because this power may
be retracted prematurely by statute, the effects of delegating such author­
ity depend on the requirement for veto overrides. Interestingly, in Peru
and Venezuela, the congress can retract a delegation of decree authority
by simple majority, even over a presidential objection, since in both cases
there is no requirement for an extraordinary majority to override a veto.
In some other cases including Colombia before 1991, an absolute majority
is required. W here one party holds a majority of all members, but less
than two-thirds, a veto requiring no more than an absolute majority to
override makes the delegation of decree powers far less costly than if a
 Legislative powers ofpresidents 145

C P
Figure 7. 1 . Presidential (P) and congressional ( C) ideal points on hypothetical
policy: indifference point for simple majority Uc1) and two-thirds majority Uc2)

greater majority is required to retract such power. In these countries,

then, the actual effect of delegating decree authority is merely to transfer
the powers of legislative initiative to the president for the duration of the
decree period. If the president were to decree legislation objectionable to
a majority in congress, we expect the decree power to be promptly re­
tracted. Under the new Chilean constitution, on the other hand, once
decree power is delegated, it cannot be prematurely retracted unless ei­
ther the president capitulates, or congress overrides the presidential veto
by a two-thirds vote of each chamber. Under these conditions, unless one
party or stable coalition holds at least two-thirds of the seats (a relatively
rare occurrence), delegation would entail an abdication of legislative au­
thority to the president.
The situation is modeled in Figure 7.1. For a given issue on which the
president might issue a decree, point P is the president's ideal policy and
point C is congress's. lc112 represents the point at which a majority of con­
gress is indifferent between the status quo and new legislation which lies
closer to the president's ideal than to congress's. This point is not necessar­
ily the same as point C, since the power of initiative now belongs to the
president. Theories of social choice (Arrow 1951; McKelvey 1976) show us
that congressional majorities will rarely be unique to any given policy
point, 6 so we expect that congress's ideal point and its indifferent point
should not be exactly the same. but lc213 , much farther from point C, is the
point at which two-thirds of members are indifferent between a presidential
initiative and the status quo. Put another way, lc112 is the point beyond
which any presidential decree would be revoked by a majority, but lc213 is
the point beyond which a presidential decree would have to go in order for
congress to override the veto that would be likely if it attempted to revoke
decree power. W hat Figure 7.1 shows is the area, in "policy space," on any
given issue that a congress abdicates if it delegates decree power to a
6 With either strong party discipline (for example, with closed-list electoral systems) or
bicameralism (see Miller and Hammond 1989) or both, an assembly may approximate a
unique majority position more closely than most models of social choice instability assume.
Even if so, the relevance of distinct "ideal" and "indifference" points in understanding
delegation remains. Over time, the ideal legislative point of even a disciplined majority of
bicameral legislature may shift from where it was when the delegation was made, while the
agent's preferences might not shift, or might shift in a different way. Thus the model that we
develop here should be applicable, to varying degrees, to any situation in which an assembly
delegates decree powers.
146 Presidents and assemblies
president that can only be revoked with legislation requiring an extraordi­
nary majority. In this model, the amount of abdication is the space between
points lct;2 and Ic213. The space between lc112 and C, in turn, is the amount of
abdication entailed by delegating decree power that can be revoked by a
simple majority.
W here delegation does not entail abdication, we might expect congress
to rely on delegation as a means of overcoming coordination problems in
the legislative process. The incentive to delegate, of course, would depend
on the ability of congress to minimize agency loss to the president (see
Kiewiet and McCubbins 1991), and the extent of its own coordination
problems. The evidence on this count is suggestive. Archer and Chernick
(1988) report that Colombian presidents held some decree powers 75 per­
cent of the time between 1958 and 1988, not necessarily because of threats
to the political order but because legislative coalitions were so fragile.
Unfortunately, Archer and Chernick do not distinguish between decree
powers delegated by congress and "economic emergency" powers declared
by the president. In Venezuela, decree authority has been granted only to
presidents with majority support in congress - three Acci6n Democratica
presidents have requested and received decree powers since the current
constitution was promulgated in 1961. In no case was the authority over­
turned prematurely. During his first presidency, however, Carlos Andres
Perez issued the staggering total of 830 decrees and established fifty-one
government commissions on a broad range of policy areas in the space of
twelve months. Coppedge (1988) reports that in the one subsequent case of
delegation that he studied, the party in congress was cautious in limiting
the extent of decree powers it authorized.
W hile the specific conditions surrounding decree powers are diverse,
they are a common form of non-veto legislative authority for presidents. In
some cases, decree powers surpass the veto as the prime source of presiden­
tial power in the legislative process. This is most clear in Peru and Venezu­
ela, where vetoes are ineffective. It is also true in Colombia, where decree
power in the economic sphere can be seized by the president without a
clear congressional check. In Chile, the prospect of delegated decree power
is formidable, but it remains to be seen whether congress would ever
delegate such authority to a president, given the difficulty to retract it.

CONCLUSION

This chapter began by noting the distinction between entrenched, constitu­

tional legislative power and delegated presidential power. We have dis­
cussed the extent of each type across systems with elected presidencies.
Regarding the first type, we concluded that although the veto is the most
common legislative power of presidents, it is neither uniformly configured
across systems, nor the only significant entrenched power. Our examina-
 Legislative powers ofpresidents 147
tion of vetoes demonstrates that a number of factors determine its impact
on presidential strength, including the conditions for override, the ability to
target specific items for veto, restrictions on what kinds of legislation may
be vetoed, and the presence or prohibition of pocket vetoes. Moreover, we
have demonstrated that the authority of the president to introduce legisla­
tion makes him or her a central player in the legislative process from the
outset, and a privileged one if the president is accorded exclusive authority
to introduce certain types of legislation, or if restrictions are placed on
congressional amendments. We have noted that those emergency powers
granted to presidents that do not require case by case approval by congress,
with a few exceptions, do not include legislative authority. And we have
argued that where other, nonemergency decree power exists, it also is
exercised based on delegation from congress more often than has been
recognized by students of presidential government.
On this subject, we stress that congresses that appear to be passive,
reactive players in the legislative process might well be those that have
delegated legislative authority to the president. But, we argue, it would be
to commit a fallacy of equilibrium to declare these congresses ineffectual
without examining the terms on which authority is delegated. As we have
discovered, in most systems where decree authority can be delegated, for
instance, it can also be retracted easily by congress.
We have noted that among students of Latin American politics (who, for
geographical reasons, most frequently encounter presidencies), the assess­
ment that decree power is more pervasive and broad than it actually is may
be based in part on a semantic misinterpretation. That is, the Spanish term
decreto is used to describe what we call in English regular rule-making
authority, as well as to demark real legislative decree power. However, we
suggest that real decree power is most likely to be delegated to presidents
where certain conditions hold. One is that the president already possesses
strong entrenched powers such as a partial veto or the exclusive right of
legislative introduction, and so is able to fine-tune legislation regardless of
congressional will. Another condition is the fragility of congressional major­
ity coalitions, perhaps in multiparty systems, which may encourage the
delegation of decree authority to break congressional deadlock. Ulti­
mately, we have seen that the legislative powers - entrenched as well as
delegated - of presidents greatly vary across systems. This should no
longer be surprising, as we have drawn similar conclusions in Chapters 5
and 6, on the comparative presidency. In the following chapter, we turn to
the task of systematically assessing the scope and configurations of all
presidential powers - legislative and governmental - and drawing conclu­
sions regarding the relationship between presidential power and regime
performance.
 8
Assessing the powers of the presidency

In this chapter we undertake a process of assessing just how powerful a

president is in constitutional terms. We identify two basic dimensions of
presidential power: one concerning power over legislation, the other en­
compassing nonlegislative powers, including authority over the cabinet and
calling of early elections for congress. Additionally, because the latter pow­
ers are directly related to the question of separation of powers, we provide
a comparison of presidential powers over the composition of cabinets to
separation of executive from assembly. There are two main, related lessons
we shall be able to draw from this exercise. First, systems that score high on
presidential powers, and in particular those that are extreme on presiden­
tial legislative powers, are often those systems that have exhibited the
greatest trouble with sustaining stable democracy. Second, systems that
give the president considerable powers over the composition of the cabinet
but also are low on separation of survival of assembly and executive powers
likewise tend to be among the "troubled" cases. There are thus left two
basic clusters that, we argue, are "safer" for the success of democracy: (1)
those with high separation of survival of powers but low presidential legisla­
tive powers; and (2) those with low separation of survival but also low
presidential authority over the cabinet. These categories happen to be
those approximating the (more or less "ideal") types we have discussed in
previous chapters: presidential and premier-presidential, respectively. The
more troubled regimes prove to be hybrids.

POWERS OF THE PRESIDENCY : LEGISLATIVE POWERS

To assess presidential powers, it is possible to devise a simple interval

scoring method on each of several aspects in which systems with elected
presidencies vary. The first set of aspects entail legislative powers constitu­
tionally granted to the president. These aspects are the veto, the partial
veto, presidential authority to legislate by decree, exclusive right to initiate
certain legislative proposals, budgetary initiative, and power to propose
referenda. Aspects of presidential power apart from the legislative domain
include cabinet formation, cabinet dismissal, lack of assembly censure , and
 Assessing the powers of the presidency 149
dissolution of the assembly. The scores on each aspect of each dimension
can then be summed to arrive at an overall indicator of presidential powers
on the respective dimension. This is not a perfect method - for instance, it
might be justifiable to weight some dimensions more than others - but it is
preferable to a purely nonquantitative, impressionistic ranking or to no
assessment of comparative presidential powers at all.

Package veto/override. As discussed in Chapter 7, if presidents are to have

any authority to shape the legislative output of the congress, they must be
provided with a veto over legislation. However, if this veto were absolute,
no piece of legislation objected to by the president could ever pass, hence
the president would be extremely powerful. The latter scenario existed in
Chile in the mid-nineteenth century. Something similar, but not quite as
strong, exists today on nonbudgetary legislation in Ecuador. A veto cannot
be overridden by congress alone, but the congressional majority can submit
the vetoed bill to a binding referendum, thereby potentially weakening the
absolute nature of the veto. In most other cases, as in the United States, a
veto may be overridden by a two-thirds majority of the congress. As seen
in Table 8. 1, we have given a score of 4 to a veto that may not be overrid­
den, 2 to one requiring a two-thirds override, and O to one that can be
overridden by the exact same majority that passed the original bill. Interme­
diate cases are also possible. 1 If the majority required is more difficult for
congress to obtain than a two-thirds majority, but the veto is nonetheless
not absolute, we score it a 3. We include Ecuador as a 3, because the
provision for a referendum means that the president's veto is not final, but
it is costly for congress to seek overrides with great frequency. At the other
end of the scale, we score at 1 any case in which the veto override requires
more than a simple majority but less than two-thirds. Included here are the
several cases in which the majority is 50 percent plus one of the whole
membership (as in Nicaragua), rather than half plus one of a quorum (as in
Venezuela).

Partial veto. As discussed in Chapter 7, some presidential systems allow the

president to veto part of a bill rather than simply being required to accept
or veto the entire package. The majority needed to override a partial
(item) veto varies, as does the override majority for package vetoes. Usu­
ally the majority is the same for both types of vetoes (Ecuador is an
1 We have not considered separately those cases in which an override is voted in a joint
session of both chambers of a bicameral assembly (Bolivia, Brazil) , rather than by both
chambers voting separately. Whether such a provision strengthens or weakens presidential
power is contingent on other factors, including how the two houses are constituted and
whether the president would be more likely to find favor in the upper house (the power of
which is diluted in a joint session with a typically larger lower house). At any rate, the
difference in scoring for these systems would be marginal.
 150 Presidents and assemblies

Table 8 . 1 . Powers of popularly elected presidents

LEGISLATIVE POWERS
Package Veto/Override Partial Veto/Override
4 Veto with no override 4 No override
3 Veto with override requiring majority greater than 3 Override by extraordinary majority
2/3 (of quorum) 2 Override by absolute majority of whole
2 Veto with override requiring 2/3 membership
I Veto with override requiring absolute majority of I Override by simple majority of quorum
assembly or extraordinary majority less than 2/3 0 No partial veto
0 No veto; or veto requires only simple majority
ovenide
Exclusive Introduction of Legis­
Decree lation (Reserved Policy Areas)
4 Reserved powers, no rescission 4 No amendment by assembly
2 President bas temporary decree authority with few 2 Restricted amendment by assembly
restrictions I Unrestricted amendment by assembly
I Authority to enact decrees limited 0 No exclusive powers
0 No decree powers; or only as delegated by
assembly

Budgetary Powers Proposal of Referenda

4 President prepares budget; no amendment 4 Unrestricted
permitted 2 Reslricted
3 Assembly may reduce but not increase amount of 0 No presidential authority to propose referenda
budgetary items
2 President sets upper limit on total spending,
within which assembly may amend
Assembly may increase expenditures only if it
designates new revenues
0 Unrestricted authority of assembly to prepare or
amend budget

NONLEGISLATIVE POWERS
Cabinet Formation Cabinet Dismissal
4 President names cabinet without need for 4 President dismisses cabinet ministers at will
confirmation or investiture 2 Restricted powers of dismissal
3 President names cabinet ministers subject to 1 President may dismiss only upon acceptance
confirmation or investiture by assembly by assembly of alternative minister or cabinet
President names premier, subject to investiture, 0 Cabinet or ministers may be censured and
who then names other ministers removed by assembly
0 President canot name ministers except upon
recommendation of assembly

Censure Dissolution of Assembly

4 Assembly may not censure and remove 4 Unreslricted
cabinet or ministers 3 Restricted by frequency or point within term
2 Assembly may censure, but president may 2 Requires new presidential election
respond by dissolving assembly I Restricted: only as response to censures
"Constructive" vote of no confidence (assembly 0 No provision
majority must present alternative cabinet)
0 Unreslricted censure

exception). The power of the partial veto when it requires a two-thirds

override is so formidable a weapon in the hands of the president that we
have scored such a power as a 3. (An absolute partial veto would, of
course, be a 4. ) W here a majority of the whole membership is required to
 Assessing the powers of the presidency 151
override (an abs olute maj ority), we give it a 2, while a simple maj ority of a
qu orum is a 1. The reas on that a simple maj ority override of a partial vet o
is n ot sc ored as O is that a part of a bill that was passed because of a l ogr oll
may n ot pass when sent back t o c ongress after a veto, as the wh ole package
m ost surely w ould. As suggested in Chapter 7, members of c ongress wh o
v oted for an amendment only t o gain assent t o other parts of the bill no
l onger have incentive t o pass the amendment when the president already
has pr omulgated the other parts of the bill.

Decree. We refer here t o the ability of the president t o make law. Such
p ower sh ould be distinguished fr om decrees that are of a regulatory or
"rule-making" nature, as we saw in Chapter 7. "Decree" here refers t o the
auth ority t o make new laws or suspend old ones without the p ower of
decree first having been delegated through enabling legislation. The m ost
severe f orm of presidential lawmaking p ower is f ound in Brazil, where the
president may declare a matter "urgent. " These matters have the f orce of
law up on such a declarati on and remain law unless c ongress explicitly v otes
to rescind within f orty-five days. Since the president is n ot prevented from
simply redeclaring the same measure upon such c ongressi onal action, a
determined president can simply make law over the head of the c ongress.
Rescissi on in this case is ineffective and we sc ore this a 4. C ol ombia before
1991 is an example of the sec ond m ost p owerful decree auth ority, f or
reas ons detailed in Chapter 7. Unanimity of president and cabinet was
required, but because the president al one app ointed and dismissed minis­
ters, this was a minimal restraint. We score such cases a 2. Cases such as
C ol ombia 1991, in which the president and a cabinet that d oes not depend
exclusively on assembly c onfidence can impose a state of "emergency" and
legislate with little or n o inv olvement by the assembly, are sc ored as 1. Als o
at 1 are pr ovisi ons granting the president discreti onary spending n ot t o
exceed a c onstituti onally specified share of the budget in the event of
executive-declared "emergencies."

Exclusive introduction of legislation. W here the assembly is barred from

c onsidering legislati on in certain p olicy areas unless the president first intr o­
duces a bill, the president p ossesses a p owerful agenda-setting p ower: If
she or he d oes n ot want a matter discussed, it will n ot be discussed. The
greatest p ower w ould be if the president c ould require a strict up or d own
vote, n ot subject t o amendment by the c ongress. Such a p ower exists
n owhere, but c ould be pr ovided f or. If it existed, we w ould sc ore it a 4.
C onsiderably weaker, but still significant, is any restricti on on amendment,
which w ould be sc ored a 2. The actual p owers in m ost systems that grant
exclusive p ower of legislative intr oduction t o the president are weaker still:
Just as with any bill, c ongress may amend it or reject it outright, s o we give
these c ountries only a 1 on this p ower.
152 Presidents and assemblies
Budgetary initiative. Several constitutions stipulate that it is the president
who initiates the annual budget bill. However, this alone would not be a
formidable power. There will be a budget, whether or not the congress may
initiate one, so giving the right of initiative to the president means little
unless there are restrictions on congressional amending ability. If the presi­
dent's budget could only be approved or rejected, while congress obviously
would be far from powerless, its ability to attend to specific political needs
via the budget would be sharply curtailed. It would be like the presidential
package veto already discussed: a blunt instrument, merely requiring that
the initiating agency (here, the executive) make the whole package mini­
mally acceptable to the majority of legislators. Such a power would be
scored a 4. Variations on restrictions on amendment exist elsewhere, from
prohibition on congressional increase in the amounts allocated to any item
or the creation of new items (scored a 3) to a ceiling on overall spending,
within which the congress may move funds around (scored a 2), to a re­
quirement simply that congress must designate the revenue source of new
expenditures not proposed by the president (scored a 1).

Proposal of referenda. In some systems, the president is granted the power

to propose referenda, usually on matters that have been previously re­
jected by the assembly. If this power is unrestricted, we score it a 4. In
some cases, there are various restrictions on the president's power to sub­
mit legislative proposals to referenda. For example, in France the constitu­
tion requires that a proposal be made jointly with the cabinet or else by a
concurrent resolution of the two houses of the assembly. In Guatemala the
president may submit any matter of "special importance" to a referendum,
but so may the congress. If only the president had this power, as is the case
in Ecuador (except on vetoed bills, as just discussed), it obviously would be
a 4. In such a case, the ability of congress to give consideration to amend­
ments to a proposed bill would be blunted by the realization that the
president could always submit the matter to a plebiscite, in which case the
president's preferred text is voted up or down. However, uniquely in Guate­
mala, the congress also has the power to submit matters to referendum,
thus bypassing a possible presidential veto. This symmetry to the process of
initiating a referendum justifies a "restricted" score (2) on the president's
power.

NONLEGISLATIVE POWERS

In the realm of nonlegislative powers are presidential authority over the

cabinet and the power to dissolve the assembly and call new elections.

Cabinet formation. In some systems the process of cabinet formation de­

pends entirely on the president's choices. If the president names members
 Assessing the powers of the presidency 153
of the cabinet without any need for confirmation of appointees by the
assembly, we score this a 4. W here the assembly (or a special committee
elected by it, as in the Philippines) must approve nominations made by the
president, the president's ability to get "ideal" ministers is weakened, thus
we score such cases a 3. Considerably weaker are instances in which the
president first nominates a premier, needing parliamentary confirmation
(or investiture), and the premier then nominates the rest of the cabinet
(with or without presidential involvement). Such cases are scored a 1. The
weakest of all, and thus scored a 0, is any case in which the president is
barred from nominating anyone to a cabinet post, including premier, ex­
cept upon the prior recommendation of the assembly, or an agent thereof,
such as a speaker of the house.

Cabinet dismissal. Many presidents have the right to dismiss members of

the cabinet (or the entire cabinet) at will. Such cases are scored a 4. In
other cases, the president has no such power, giving the president a 0. We
have a few instance in which the president may dismiss ministers, but the
power is restricted. In Portugal since 1982, for example, the president
may take such action only when he or she can justify it as a response to a
threat to the democratic institutions themselves. While presidents need
not be very creative to find a "threat" to democracy when they dislike the
policies of some ministry that enjoys the confidence of the assembly, such
a move in times of no obvious threat to the country could be subject to
judicial review and may entail political costs. Thus the power is weaker
than if the president may simply dismiss a minister or cabinet without
even being required to give a justification. There could also be a stronger
restriction requiring the president to propose an alternative minister or
government that would need assembly approval before the incumbent
minister or cabinet could be replaced. Such a "constructive" dismissal
would be scored a 1.

Censure. The power of censure is actually, of course, not a presidential but

an assembly power. However, it directly concerns the authority of the
president over the cabinet and is clearly separate from cabinet formation
and dismissal. Accordingly, when there is no censure, meaning that only
the president may remove ministers, the president's authority is strongest,
hence the score 4. W hen censure is unrestricted, the president's score is 0.
There are some intermediate cases - for example, Cuba, in which a minis­
ter could not be censured in his or her first six months in a given post - that
are scored 3 because the restriction limited the extent to which ministers
were genuinely subject to assembly confidence. If a vote of censure is tied
to a process that leads to dissolution of the assembly, the score is 2. When a
censure must be "constructive," requiring the assembly majority to pro­
pose a new minister or cabinet at the time of censure, the score is 1.
154 Presidents and assemblies
Dissolution of the assembly. Some presidents are permitted to dissolve the
assembly at any time, a power that is scored a 4. Others have dissolution
powers but with restrictions. If the restriction is defined either in terms of
frequency - once a year, for example, as in France - or in terms of time
point within the term - not in the last six months of the president's term, as
in Portugal - we score it a 3. Another restriction that renders the power of
dissolution more symmetrical with regard to the assembly's powers is
where the president, too, must stand for reelection along with the assembly
after a dissolution (as in Namibia). Hence we score such situations as 2.
Weaker still (scored as 1) is where dissolution may be invoked only after
censure or a certain number of censures by the assembly (as in Peru and
Uruguay).
Table 8.2 shows the scores on our two dimensions of presidential power for
a nearly exhaustive sample of democracies that have had popularly elected
presidents. In a few instances, two lines are needed for a given case because
the president's powers are different according to the type of legislation in
question. In such cases only the aspects of presidential power that differ are
shown separately, with the main line being simply an average of the two
lower scores. We now turn to graphing these regime's scores in order to
visualize how powerful different presidents are in different dimensions.

POWERS : LEGISLATIVE VERSUS NONLEGISLATIVE

Figure 8.1 graphs the strength of the presidents across all the cases in two
dimensions. Thus the very strongest presidents in both dimensions would
be found in the upper right region, while presidents with no power in either
dimension would be at the origin (0,0). Most of the systems commonly
understood to be presidential are located on the vertical line running at 12
on the dimension of nonlegislative power. These systems are thus all identi­
cal in the extent of authority they give to the president over the government
and the assembly: All have exclusive authority over the cabinet and none
may dissolve the congress. These systems differ only on the dimension of
presidential legislative powers, but there is a very wide range of variation
on this dimension. Two systems are located farther to the right - Chile 1989
and Paraguay - owing to the additional power granted to the president,
who may dissolve the congress. To the immediate left of this group are
those presidential systems in which the assembly must confirm cabinet
appointments. All those farther left provide for censure, including the
several cases of premier-presidentialism. We shall see later another way to
visualize these cases that clarifies the distinction between presidential and
premier-presidential regimes. For now, we attempt just to assess an overall
measure of presidential power, as located in a two-dimensional space.
We have divided the space of Figure 8.1 into six regions, numbered
counterclockwise from the upper right. Region I consists of the very power-
 Assessing the powers of the presidency 155

Table 8.2. Powers of popularly elected presidents by country

Pack. Part. Exel. Budg. Cab. Cab. Cen- Disso.

Country Veto Veto Decree Intro. Power Refer. Tar. Form. Dism. sure lution Tar.
Argentina
current 2 0 0 0 0 0 2 4 4 4 0 12
proposal 2 0 0 0 0 0 2 1 2 0 4 7
Austria 0 0 0 0 0 0 0 1 0 0 4 5
Bolivia 2 0 0 0 0 0 2 4 4 4 0 12
Brazil
1946 2 3 0 I 1 0 7 4 4 4 0 12
1988 I 2 4 1 I 0 9 4 4 4 0 12
Bulgaria 0 0 0 0 0 0 0 0 0 0 0 0
Chile
1 891 2 3 1 1 1 0 8 4 4 0 0 8
1925 2 3 I 1 I 0 8 4 4 4 0 12
1969 2 4 I 2 1 2 12 4 4 4 0 12
1989 2 0 0 1 2 0 5 4 4 4 3 15
Colombia
pre-1991 1 .5 2.5 2 I I 0 8 4 4 4 0 12
(expend.) 2 3
(other) 1 2
1991 1 2 I 0 1 0 5 4 4 0 0 8
Costa Rica 1 0 0 0 0 0 1 4 4 4 0 12
(budget) 0
(other) 2
Cuba 1940 2 0 0 0 0 0 2 4 4 3 0 II
Dominican 2 0 0 0 0 0 2 4 4 4 0 12
Republic
Ecuador 1 .5 1 .5 1 0 0 2 6 4 4 0 0 8
(budget) 0 0
(other) 3 3
El Salvador 2 0 I 0 0 0 3 4 4 4 0 12
Finland 0 0 0 0 0 0 0 4 0 0 4 8
France 0 0 1 0 0 0 1 I 0 0 3 4
Germany 0 0 2 0 0 2 4 4 4 0 4 12
(Weimar)
Guatemala 2 0 0 0 0 2 4 4 4 4 0 12
Haiti 0 0 0 0 0 0 0 0 0 0 0 0
Honduras 2 0 0 0 0 0 2 4 4 4 0 12
Iceland 0 0 I 0 0 2 3 4 0 0 4 8
Ireland 0 0 0 0 0 0 0 0 0 0 0 0
Korea
1948 I 0 I 0 0 0 2 4 4 0 0 8
1962 2 0 1 0 3 0 6 4 4 4 0 12
1987 2 0 I 0 3 0 6 1 4 4 0 9
Mexico 2 3 0 0 0 0 5 4 4 4 0 12
Namibia 2 0 0 0 0 0 2 4 4 0 2 10
Nicaragua I 2 0 0 0 0 3 4 4 4 0 12
Nigeria 2 0 0 0 0 0 2 3 4 4 0 11
Panama 2 3 0 0 0 0 5 4 4 4 0 12
Paraguay 2 2 2 0 0 0 6 4 4 4 4 16
Peru 0 0 0 0 0 0 0 4 4 0 I 9
Philippines 2 3 0 0 0 0 5 3 4 4 0 11
Portugal
1976 4 0 0 0 0 0 4 I 4 0 4 9
1982 1.5 0 0 0 0 0 1 .5 1 2 0 3 6
(for. pol.) 2
(other) I
Romania 0 0 0 0 0 2 2 1 0 0 1 2
Sri Lanka 0 0 0 0 0 4 4 4 4 0 4 12
United States 2 0 0 0 0 0 2 3 4 4 0 11
Uruguay 1 1 0 2 2 0 6 4 4 2 I 11
Venezuela 0 0 0 0 0 0 0 4 4 4 0 12
156 Presidents and assemblies
1 5 �--------------------------�

0 Chile 69

I
10
III II

.. □ Chile pre-25
D Brazil 88

.. � ��!:!�ia pre-91

0 Brazi1 46

��=·
Ecuador Uruguay Paraguay
□ □ Korea 87
0 0 Korea62
Philippines Chile 89
D Colombia 91
5 □ [fu □
Portugal 76
r.'cl Germany (Weimar),

--- ---- □ 'C!J Guatemala, Sri Lanka

-Ice;:d -
I □ !cu:: C?J EISa�do;:- - -
IV Nigeria, Nicaragua
I Korea48 lJ-� .,..c:i Argentina
Argentina
E] Romania □ I
D
Narnib� I 'O l.:T
Dominican Republic
Portugal 82 E]
(proposal)
VI
Bulgaria, France [] D Costa Rica
I V
Haiti,
Finland r
Ireland Austria I " Pe u I ,..., Veneruela
0 c;i --,---r� □
- �--,--.--------.----1
0 5 10 15

Nonlegislative power

Figure 8. 1. Powers of popularly elected presidents ( democratic regimes that have

broken down indicated by circles)

ful presidents. Presidencies in Region I have great powers in both dimen­

sions. Region II comprises presidents with great legislative powers, but
whose powers outside the legislative process are somewhat weaker. Region
III is empty, reflecting the lack of any logical reason to give great legislative
powers to a president whose power over the composition of governments is
even weaker than those in Region II. Region IV, in the lower left of Figure
8.1, contains our weakest presidents in both dimensions. Almost all of the
premier-presidential regimes are located here. Region V also includes
premier-presidential systems, as well as other systems with moderately
powerful presidents. Finally, Region V I includes only presidential systems
with relatively weak legislative powers but great powers over government
formation.
If we consider the performance of the regimes in Figure 8.1, we find
reason to believe that the more powerful presidencies are also the more
problematic. First let us remove from consideration those cases that are
 Assessing the powers of the presidency 157

Table 8.3. Frequency of democratic failure, grouped by regions in Figure 8.1

Region of Number of Number of Percentage

Fi�ure 8. 1 cases breakdowns of breakdowns
I 12 6 50
II 3 0 0
III 0 0 0
N 4 0 0
V 4 1 25
VI 9 3 25
Note: Nondemocratic systems and constitutions under which a full electoral cycle had not
been completed as of 1990 are excluded.

clearly nondemocratic, or have been nondemocratic for most of their exis­

tence. This criterion eliminates Mexico and Paraguay, which happen to be
among the Region I very powerful presidencies, lending support to our
suggestion in Chapter 3 that there might be an affinity between very strong
presidencies and authoritarianism. By the criterion of nondemocracy, we
should also eliminate Sri Lanka, since elections were suspended for a long
period during emergency rule and civil strife. Other cases to be set aside in
order to investigate the link between presidential powers and democratic
performance are those too new to evaluate: Brazil 1988, Bulgaria, Colom­
bia 1991, Haiti, Korea 1987, Namibia, Panama, and Romania, as well as
the mere proposal made for Argentina. This leaves us with thirty-two
cases, which are grouped by region in the manner shown in Table 8.3.
Although the sample is too small to make any claims of statistical signifi­
cance, six of the ten breakdowns have occurred in the region we have
suspected would be most problematic: very strong presidencies (Region I).
Fully half of these cases have suffered breakdowns at some point. We thus
have reason to be concerned (absent constitutional revision) about the
future viability of democracy in three countries. In the 1980s, Brazil and
the Philippines have adopted constitutions with powers at least as strong as
in the earlier, failed constitutions. Chile has adopted a new constitution
that, while weakening legislative powers compared to the earlier period,
expands nonlegislative powers.
Regions II and V of Figure 8.1, which contain the "confused" cases of
shared presidential-parliamentary authority over cabinets, also include sev­
eral cases of dubious democratic performance, although only one break­
down. (Two more such regimes, the Weimar Republic and Sri Lanka, are
found in Region I.) Besides Peru, where a breakdown occurred, the Chil­
ean parliamentary republic and Portugal 1976 both engendered discontent
that resulted in significant constitutional revision, strengthening overall
presidential powers in Chile and weakening them in Portugal. Finally,
158 Presidents and assemblies
among this group is Ecuador, arguably the Latin American case with the
most aggravated presidential-congressional relations in the 1980s (Conag­
han 1989).
We are left with two regions. One, Region IV contains three premier­
presidential systems and no breakdowns. 2 The other, Region V I, contains
the presidential systems with relatively weak presidential legislative pow­
ers. Of the nine cases, two (Argentina and Nigeria) have suffered break­
downs at some time. Cuba's ambiguous system (see Chapter 6) is also
found here. We also observe that the longest-lived presidential systems -
all scoring 11 or 12 on the dimension of nonlegislative power - may be
found at the low end of the scale of presidential legislative powers. These
cases include:
Costa Rica United States
Dominican Republic Venezuela
These are the four of the five longest-lived presidential democracies in the
world (the other is Colombia, where presidential powers recently were
reduced in both dimensions). If there really is a link, then there are reasons
to be optimistic about several fledgling presidential democracies that rank
low on presidential legislative powers. Regimes such as those of Argentina
and El Salvador, for example, have weathered severe crises that have not
been allowed to become clashes between the two elected branches of gov­
ernment over constitutional powers, in part because the assembly is clearly
the dominant branch.

SEPARATION O F SURVIVAL AND PRESIDENTIAL CAB INET

POWER : DEV E L O PING A COMPREHENSIVE TYPO LOGY O F
REGIMES

We next consider the relation between presidential authority over the cabi­
net and the extent to which the survival of executive and assembly powers
are separate, themes discussed in Chapter 6. Legislative powers do not
enter into the consideration at all. We are concerned here with the nature
of the regime and the power of the president over its officials. As noted in
Chapter 2, these are the dimensions on which the various types of regime
diverge. Considering these dimensions will allow us to develop a compre­
hensive typology of regimes. The value of this typology is not simply in
providing cells into which empirical regimes may be placed, but, more
fundamentally, in letting us perceive what has not previously been an obvi­
ous point to scholars or regime designers: Any given democratic regime
involves some shading on the degree to which powers rest only in the
2 This lack of breakdowns among the premier-presidential systems provides support for
Lijphart and Rogowski's (1991) suggestion that such regimes may be especially well
equipped to manage political cleavages.
 Assessing the powers of the presidency 159
assembly or in an official elected outside the assembly (a president). We
still have ideal types and definitions, as presented in Chapter 2, but we also
have many intermediate types. This is an important finding because it
makes clear that the basic types (presidential and parliamentary and, for
that matter, premier-presidential) are not based on incompatible principles
but rather on mixtures of elements, such as separate survival of powers or
assembly sovereignty over cabinets, that may be applied in varying degrees
to come up with different regime constellations.
In developing this typology, a regime's score on Presidential Cabinet
Authority is the sum of the scores on Cabinet Formation and Cabinet
Dismissal from Table 8.2. The other dimension takes into consideration
Censure and Dissolution. For the latter score on this dimension, however,
it is necessary to reverse the scoring order from that used in Table 8.2, for
here we are not measuring dissolution as an indicator of presidential power,
but rather as an indicator of the degree to which the assembly's survival is
separated from the president. W here a president has the power to dissolve
the assembly, the latter's survival is not separate, hence the score of such a
regime is lowered on that dimension.
Figure 8.2 shows how the cases fit in these two dimensions. The groups
defined above as being near the two extreme corners, (0,0) and (8,8), thus
encompass those regimes in which the definition of authority and survival
are clearest: premier-presidential and presidential, respectively. At the far
upper right are the "pure" presidential systems, with maximum separation
of survival and maximum presidential control over the cabinet. As we
move away from this corner but remain close to it, a regime still exhibits
the characteristics of presidentialism, as with the systems that are slightly
lower on the dimension of authority over the cabinet: Nigeria, the Philip­
pines, and the United States. Moving to the left we find several regimes
that begin to deviate from the basic principles of separation of powers that
define presidentialism. These regimes include Chile 1989, Cuba 1940, and
Uruguay, which, while not "pure" types, resemble presidential systems
more than any other type that we have identified.
At the lower left are the premier-presidential systems. In the far lower
left corner (at 0,0), where we have no empirical cases, would be a system in
which the premier, whose authority stemmed only from the parliamentary
majority, was constitutionally dominant over the president. Nevertheless,
since there would be no separation of survival in such a system, the presi­
dent could dissolve the assembly. In this hypothetical system, the only way
presidents could hope to have any say in the composition of the cabinet
would be to use their power of dissolution. At times this could work for the
president, with the voters returning a majority to the president's liking; at
other times it might not work. Thus the compatibility of president and
premier (and the cabinet headed by the latter) would depend entirely on
what cabinet could gain the assent of the assembly. This feature of such a
160 Presidents and assemblies

8 -01-
�=•
------, ----8--fl�--O
Germany Peru
□
Nurub1a Onie pre-25
Colo 91, Ecu
Korea 48
I
I
Chile 89
Uruguay

ri: t it:,g � �:�k���t2,

r ie
G
Mex , N,c , Pan , Ven
e.9l
,
/b
�ft"•1
President­ vs
6 parliamentary I

1 Presidential

Korea 87

Directly elected
I "prime minister"
(hypothetical)
,{] - - -

Argeolina
(prnposal)

2
Premier-presidential
Assembly­
I independent
D �r':n�: D Romania

m ary
o -1---�---------IH
: 1:
_B_ ::
_
ul.;_ �
_. �·.. _
�-
· Ir-�-
�-!�-���-�_\\�_iH
)
0 2 4 6 8

Separate survival

C.R. Costa Rica

D.R. Dominican Republic
E.S. EI Salvador

Figure 8.2. Separate survival and presidential cabinet power: a comprehensive

typology of democratic regime types

system justifies placing the title of the executive who depends on parliamen­
tary confidence ahead of that of the popularly elected official: premier­
presidential. As we move away from the lower left, either the president
gains increasing authority over the composition of the cabinet (upward
movement) or the president's powers of dissolution are restricted, at least
to som e degree (rightward m ovement).
There is also a considerable cluster of cases near the upper left of the
figure. Regimes in this area have divided authority over the cabinet as
well as a lack of separate survival of powers. The extreme point is repre­
sented by Weimar Germany and Sri Lanka: no separation of survival plus
presidential authority to appoint members of cabinets (subject to censure)
as well as full authority to dismiss ministers. Because of the greater role of
the president in these regimes than in premier-presidentialism, and the
likelihood that the premier (if any) would be unable to serve as a viable
 Assessing the powers of the presidency 161
head of government, we have suggested that these regimes be identified
as "president-parliamentary," reflecting their lack of clarity on the ques­
tion of who wields authority over the government. Some cases, such as
the new constitution of Colombia and the Chilean parliamentary republic,
are intermediate cases between the presidential and president-parlia­
mentary types.
The group of regimes we have identified as president-parliamentary can
be regarded as troubled, or even confused, as each has suffered from
regime crises and each has broken down at some time or been re-formed to
make it conform to one of the pure types, either presidentialism or
premier-presidentialism. For instance, Chile's parliamentary republic was
transformed into a presidential system after a pair of presidential coups in
1925, and Portugal was transformed toward the premier-presidential type
by an assembly-sponsored constitutional amendment in 1982. W hile this
group consists of several cases and thus constitutes an identifiable regime
type that has proven attractive to some regime designers, most recently to
delegates to Colombia's and Namibia's3 constituent assemblies, we suggest
that it is a type best avoided.
Now let us consider the lower right, for the final "pure" type that is
suggested by the two dimensions of Figure 8.2. Literally understood, such a
regime would be one in which there was maximum separation of survival of
powers - neither censure nor dissolution - but no presidential role in the
composition of the cabinet. Such a regime has never existed, to our knowl­
edge, with a popularly elected president. However, the scenario just de­
scribed comes close to fitting a regime such as that of Switzerland, in which
the entire cabinet is chosen by the assembly but is not responsible thereto.
Bolivia would be similarly situated (unless a presidential candidate's slate
were to win a majority of the popular vote), even though the assembly
formally elects only the president, who then chooses the rest of the execu­
tive. The dependence of the president, and therefore ultimately the whole
executive, on selection by the assembly - but for a fixed term - implies
assembly primacy in the origin of the executive to an extent that nearly
approximates that of Switzerland. Therefore these regimes could be placed
in this cell and called assembly-independent regimes, in order to identify
both what institution has control over the origin of executive power, and
the separation of survival. 4
3 As we have discussed, Namibia's constitution contains one provision that may make for
somewhat less contentious presidential-assembly relations than in cases like Weimar or Sri
Lanka: dissolution in Namibia requires a new presidential election.
4 Because the terms of comparison literally require a popularly elected president, and there is
much room in the lower right of Figure 8.2 for intermediate cases, is there any basis to
imagine a hypothetical regime in which a president would have some (limited) powers over
the composition of the government but there would be separation between the survival of
assembly and executive powers? Imagine a popularly elected president who could propose a
cabinet to the assembly, which would then have to ratify (or reject) the proposal. Once
162 Presidents and assemblies
A few regimes appear as borderline between the main categories. We have
already referred to those, such as the new constitution of Colombia, that
are basically president-parliamentary but tending more toward presiden­
tialism because of the lack of a power of dissolution. Finland and Iceland
appear to be intermediate regimes between president-parliamentary and
premier-presidential types. W hile this placement is somewhat surprising, it
results from the provision in both constitutions that the president alone
appoints all ministers. 5 This feature, combined with the provision that only
the assembly may remove ministers, gives the president at least marginally
greater influence over cabinet composition, as we argued in Chapter 6. As
a result, greater instability of cabinets is a possibility, although, as Finland
demonstrates, this need not be regime-threatening, as is the combination of
features that characterizes the Weimar Republic or Sri Lanka. If Finland
and Iceland represent cases that deviate from an ideal type of premier­
presidentialism by granting the president marginally greater power over
cabinets, Korea's 1987 constitution represents a deviation from presiden­
tialism that grants the president marginally less power over cabinets. It
does so by providing that only the premier is directly appointed by the
president, subject to confirmation by the assembly. The premier then ap­
points all other ministers. All ministers, including the premier, may be
removed by the president but are not subject to censure, thus accounting
for this regime's placement on the far right-hand axis. As discussed in
Chapter 6, this design makes the president's ability to get his or her pre­
ferred cabinet dependent entirely on how good an agent the premier
proves to be, but the president retains the key powers of initiative in the
process.

Parliamentary regimes. Another small cluster of intermediate cases in Figure

8.2 turns out to be identical to one of the most basic regimes types: parlia­
mentarism. Literally, as with assembly-independent regimes, parliamentary
regimes are off this figure unless there also happens to be a popularly elected
president. However, in the same sense that assembly-independent regimes
can be accommodated conceptually within the dimensions of whether or not
authority over the cabinet rests within a separately elected constitutional
power, so can parliamentary regimes. Such regimes feature assembly cen­
sure of the executive, but no popularly elected president with authority
either over the government or to dissolve parliament. Thus, parliamentary
regimes have a score of 4 on separate survival and O on presidential authority
ratified, however, the government would not be subject to parliamentary confidence, nor
could the president dismiss the ministers. (Of course, if the president could dissolve the
assembly, the regime would be located to the left side of the lower-right cell.)
5 Recall that our definition of premier-presidentialism requires only that the president have
some powers, but that those powers not include unrestricted authority to dismiss ministers
or cabinets that enjoy parliamentary confidence.
 Assessing the powers of the presidency 163
over the cabinet. Such cases in Figure 8.2 are Bulgaria, Haiti, and Ireland. In
none of these constitutions does the president have any constitutional author­
ity, rendering these cases effectively parliamentary. If we ignore the question
of whether or not there is a popularly elected president in a parliamentary
system on the grounds that such a president would be utterly feckless any­
way, then all parliamentary systems are located at the point (4,0) on our
graph, along with Bulgaria, Haiti, 6 and Ireland.
Another intermediate type of regime is represented by the case of Leba­
non, which was discussed briefly in Chapter 5. Lebanon and similar re­
gimes that have existed at one time or another in Indonesia, Kenya (before
1969), Turkey, and elsewhere, have an executive chosen entirely by the
assembly, but part of that executive depends on parliamentary confidence
while another part of it serves a fixed term that is, under some constitu­
tions, longer than that of parliament. If the fixed-term part of the executive
is more than just a ceremonial head of state, the regime's score on the
horizontal axis of Figure 8.2 is somewhat greater than 4 but less than 8. (It
still scores O on the vertical axis, given the lack of a popularly elected
president with authority over the cabinet.) Some of these assembly­
selected presidents have initiative in cabinet formation, the power to dis­
solve the assembly, a veto on legislation, or some combination of these and
other powers. As with assembly-independent regimes, the "president" in
this regime format remains an agent of the parliament, but a more indepen­
dent one than a purely parliamentary executive, given the lack of a vote of
no-confidence. If the agent is eligible for reappointment, there should be
little slack between the wishes of the parliamentary majority and the ac­
tions of the agent, since the agent can anticipate the possibility of being
denied reappointment. In some of these regimes, however, the president
cannot be reappointed and is thus potentially truly independent. Still, given
the mechanism of parliamentary confidence binding a part (usually the
more important part) of the executive to the assembly on an ongoing basis,
these regimes, while not strictly parliamentary, are closer to that type than
to any other.

A possible alternative regime: the directly elected "prime minister. " A ques­
tion that might arise is whether or not there is any conceivable regime
6 One interpretation of the military coup that ousted Haitian president Jean Bertrand
Aristide in September 1991 is that the president, while enormously popular, had failed to
win a majority in either house of the parliament and therefore was powerless. He proceeded
to use his popularity to incite mob action against the parliament, compelling it to accept a
pro-Aristide prime minister that the president had no constitutional authority to impose.
The military subsequently removed the president on the not entirely unfounded grounds
that he was acting unconstitutionally. The Haitian experience suggests that a president
should not have zero constitutional powers, even at the same time that several other cases
that we have studied have led us to the conclusion that presidential powers should be quite
limited in the constitution.
164 Presidents and assemblies
design that would be placed at or near the center of Figure 8.2, where we
have no empirical cases. W hile not the only possible configuration that
would place a regime at the center of the figure, there is an important
alternative regime type that has been proposed by reformers in countries
such as Israel (Libai et al. 1990), Italy (Barbera 1990), and the Netherlands
(Lijphart 1984). Such proposals are typically referred to as featuring the
direct election of the "prime minister," and Barbera calls his proposal
"neo-parliamentary"; however, the elected head of government would be a
president by our definition. The proposals consist of the concurrent elec­
tion of the head of government (whom we �hall call a prime minister, in
keeping with the terminology of the propcsals) and the assembly. The
prime minister would be permitted to dissolve the assembly, but then she or
he would have to stand concurrently for reelection. Similarly, the assembly
could censure the prime minister or the entire government, but then it
would have to stand for reelection along with the prime minister. Thus, on
separation of survival, we would score a 4, that is, 2 on each aspect, as the
powers of censure and dissolution are both restricted by requiring the new
election of the power that initiated the process. The placement on the
vertical dimension would depend upon the specific process by which mem­
bers of the government other than its elected head would be appointed and
dismissed. The Israeli proposal would require the elected prime minister to
present her or his government to the assembly for investiture, thus suggest­
ing a score of 3 on (President's) Cabinet Authority. The proposal would
also allow the prime minister to dismiss ministers at will, leading to a
combined score on this dimension of 7. Because of the possibility that,
especially in a multiparty context, the prime minister and the assembly
majority might not always be entirely compatible, it might be advisable to
weaken the prime minister's power of dismissal. Otherwise some of the
same kinds of conflicts that befall the president-parliamentary regimes
could result, although the means of breaking deadlock would be more
symmetrical, given the assembly's ability to oust the "president." Some
restrictions - for example, by number of dismissals or how long a minister
who enjoys parliamentary confidence must be allowed to serve before
being dismissed by the prime minister - would reduce the overall score to a
5. If the prime minister's power of dismissal were made "constructive"
(akin to the constructive vote of no-confidence) by requiring that the prime
minister present for the assembly's approval an alternative candidate for
the minister that she or he wants to replace, then the regime would score 4
on both dimensions, with neither elected body being more powerful than
the other. We would thus have the "ultimate" compromise between the
principle of parliamentary sovereignty over the composition of cabinets
and that of placing governmental authority in the hands of an extra­
parliamentary offical. Even without the suggested restrictions on the power
of the directly elected prime minister, such a regime is indeed a compro-
 Assessing the powers of the presidency 165
mise regime. Thus it is not surprising that it would be developed as an
alternative in some countries where there is dissatisfaction with existing
parliamentarism. Perhaps what should be surprising is that there are no
empirical cases, as of 1991.

C ONCLUSI O NS : PRESIDENTIAL PO WERS

Our examination of the dimensions of presidential power suggests the re­

gimes with great presidential legislative powers are problematic, as are those
in which authority over cabinets is shared between assembly and president.
These issues need to be addressed in turn. On matters of legislation, we
suggest that relatively strong assemblies should be associated with more
stable and effective government relative to strong presidencies because as­
semblies serve as arenas for the perpetual fine-tuning of conflicts. An assem­
bly represents the diversity of a polity far better than an executive dependent
on the president's whims is likely to do. Because of the diverse forces repre­
sented in an assembly, such a body has the potential for encompassing diver­
gent viewpoints and striking compromises on them. The dual democratic
legitimacies decried by critics of presidentialism - the claim that no demo­
cratic principle exists to resolve conflicts over who better can claim to repre­
sent the "will" of the electorate - are minimized to the extent that an assem­
bly is accorded a more powerful role in legislation than is the president. Thus
presidentialism with a strong congress indeed does afford a democratic prin­
ciple for the regulation of interbranch conflicts; that principle is that the
assembly prevails, subject to a need for compromise with the president. The
relatively weaker presidents (those of Costa Rica and the United States, for
example) cannot use decree legislative authority to break a "logjam" in
congress, as many presidents can do with perfect legality. Thus, a fundamen­
tal conclusion is that the criticisms of presidential regimes should not be put
forward as if all presidencies were created equal; rather, these criticisms
apply with greatest force to strong presidents.
On the matter of authority over cabinets, again the shared control that
we said typifies the president-parliamentary type goes right to the heart of
the concern of many with dual democratic legitimacies. If there is no
"democratic principle" that defines who fills cabinet posts, one of the most
basic elements to any democracy, then conflicts of a very basic nature are
likely. Are the ministers the president's ministers, or are they the assem­
bly's? In some regimes the answer is both. In either a premier-presidential
regime or a presidential regime, on the other hand, the primacy of one
branch over the other is clear. Both types make for a cabinet subject
(whether exclusively or primarily) to one branch or the other even when
the branches are controlled by different political tendencies. This does not
prevent conflict, but it is not as clearly guaranteed to generate conflict as is
the president-parliamentary type. Democratic institutions are supposed to
166 Presidents and assemblies
be conflict regulators, not conflict generators. Either giving great legisla­
tive powers to the most majoritarian component (the presidency) of a
regime meant to be consensual, or granting shared authority over the com­
position of the cabinet, is a potentially dangerous arrangement. Finally, we
saw in this chapter that actual regimes, as well as a promising proposal for a
directly elected "prime minister," entail varying combinations on two
dimensions - separation of survival of executive and assembly, and presi­
dential power over cabinets. In the next chapters we shall see how presiden­
tial powers interact with the means of electing the assembly and president.
 9
Elector al dyn amics: efficiency and inefficiency

In this and the following three chapters we discuss institutional variations

related specifically to elections. The way in which the checks and balances of
presidentialism or the relations between president and cabinet in premier­
presidentialism play themselves out depends in part on how likely it is that
the president confronts an assembly that does not reliably conform to the
president's will on legislation. As we shall see, such factors as the method of
electing the president and assembly and the relative timing of elections to the
two branches are crucial factors in affecting the number of competitors. Thus
we must return to the issues with which we started this book and which
constitute one of its major themes: the ways in which the processes of elect­
ing representatives and of executive formation interact. Here we deal with
electoral dynamics, by which we mean the ways in which the practices used
for electing the assembly interact with the form and powers of the executive
to shape the functioning of democratic regimes. This chapter concerns itself
primarily with presidential systems, while the following three deal with both
presidential and premier-presidential regimes.
We begin this chapter by reconsidering the tension between representa­
tion based on parochial interests versus that which articulates national
policy perspectives. We develop an archetype of a presidential system in
which the two forms of representation coexist. In so doing, we are drawn to
a novel conclusion: that regimes that maximize the articulation of local
particularism in congressional elections tend to be associated with very
powerful presidencies. We develop a theoretical consideration of why this
would be so, then contrast these regimes with those that entail weaker
presidents, concluding that the latter regimes are more conducive to the
articulation of national policy alternatives, especially when elections to the
two branches are concurrent. Thus we build in this chapter on the concerns
of the previous one on presidential powers, while introducing new concerns
about the organization of elections for president and assembly.

P A R OC H I A L I S M V E R S U S N A T I O N A L P O L I C Y

There is a certain tension inherent in presidential systems, since the legisla­

tive assembly and the executive have separate origin and survival. On the
168 Presidents and assemblies
one hand, unlike parliamentarism, presidentialism has a directly elected
national executive. This allows, at least in theory, the voters to confer a
mandate upon a particular political force to govern the country for a speci­
fied period of time. At the same time, the assembly can represent diversity
without compromising the survival of the executive. Recall from Chapter 1
that we have identified two basic forms of representation: that based on
local particularism and that based on group interests and national policy
preferences. The two are not mutually exclusive, but rather opposite ends
of a continuum, with real systems exhibiting some mix of these two ideal
types. Recall further that we indicated that there is a trade-off between
representativeness and efficiency, but that the trade-off is theoretically less
acute under presidentialism than under parliamentarism. We shall sketch a
bit more here what we touched on in Chapter 1, the origin of policy-related
voting in Britain. We do so mainly to provide a contrast to how the process
of executive formation affects local representation under presidentialism
and in order to define an "archetype" of presidentialism.

The articulation of national policy in parliamentary systems

How are parochial interests transcended in the process of forming an execu­
tive, thereby permitting national policy to be formed and approved despite
the narrowness of regional representatives' concerns? In a parliamentary
system, the primary mechanism is the requirement that some group of those
narrow-minded district representatives must come together and sustain a
government in office. Thus developed party organization in the parliament.
Although the degree to which the process is completed varies empirically
across cases, the theoretical process of transcending district-level parochial­
ism implies the following, drawn from Gary Cox's the Efficient Secret, in
which the emergence of parties in the British case is examined.
Once the practice of the government's dependence on the confidence of
the parliamentary majority became established in Britain, backbench mem­
bers saw the progressive erosion of their powers of legislative initiative.
The cabinet leadership more and more monopolized the agenda. As this
process continued, members of parliament lost most of their ability to curry
voter support in their districts by means of attention to purely local con­
cerns and the provision of specific benefits.
If members' time for attending to parochial concerns is curtailed - and if
pursuing such concerns might imperil one's own government - then candi­
dates in the districts must differentiate themselves on the basis of what the
policies are or would be of a government formed by the party to which they
belong rather than on servicing their constituencies. That is, rather than
seek to promise to deliver goods, candidates begin to identify with a pack­
age of national policies (presumably always being careful to explain what
benefits said policies will provide to one's district). The process just de-
 Efficiency and inefficiency 169
scribed, then, is one in which parliamentarians, rather than continuing to
be narrow-minded so as to reflect only the needs of their parochial districts,
begin to act collectively. This collective action is manifested in the develop­
ment of more programmatic, nationally oriented parties and a policy­
oriented electorate (Cox 1987). W ith voters having to choose from among
only one candidate from each of two (or more) parties, rather than from
competing members in each party (as had been the case in many British
districts before the Reform Acts of the mid-nineteenth century), the compe­
tition became categorical: one party/cabinet or the other, rather than com­
petition over the provision of specific benefits to supporters. As this pro­
cess plays out, an "efficient secret" has emerged: The party now has a
"brand name" by which to identify itself to the voters.
The process just discussed has been inhibited in those parliamentary
democracies in which the electoral system affords voters a choice of candi­
dates within parties. The interaction of executive form with the electoral
system is seen very clearly in Japan, where the ruling Liberal Democratic
Party, in order to maximize its majorities, not only needs the party program
that goes along with parliamentary confidence, but also must provide
"pork" to its rank and file to cultivate support groups within the districts.
In this way, the LOP - like similar parties elsewhere - is able to ensure the
loyalty of core groups of voters to particular members of the Diet without
having a situation in which intraparty competition dilutes the overall repre­
sentation of the party, as would happen if members of the party competed
with one another on the basis of policy programs. The result, relative to the
United Kingdom, is inefficiency: The success of LOP members in securing
a large portion of the total vote on non-policy grounds limits the potential
for voters to designate alternative policy programs, a defining criterion for
electoral efficiency. Moreover, the electoral system also encourages ineffi­
ciencies in budgetary allocations. For example, according to McCubbins
and Rosenbluth (1992) , around 30 percent of the Japanese budget goes to
particularistic allocations (public works, subsidies, etc.), while a compara­
ble figure for either the United Kingdom or the United States would be
about half that. Such "inefficiency" may be expected, ceteris paribus, in
any electoral system with intraparty competition,1 where such means of
cultivating votes become relatively more important compared to voting on
broader policy options than in systems that do not afford intraparty choice.
If we are correct that parliamentarism without intraparty preference
voting (as in the United Kingdom) encourages national outlook on the part
of local representatives, then we expect presidentialism to create a very
1 We refer here to any electoral system in which the criterion determining the order in which
candidates get elected is how many personal votes they have amassed. Other electoral
systems, including those of Belgium and the Netherlands, allow voters to express votes for
candidates within lists, but the party leadership's rank order of candidates on the list is the
dominant criterion in determining election.
170 Presidents and assemblies
different electoral dynamic. This should be so especially with a congres­
sional electoral system requiring intraparty competition. Members of con­
gress need not submerge their district interests beneath a national focus in
order to maintain an executive in office. Also, control over the agenda of
the assembly is not the imperative for the executive that it is in a parliamen­
tary system. To the extent that the assembly itself takes on an important
policymaking role, the necessity of acting collectively will pull our paro­
chial representatives in the general direction of national focus. However,
even if this process occurs in a presidential system's assembly - and it may
not - there is no theoretical reason to believe that it must take place in
some sort of harmony with the national policy being debated and chosen in
the presidential election. Moreover, where the electoral system favors the
distribution of "pork," the absence of collective responsibility for the gov­
ernment may (theoretically) render the process of budgetary allocation
even more inefficient than in parliamentary Japan. 2

The articulation of national policy in presidential systems

Having reviewed the process by which local interests are transcended
within parties in parliamentary systems and institutional arrangements that
favor particularistic allocations, we turn more directly to the question of
presidential systems. It is self-evident that the election of the president
favors the articulation of national concerns, at least if there either is no
regionally elected electoral college or if campaigns for electors to such an
electoral college are conducted as if they were campaigns for president
directly. Indeed, that the process of choosing electors in two presidential
systems with electoral colleges - Argentina and the United States - should
have developed rather early into a nationwide campaign for the single
office of the presidency is testimony to the nationalizing tendencies of
elections for a national executive. As Cox (1987:143) suggests, it was a
mistake of the U.S. Founding Fathers "to believe that voters would entrust
a decision to others which they could make themselves." If the office to be
elected is constitutionally disposed to be the representative of the "whole
nation," then candidates and parties have strong incentives to conduct
2 While we should expect a theoretical and empirical link between electoral and budgetary
inefficiency, the case of electoral efficiency is less clear. The ability of voters to make a
categorical (efficient) choice does not guarantee budgetary efficiency. It is conceivable that
electoral efficiency might simply consist of voters choosing which group of spoils they
prefer, as indeed was the case for the pre-civil service United States. The difference is that,
without intraparty competition, spoils are under leadership control to a greater extent than
in systems with intraparty competition. Reform away from inefficient economic and bureau­
cratic policies should be more difficult when the electoral system itself is inefficient. in the
sense used here. See Geddes (1991) for a similar argument. This relation between relative
electoral efficiency and efficiency (in the economic sense) of policy outcomes is a highly
promising area for future research.
 Efficiency and inefficiency 171
campaigns for that office on the basis of national concerns. This observa­
tion applies when there is an electoral college as much as when there is a
direct popular vote, as candidates to be delegates to the electoral college
have only one thing to offer voters: their voting intentions for president.
Thus we find, for the election of the national executive, the emergence
of electoral efficiency in the form of the articulation of policy options. Cox
applies the term to elections for members of parliament in the British case,
quoting Bagehot's very same analogy of the House of Commons to the
U.S. electoral college: "The main function of the House of Commons is [as]
an electoral chamber; it is the assembly which chooses our President"
(quoted in Cox 1987:143). Efficiency in voters' ability to designate a policy
direction in the national executive, then, stems from their voting for either
an executive directly or an assembly that functions mainly as a means of
translating voter preferences into the formation of an executive. This ele­
ment of efficiency we should expect to exist in a typical presidential system.

Preservation of local representation in presidential systems

Where presidential systems may be likely to fail the test of efficiency is in
that other branch that is also elected. The assembly need not be chosen
with an eye to the national executive or national policy direction preferred
by the plurality of voters, especially when the electoral rules favor the
cultivation by candidates of particularistic support groups. The separation
of powers does not by itself guarantee linkage between the executive and
assembly either in elections or in the functioning of these institutions be­
tween elections. In particular, when elections for president and for con­
gress occur at separate times, congressional campaigns are more likely to
focus primarily upon local issues, or, to take the other form of representa­
tion discussed in Chapter 1, on narrow ideological or partisan interests.
While a calendar of elections that entails separation of campaigns is
especially likely to reduce the salience of national issues in congressional
races, the process that retards nationalization of congressional campaigns is
inherent in the institutional basis of presidentialism. The separation of
powers that characterizes a presidential system means that candidates for
congressional seats need not tie their fortunes to a national policy position
in order to be elected. W here this process inhibiting nationalization of
congressional campaigns has been transcended, we should expect most
often to find closed party-list systems, which enhance the value of the party
label in elections by not requiring members of a party to compete against
one another for personal votes and by giving party leadership the ability to
reward its most loyal rank-and-file members by reserving for them the list
ranks most certain to secure election.
The inherent separation of campaigns has been a source of frustration for
advocates of reform in the United States who have decried the localism that
172 Presidents and assemblies
so dominates congressional campaigns and the "errand boy" functions per­
formed by members of Congress (e.g., Stedman 1968; Robinson 1985;
Sundquist 1986). Many of these would-be reformers, whose arguments are
predicated on their own presumption that national issues should dominate
congressional as well as presidential elections, advocate abolishing mid­
term elections. They reason, with good cause, that the salience of national
issues is diminished in midterm elections by the lack of a national office
being elected simultaneously. For many of these critics, however, even
eliminating midterm elections would be insufficient to accomplish their
stated goals, and they advocate other reforms to make parties more "re­
sponsible" to a national platform (see Ranney 1962 for a review). Some
even go so far as to recommend doing away with separate elections alto­
gether, by requiring a "team ticket," as in Bolivia and Uruguay and some
elections in the Dominican Republic, such that voters could not split their
votes between presidential and congressional candidates of opposing par­
ties (see Robinson 1985:177-79). 3 These frustrations over the different
constituencies of the Congress and executive in the U.S. presidential sys­
tem stem from the inherent tension of presidentialism between national
and particularistic representation.
The tension in presidentialism, stemming from the separation of powers
and separation of campaigns, may be a source of frustration for reformers
who value the national focus that presidential elections can give to the
polity and disdain the "obstructionist" role of the particularistic congress.
Yet these same frustrations also point to a key advantage of the separation
of powers and functions. If one values or at least recognizes a conditional
utility to an electoral system that is designed to facilitate the expression of
the interests of localities in the legislative process, presidentialism is more
likely to serve the goal than is parliamentarism.
In a parliamentary system, one must choose between a national policy
focus in elections and parties or else localism. Either one gets an efficient
secret, in which politics is focused on party-centered national policy goals,
but some of the benefits of locally oriented politics are lost. Or one main­
tains the advantages of politics based on provision of local services, but
parties are weak, lacking national focus. Then governing becomes highly
problematic, as with the famous trasformismo in Italy or the French Third
and Fourth Republics, as well as several failed parliamentary systems in
less developed countries (e.g., Burma and Nepal in the 1960s).
W here parliamentary systems have become fully developed, the existence
of the institution of the vote of no-confidence has mandated an imposition of
discipline. This party discipline in an archetypal parliamentary system (such
3 Curiously, few advocates of reform have argued in favor of a party-list electoral system,
which would serve the reformers' desired end of more nationally focused parties without
touching the basic separation of powers. Duverger (1984) and Amy (forthcoming) are
exceptions to the general neglect of list PR as an alternative for the United States.
 Efficiency and inefficiency 173
as the United Kingdom, New Zealand, or Norway, for example) is sure to be
far stronger than in a typical presidential system's assembly, barring other
mechanisms, such as a closed party list, by which party leaders exert disci­
pline over their rank and file. Even if the assembly of a presidential system
takes on an important policymaking function, as in the United States, the
parties can afford more deviations from the party line than can parliamen­
tary parties. If the party that has a majority in the congress also holds the
presidency, there is no danger of losing the executive if individual members
are allowed to continue to pursue their own districts' interests. If the party
with the congressional majority does not hold the presidency, it cannot hope
to displace the executive, but it can continue to offer opposition to the
existing executive. Either way, it can afford more slack within the ranks than
can a parliamentary party. Indeed, this affording of greater intraparty slack
and articulation of local interests is a potential strength of presidential sys­
tems, even if it does cause frustrations for those seeking to carry out reform­
ist policies on a national scale. It can allow for more accurate representation
of constituent interests in the assembly (see Powell 1989).

A N A RC H E T Y P A L P R E S I D E N T I A L S Y S T E M

We shall now propose a n ideal type of system based on a separation of

powers. The system in question is archetypal in that it would allow, to the
greatest extent feasible, both branches to do what they were defined in
Chapter 1 as doing best. We must caution that by "archetypal" we are not
proposing the most desirable system. Indeed, we shall argue that the arche­
type is not the most desirable form of presidentialism; moreover, its com­
monness as a form of presidentialism - and a frequently failed form - has
given presidentialism as a regime type a "bad name." Our purpose is more
limited than endorsing a particular regime type, however: We are merely
articulating a conceptual construct.
The assembly in such a system would be highly representative, but the
executive would be based upon a broad set of national goals endorsed by a
majority (or nearly so) of the voting population, hence electorally efficient.
To maximize these ends, presidential and congressional elections would
always be conducted at different times and therefore on their own terms. If
the basis of particularism in the assembly were localism, parties in congres­
sional campaigns would not even have to operate on a national scale. Their
purpose, after all, would be primarily to provide vehicles for the allocation
of patronage or to give structure to local issues in locally run campaigns for
congressional seats. 4 Come time for a presidential election, however, vari-
4 All of these seats, in order to meet the criterion of local representation would be allocated
in districts. This should not, however, be taken to imply necessarily single-seat districts;
indeed, as we have suggested, representation of local interests is maximized by systems that
afford intraparty competition, meaning (small) multi-seat districts.
174 Presidents and assemblies
ous parties would coalesce to present two major candidacies for the presi­
dency, offering voters a choice (an efficient choice indeed) for the direction
of national policy over the next term.
The sketch just provided of an archetypal presidential system closely
resembles Chile under the constitution of 1925. There are also strong simi­
larities to the Brazilian constitution of 1946 and to Colombia from 1886
until 1991 (with the possible exception of the National Front period). How­
ever, as we shall see farther into this chapter, electoral reforms in Chile
beginning in 1958 and, especially, a constitutional reform in 1970 changed
that system considerably, moving it out of the category of the archetype.
During the earlier period, congressional particularism was indeed high and
based on local project- and patronage-oriented concerns. As a result of
electoral and constitutional reforms intended to increase the efficiency of
the system, the congress was transformed toward the other type of assem­
bly discussed in Chapter 1: a body representing ideological and partisan
differences that transcended many of the previously more prominent local
matters. Simultaneously, as we saw in Chapter 8, the powers of the presi­
dency were further increased. Once these transformations occurred, the
system became so polarized that a constitutional crisis eventually led to a
military dictatorship. Thus, Chile demonstrates both how a regime based
on rather inefficient congressional elections can function as a stable democ­
racy and how not to reform a system if one hopes to make it at once
efficient and stable. To foreshadow a conclusion of this chapter's study, let
us indicate that the more promising means of increasing efficiency rest in
enhancing the strength of party leadership while simultaneously decreasing
the strength of the presidency.

P A R T Y S T R E N G T H A N D P R E S I DE N T I A L S T R E N GTH

We shall now look more closely at the connection between form of

representation - as determined by the electoral system - and strength of
presidents. The Chilean experience is discussed here at some length be­
cause it suggests that an archetype of presidentialism - extreme separation
of congressional and executive constituencies - is most viable when the
congress is highly local in its own constitution. However, Chile and other
cases suggest that this kind of congress also generates pressures to delegate
great constitutional powers to the presidency. This process may plant in the
regime the seeds of its own destruction, especially once national "reform­
ist" policy options come to be articulated. And recall that, owing to the
nature of presidential elections, the articulation of such options is easier at
that level, enhancing the prospects of interbranch conflict. Articulation of
reformist options also becomes easier in congressional elections to the
extent that party organizations are strengthened, thus decreasing the sa­
lience of particularistic campaigning. Should the assembly itself then come
 Efficiency and inefficiency 175
to be divided by ideological cleavages, presidentialism may be poorly
equipped to handle the stress, especially if the president is accorded signifi­
cant legislative powers. For now, let us observe that several cases of
multiparty presidentialism, including Chile and Brazil,5 as well as other
cases, such as Colombia, the Philippines, and Uruguay, in which local
particularism prevails, have also been among the systems with the most
powerful presidents. Perhaps, therefore, these systems were designed pre­
cisely with the idea that a president with strong legislative powers was
necessary to provide some efficiency and continuity to policy output. We
shall develop this point further below.
We, along with Mainwaring (1992a), have wondered why the fact that
Chile remained a stable multiparty presidential system for so long has gone
unnoticed in the comparative literature. Still more remarkable is that Chile
so closely resembled the archetype of a presidential system in which the
electoral constituency of the congress is predominantly local and parochial
while the executive is electorally connected to broader national coalitions.
Before proceeding with a more detailed examination of the Chilean case, let
us consider the relationship more generally between presidential strength
and fractionalized party systems, including multiparty systems such as those
of Brazil and Chile, as well as others that provide considerable autonomy to
local rank and file.
We propose to develop a rough indicator of the strength of parties accord­
ing to the constitutional and electoral rules that influence party leadership's
control over local activists and elected officials. We shall consider five as­
pects of party strength. The first is control over nominations. Does the
national party leadership determine who may run under the party name, or
are there laws and constitutional provisions that decentralize the decisions
on nominations? The second aspect is whether or not the party leadership
controls the order in which candidates are elected by means of a list-rank
order. The third aspect is whether votes for one candidate ( or factional list
within the party) help increase the votes and therefore seats won by the party
as a whole by means of some procedure for pooling votes. Fourth, we shall
ask whether there is internal party competition simultaneous with interparty
competition. Fifth, how great are the barriers imposed by the district magni­
tude to forming new parties (or new intraparty lists)? As shown in Table 9.1,
we score each of several countries as either lacking an aspect of party leader­
ship strength (0), possessing it ( + ), or possessing it to a great degree ( + + ).
Each " + " will count as a point, leading to a maximum of 10 points for any
party system. The lower the total score, the more decentralized the party
system is and therefore the more the parties are under the control of local
rank-and-file candidates and campaign workers, or caciques.
S We could add the case of Ecuador but, as we have seen, it is not a true presidential regime,
given the provision for censure. Still, the president has significant powers and much of the
logic of the present argument should apply.
176 Presidents and assemblies

Table 9 . 1 . Strength of party leadership over rank and file in presidential democracies

Control Control order Pool of Lack internal Entty

Coun!!J: nominations? of election? votes? COmJ!!:tilion? barriers Score
Argentina + ++ ++ ++ 0 7
Brazil 0 0 + 0 0 I
Chile
(pre-1958) 0 0 + 0 + 2
(post-1958) ++ 0 + 0 + 4
Colombia 0 0 0 0 + I
CostaRica ++ ++ ++ ++ + 9
Dominican ++ ++ ++ ++ + 9
Republic
Ecuador ++ ++ ++ ++ 0 8
El Salvador ++ ++ ++ ++ ++ JO
Nicaragua ++ ++ ++ ++ 0 8
Philippines 0 0 0 ++ ++ 4
Uruguay + + ++ 0 + 5
United States 0 0 0 ++ ++ 4
Venezuela ++ ++ ++ +-+ 0 8

We consider here primarily those presidential systems with a fairly long

record of democratic governance at some point in the twentieth century, as
well as some others for which data on electoral systems were available, for
a total of fourteen cases. Some of our rankings may require some explana­
tion. The closed-list systems (Argentina, Costa Rica, and Venezuela) all
provide leadership with control over nominations, but in Argentina there is
some evidence (McDonald and Ruhl 1989) that this is weakened by state­
level party organizations within the federal system. In Brazil a law known
as the candidato nato makes it possible for any candidate once elected to a
post by that party to continue to have a place on the party's future lists for
that office regardless of the desires of the party leadership (Mainwaring
1991); thus we score this case as not granting the party control over nomina­
tions. For Chile, the electoral rules before and after a reform in 1958 must
be scored differently. In 1958, a center-left coalition enhanced the meaning
of the party label in the name of efficiency (Valenzuela 1989b), while
before 1958 the rules more closely resembled those of Brazil. Control over
order of election is generally straightforward, being granted by closed lists
but not by open lists (and being irrelevant when M = 1), but is weakened
when there are multiple lists within the party. In such cases, whether there
is any effective institutionally determined control or not depends on
whether the party can control the use of the party name by the factions in
the first place. Colombia's parties lack such control, while Uruguay's have
it. Barriers to entry depend largely on district magnitude, and we have
scored magnitudes of 4 or less as high barriers, magnitudes of 10 or more as
low, and intermediate cases as moderate ( + ).
We now shall compare our indicator of party leadership strength to our
indicator of presidential legislative powers (from Table 8. 1 ). W hile we must
 Efficiency and inefficiency 177

12 -----------<�1--------------� Chile 70

10 -

Brazjl 88

Colombia
8 _ 0 0 g'�_e58) 0 Chile (post-58)
Brazil 46
O Inefficient

6 - O Uruguay A Ecuador

0 Colombia 91 [J Phillipines

4 -
Nicaragua

□ El Salvador

Argentina
2 - A U.S. A □
Dominican
Republic

Efficient D co.sta
R.ca
Venezuela
0 --t-----r---r--
, --..-----.-,--.---�,---r----l · 1--�--1 ';"
0 2 4 6 8 10

Party strength

0 Concurrent elections
0 Non-concurrent elections
A Concurrent and midterm elections

Figure 9.1. Party strength and president's legislative powers

be careful about drawing conclusions from such a small number of cases,

Figure 9.1 suggests a striking relationship. We have seventeen cases - the
fourteen different electoral systems from Table 9 .1, plus three additional
cases created by revised constitutions applied with identical electoral sys­
tems in Brazil, Chile, and Colombia. Most of the cases cluster in one of two
groups: Those with strong presidents and weak parties constitute one
group, while those with weak presidents and strong parties constitute an­
other cluster. The latter group includes the three presidential systems that
would have to be considered among the most successful as democracies:
Costa Rica, the Dominican Republic and Venezuela. In the upper-left
quadrant of Figure 9 .1, we find that all the regimes except Colombia and
the new regime of Brazil have suffered a breakdown, although two, Chile
178 Presidents and assemblies
and Uruguay, were democratic for more than two thirds of the twentieth
century. The graph also shows by different symbols some characteristics of
the electoral cycle each country employs - whether or not congressional
elections run concurrently with presidential elections. The importance of
this variable will become apparent as we develop the arguments of this and
the following chapters.
Most of the cases in Figure 9. 1 fall close to an imaginary axis running
from the upper left to the lower right. Two cases are obvious outliers. Of
the long-term effective democracies, the United States is our clearest
outlier, having relatively weak parties along with a president whose legisla­
tive powers are weak. Ecuador, a highly troubled democracy and not
strictly a presidential system, anyway, lies alone among cases with presi­
dents and parties that are both strong. 6 Colombia's new regime, also not a
genuine presidential system, has weak presidential legislative powers com­
pared to other cases with such weak parties. As we shall see, the axis runs
from the inefficient presidential regimes (upper left) to the efficient ones
(lower right).
Besides observing that strong presidents and weak parties appear to go
together, can we draw any theoretical link that explains why these phenom­
ena constitute a distinct cluster of presidential regimes? We shall suggest
that, indeed, the designers of this group of regimes created strong presiden­
cies precisely to serve certain interests. Given that this group also includes
some conspicuous failures and problematic regimes, any theoretical con­
struct that accounts for their existence may also allow us to address how
presidential regimes can function more satisfactorily. Before proceeding
with a theoretical argument, however, there are some other observations
that are especially pertinent to the Chilean approximation of the archetype.
These observations concern the number of parties, for the Chilean party
system was most fragmented at precisely the time that it was least prone to
crises, which was also the time that the regime was most responsive to local
particularistic interests.

The number ofparties. A look at the number of parties in Chile, in addition

to their relative decentralization as organizations, will shed further light on
the functioning of the archetype as well as its unraveling in Chile. We shall
6 This contradicts Conaghan (1989), who argues that Ecuadorian parties are "loose. ·· In the
sense that it is easy to form new parties (low barriers to entry) and that the majority runoff
rule for presidential elections does not impose the incentives for broad coalitions that
plurality election does (see Chapter 10) , she is right. However, members' ability to be
elected depends on their ranking in closed lists, controlled by party leadership. Thus the
appearance of weak parties is dependent entirely on institutions favoring a fragmented
multiparty system, not on the rules for allocating seats within parties, as in, for example,
Brazil and Colombia. It is perhaps also possible that the provision for censure encourages a
proliferation of parties because politicians can form parties as vehicles for achieving ministe­
rial positions.
 Efficiency and inefficiency 179
look to the "effective" number of parties, by which we mean a weighted
measure that expresses the degree of fractionalization of a party system in
terms of how many "significant" parties there are. 7 The use of this indicator
is justified because it is the closest that we can come to expressing, with a
simple number, both electoral efficiency and representativeness. Because
efficiency is closely linked to the choice of competing policy options and
therefore to identifiability, it follows that as the effective number of parties
approaches two, we are approximating an efficient choice. 8 At higher val­
ues of N, it is likely that the party system is more representative (and
clearly tends to be less efficient). The correlation between higher N and
greater representativeness is not perfect - there might be only two impor­
tant political tendencies to be represented, as Riker (1982b) argued for
Austria, or maybe a low N disguises myriad intraparty factions from which
voters may choose, as in Colombia and Uruguay. The use of any simple
indicator conceals some information and reveals other. Still, we maintain
that N does a reasonably good job of capturing the trade-off between
electoral efficiency and representativeness.
Chile's effective number of parties based upon votes (NJ is displayed
over time in Table 9.2, along with the same measure for presidential candi­
dates. The values of N in congressional elections were at their highest when
the congress was most heavily based on local brokering of projects and
other constituent services, as explained by Valenzuela (1989b). Many of
these parties elected members in only one or two provinces, reflecting the
regional basis of Chilean politics at the time, but there were few single­
party regions. For the period from 1932 until 1957 the median Nv was 6.9,
an exceptionally high figure cross-nationally (Taagepera and Shugart 1989)
and one that implies little obvious connection between the vote and legisla­
tive coalition building. There was an extreme peak, at Nv = 11.8, in 1953,
when the "anti-politics" candidate, Carlos Ibanez, had recently been
elected, but the effective number of parties was in excess of 6 for this whole
period. For the later elections (1961-73) before the military takeover, the
median N is 4.9. In the last election, months before the coup in 1973, it was
2. 1. One puzzle here is that Chile's constitutional system did not face any
crises that threatened its existence when it had an extremely high number
of parties, but experienced a bloody overthrow as the culmination of a
7 The effective number of parties, N, presented by Laakso and Taagepera (1979), is a "self-
weighted" measure of the number of significant parties in the system. Rather than apply
arbitrary cutoffs below which a party is not considered a "party" for purposes of counting or
some other arbitrary method of assigning weights, N allows parties to weight themselves by
simply squaring the fractional share (whether seats or votes) for each party. These squares
are summed, and the reciprocal of the sum yields N. See Taagepera and Shugart (1989:259-
60) for a brief methodological discussion.
8 If there is a pure two party system, N will be less than 2.0 unless the two parties have equal
vote (or seat) shares. If there are two major parties, but also some smaller ones, N will be
greater than 2.0, perhaps around 2.5 or 3.0.
180 Presidents and assemblies

Table 9.2. Effective number of parties contesting congressional elections (Nv) and
effective number of presidential candidates (Np), Chile, 1 931-1973

�8
Date Nv
1931
1932 2.7
1933 10.6
1937 6.9
1938 2.0
194 1 7.7
1942 1.9
1945 6.5
1946 3. 1
1949 7.4
1952 2.9
1953 1 1.8
1957 7.4
1958 4.0
1961 6.4
1964 2. 1
1965 4.3
1969 4.9
1970 3.0
1973 2.1

period of a steadily consolidating party system, implying a failure just as

the regime - or at least its congressional election process - was becoming
efficient. Indeed, this puzzle is problematic for arguments such as those of
Linz (1987) and Mainwaring (1992a) that presidentialism and multipartism
are inherently incompatible. Another puzzle is that presidential elections
from 1931 until 1970 show a strikingly different side of Chile's party system.
In three of the first four presidential elections of the period, coalitions
were structured behind two principal candidates, as reflected in values of N
of 2.0 or less. The parties thereby offered voters the clear choice of two
competing alternative executives. This feature accords with the archetype
of presidentialism and with efficiency in the channeling of voters' expressed
preferences into executive formation. In only one of the next five presiden­
tial elections, that of 1964, did a two-candidate race occur again. The
divergent direction of Chile's party systems - simultaneous consolidation
in assembly elections and increasing fragmentation of executive elections -
may be seen in Figure 9.2.
Three of the two-candidate races occurred at precisely the same time that
the party system for congressional elections was most highly fragmented: in
1931, 1938, and 1942. For the congressional elections closest in time to
these presidential elections, the lowest value of N was 6.9 in 1937. Yet this
extreme fragmentation of the party system did not prevent the structuring
of national coalitions capable of generating majority electoral support be-
 Efficiency and inefficiency 181
1 2 .,...--------------------------,
a
a Nv
Nv average
a • Np
Np average
10

&

2 &
& a
&

0 -t--.---.-r--r---.---.r--,---r---r---ir---.----.---,---,,--.,.....--,-----,----1
1 930 1 935 1 940 1 945 1 950 1955 1 960 1 965 1 970 1 975

Year

Figure 9.2. Effective number of parties contesting congressional elections (Nv) and
effective number of presidential candidates (Np), Chile, 1931-1973

hind an executive. Even in the one election in which there were several
candidates, in 1932, one candidate brought together a coalition sufficient to
garner a majority of votes.

Majority-building in presidential elections. It was not until 1946 that Chile

experienced its first true multiparty presidential election under the 1925
constitution. The candidate who won the plurality of the popular vote,
Gabriel Gonzalez, had 40.2%, while the next two candidates, both of the
right, had 29.8% and 27.4%. This was thus the first election in which the
Chamber of Deputies would have the final say over who would become
182 Presidents and assemblies
president. The chamber ratified the electorate's plurality, voting for Gonza­
lez by a vote of 138 to 46 (Urzua 1986:66-8). Thus was institutionalized the
precedent, followed in future cases when there was no winner of a popular
majority, of choosing the plurality candidate. This respect for the plurality
winner was aided by (and probably depended on) the usual practice of the
parties' having put together a broad preelectoral coalition that could not be
beaten in congress by any new postelection coalition. This preelection
coalition-building resulted from the desire of parties to support a winning
candidate for an office that had substantial power over legislation. Until
the presidential elections of 1958 and 1970, parties usually did structure
broad and decisive coalitions before the presidential election; that of 1946
was actually one of the less broad coalitions, but still one candidate led the
rest of the field by a considerable margin.
In the next presidential election, 1952, while no candidate won a major­
ity, Carlos Ibanez came close, at 46.8%, and did so by offering a national
vision of sorts, if a highly negative one. He was the "outsider," campaign­
ing against the established parties. Despite his anti-party stance, the con­
gress, while dominated by members of those same parties, selected Ibanez
without incident. Thus the plurality principle was again upheld; it would
have been hard for parties to buck the bandwagon that had ensued behind
Ibanez even though constitutionally they could have chosen the runner-up.

Inefficiency in assembly elections. During the period just reviewed, in

which the number of parties was at its highest, congressional politics was at
its most local. Electoral lists in districts were regularly drawn up for ex­
pressly local purposes, rather than for the purpose of providing voters with
a set of ideologically coherent policy priorities at the national level. Mem­
bers of diverse parties, as well as local notables running as "independents"
would structure preelectoral pacts in which a vote for any of them helped
all their partners. After 1958, voters voted directly for a candidate within a
party list.
Valenzuela (1978, 1989a, 1989b), in various writings on Chile, has stressed
how that extreme localism and personal brokering of services rendered by
those elected coexisted with parties that were often highly polarized on
national issues. Parties thus were Janus-faced: highly personalistic in elec­
tions, highly combative in worldview. Indeed, one might say that the local
particularism was the "glue" that held together an otherwise polarized party
system. Despite the possible ideological gulf between them, members of a
particular locale's congressional delegation frequently came together across
party lines for the purpose of rendering services to their constituents. While
most presidential elections until 1958 were rather efficient in offering at least
one candidate who could articulate a national vision via a broad coalition,
congressional elections were conducted on entirely different kinds of cam­
paigns. As we shall see, it was when congressional elections also began to be
 Efficiency and inefficiency 183
conducted primarily around support or opposition to the national coalition
and the more ideologically driven policy proposals offered at that level, that
the system began to become unglued.
We thus might call the success of the Chilean system an approximation of
the archetypal presidential system before the late 1950s the "inefficient
secret." The secret was that the very inefficiency prevailing in the assembly
accommodated both logrolling and polarization without threatening the sta­
bility of the system, while broader coalitions, sometimes even ones that were
policy-based, formed for presidential elections. Congressional elections dur­
ing the period in question (about 1930 until the late 1950s) were highly local,
centering on district-specific and even citizen-specific benefits, what we
know as "pork barrel," patronage, and casework. However, the existence of
presidential elections, and especially the early emergence of a plurality char­
acter to those elections, meant that parties would need to present a broader
focus in those elections. Thus the extreme localism of congressional elec­
tions did not inhibit substantial policy innovation in the case of the Popular
Front, which was a majority coalition of the middle-class Radical Party and
leftists contemporary to, and bearing much substantive resemblance to, the
U.S. New Deal. We shall see shortly what changed to undercut the inefficient
secret of Chile's presidential regime, but first let us see if any other systems
have been characterized by this inefficient secret.

Other examples of the inefficient secret. We can identify some other cases
that could be considered examples of an inefficient secret, although each
has a considerably more powerful president, and this skewing of the bal­
ance can be linked to considerably less adequate functioning of these sys­
tems. Brazil from 1946 until the military takeover in 1964 bears some
similarity to Chile. The new regime in Brazil since 1988 is similar in many
respects. Colombia was another example of inefficient presidentialism.
The congressional electoral system in Brazil is one based on personal
voting within parties. Indeed, according to the rating system used in Table
9.1. Brazilian party leadership is weaker than its Chilean counterpart.
Members of congress pay special attention to bringing pork barrel benefits
back to their districts (or "bailiwicks" within their districts) in order to
cultivate a personal electoral following (Ames 1987). Brazilian parties are
extremely uncohesive, because the electoral system contains several provi­
sions that undermine party leadership in addition to simple candidate vot­
ing (Mainwaring 1991). We should like to emphasize another reason, not
mentioned by either Ames or Mainwaring, why party building is ham­
pered: the strength of a presidency that rendered the congress less effective
as a national policymaking institution than its Chilean counterpart. Just as
much as the electoral rules themselves, as we shall show, these constitu­
tional rules inhibiting party development were products of a congressional
rank and file that wanted weak parties.
184 Presidents and assemblies
As the rules of the policymaking game are skewed in favor of the execu­
tive, congress is left with little role beyond providing for its regionally based
constituents. Mainwaring characterizes Brazilian presidents at mid-century
as "circumventing" congress (Mainwaring 1992b), and he and Ames empha­
size the greater importance of pork and patronage rather than policy debate
in congressional elections. We are confident that the relative strengths of the
two country's presidencies, coupled with their electoral systems, account for
the regimes' divergent party systems and policy processes.
Colombia, especially before the 1991 constitutional reform, is another
case of a congress heavily based on localism, and with a very powerful
presidency (see Figure 8.1). Additionally, the party system, while generally
consisting of only two major parties, had been highly factionalized, mainly
along regional and personalistic lines. Splits and crises within the tradi­
tional Liberal and Conservative parties were every bit as critical as in­
terparty disputes in leading to the breakdown of Colombia's oligarchical
democracy in the 1940s. W hile intraparty splits are no longer viewed as a
threat to a democratic regime (Hartlyn 1987), the electoral system has
continued to reflect a legacy of regionalism and personalism by allowing
factions of each party to compete with one another. The progression of
institutional changes over time in Colombia demonstrates the importance
of both the electoral system and the constitutional allocation of powers in
facilitating or hindering the articulation of national policy goals.
In the 1930s the presidency had not yet been granted by congressional
majorities the sweeping legislative decree powers and primacy in economic
affairs that were parts of later constitutional reforms. Nor had the leaders of
the principal factions yet lost control over congressional nominations, as
happened when interparty competition was banned under the National
Front regime. Thus in the 1930s, the Colombian Liberal party was able to
carry out fundamental social reform in the so-called Revolution on the
March (Dix 1967). By the 1980s, however, Colombian parties were generally
regarded as almost completely feckless in addressing national policy chal­
lenges (Archer and Chernick 1988). Electoral lists for congress were not
drawn up with policy options in mind so much as to garner personal votes on
behalf of local vote brokers (Archer 1989, Shugart 1992a). The predictable
result of a personalistic and faction-ridden congress coupled with a president
holding dominant lawmaking powers was a congress ill-equipped to deal
with fundamental challenges facing the country. As in Brazil, only perhaps
more so, members of congress were reduced to being little more than lobby­
ists for the projects and patronage that enabled them to continue to deliver
votes in congressional elections. The congress, through the electoral system,
was structured so as to provide benefits on a particularistic, geographic basis.
Power to address broader interests was ceded to the presidency, but congress
retained important negative powers with which (until 1990) it was able to
check "reformist" tendencies in the executive (cf. Archer and Chernick
 Efficiency and inefficiency 185
1988). It was because of this preference of Colombia's rank-and-file congress
members for a constitutional structure that permitted them to service their
clienteles that an ironic situation had developed by 1990: an executive­
sponsored constitutional revision to weaken the executive's power over legis­
lation (cf. Shugart and Fernandez 1991). The explicit goal, only partially
achieved, was to make the system more efficient so that the congress and
executive alike could address serious policy challenges confronting the coun­
try. In short, Colombia as of 1990 perhaps had no longer an inefficient secret,
but simply an inefficient regime.

The U. S. case. The United States comes to mind as a case with some
similarities to the extreme electoral separation of the two elected branches.
However, we saw that the United States occupies a distinctly different
region of Figure 9. 1. Does this accord with the regime's actual perfor­
mance? In the United States, voting based on local interests contributes to
the frequency with which opposing parties control congress and the presi­
dency (Jacobson 1990), despite the fact that half of the congressional elec­
tions are held concurrently with the presidential elections. The safety of
incumbency has also been explained by the ability of members to cultivate
a "personal vote" (Cain, Ferejohn, and Fiorina, 1987), hence shielding
themselves from national policy swings.
In this limited sense, the United States might appear to resemble the
inefficient secret. However, there are important distinctions between the
United States and the other cases discussed in this section. One is that,
while congressional candidates are afforded considerable latitude by the
congressional party leadership to tailor their messages to their constituents,
the vote for Congress is still categoric by party: either a Democrat or a
Republican. 9 Thus the salience of pork need not be as great as in intraparty
preference electoral systems. A second critical difference is the lesser pow­
ers of the president over legislation than in those other cases. These two
features, of course, account for the regime's placement in Figure 9. 1. As a
result, while we lack an easy way to scale it, policy voting in congressional
elections is surely higher (hence less inefficient) in the United States than in
Brazil, mid-century Chile, or Colombia.
W hile the U.S. Congress has generally maintained a frequently maligned
tendency to deal with local services even at the expense of an articulation of
a national policy perspective in congressional campaigns (for a review, see
Cain et al., 1987), Cox and McCubbins (1992) have offered a convincing
counterargument to those who would sound the death knell for American
parties as organizations that structure policy debate at the national level.
Nevertheless, even the case for the strength of parties acknowledges the
9 Primary elections do not significantly undercut this efficiency, since in all but a few states,
voters are permitted in primaries to select candidates only within one party; unlike in the
other cases examined here, the intraparty choice is not simultaneous to the interparty race.
186 Presidents and assemblies
importance of local pork in maintaining personalistic loyalties between
congresspersons and their constituents, and the fact that party organiza­
tions in the United States have traditionally been strongest at the state and
even local levels, all indicators of relative inefficiency. National policy is­
sues remain less important in elections to the representative assembly in
the United States than in many other countries' assemblies, particularly
Britain's, and less important also than in U.S. presidential elections.
Our conclusions about the United States, then, are that despite their
considerable local tendency, U.S. parties are able to attend to policy mat­
ters in elections to a greater extent than in the more inefficient cases
primarily because of the electoral system and limited presidential powers.
If a congress has preeminent legislative power, parties cannot afford exces­
sive slack in discipline and inefficiency and still carry on with legislation.
Because the electoral system reduces the salience of pork in campaigns,
compared to the systems of Brazil, Chile, and Colombia, parties in the
United States do not need to distribute particularistic benefits to regional
politicians to the same great extent as in those other systems.

LOCALLY B ASED PARTY SYSTEMS :

WHY S U CH POWER F U L PRESIDENCIES?

Now we have seen that presidentialism and electoral systems interact to

shape the party system, but why is there a tendency for locally based party
systems and strong presidencies to be associated? Here we argue that the
answer lies in the origins of the regimes themselves. The Brazilian (1988) and
Colombian (1886) systems were devised by congresses or similarly elected
constituent assemblies, while the Chilean regime represented the culmina­
tion of a political tug-of-war between president and congress that had even
resulted in a civil war in 1891. Thus it is not surprising that all have afforded
great opportunities for representatives to attend to their local constituents.
W hat is perhaps more paradoxical is that, as we have observed, these very
same systems have tended also to have very strong presidents. Yet this is
what is suggested by Figure 9.1, above. Why would these assemblies have
approved constitutions allocating major budgetary, legislative, and even
"emergency" decree powers to an institution outside the congress?
The model that follows is an abstraction derived from a theoretical con­
sideration of two of the most extreme cases of strong presidents and weak
parties: the regimes of Brazil after military rule and the Colombian system
as it operated through most of the twentieth century. Later we suggest
modifications that help account for variations in some of the inefficient
systems, including the earlier Brazilian democratic regime and Chile. At
first, our discussion applies to the class of inefficient regimes more gener­
ally, as all contain certain similarities in structure and origin.
Let us start from a very basic assumption. What do members want? We
 Efficiency and inefficiency 187
assume that they want to further their careers as politicians. To do so, they
want to be able to bring benefits back to their regions in order that they
might amass still greater political capital. The institutional arrangement
that emerges is one permitting these representatives at once to free them­
selves and to bind themselves: to free themselves of direct involvement with
broader national policy matters by creating a central, accountable agent to
handle such matters without impinging on their ability to attend to constitu­
ent needs, and to bind themselves against excessive spending that could
undermine the system itself. Let us now develop this idea further, first by
asking the counterfactual: W hat if these systems had been parliamentary?

Freedom to service constituencies

Under a parliamentary system, with its vote of confidence and the possibil­
ity of dissolution, these representatives would have been inevitably drawn
into forming and supporting cabinets responsible for national policy. As we
have discussed, in a parliamentary system, the only institution disposed to
design and implement national policy is the cabinet. The responsibility of
the cabinet to parliament makes parliament itself responsible for policy.
Then either party whips become critical in sustaining governments in office
through enforcing party discipline (e.g. Britain), or else the executive itself
becomes unstable through frequent votes of no-confidence (e.g., Italy's
trasformismo) .
The very separation of powers afforded by a presidential system, on the
other hand, allowed these regionally based vote brokers, often tied to big
landowners on whom election to congress depended, to delegate national
policy responsibilities to an executive such that they would not bind them­
selves to a particular set of policies that could impinge upon the local
bosses' autonomy. We do not imply by this that congress simply abdicated
policymaking because its members did not care about these matters or
preferred to leave the "tough questions" to someone else. Instead, in a
manner similar to that described by Kiewiet and McCubbins (1991) for the
United States, we see a congressional majority delegating to achieve cer­
tain goals. As we have stated, when the method of election is one that
places overwhelming importance on providing services to constituents be­
cause of factional competition, a paramount goal of majorities will be to
obtain the resources to protect members' personal followings. Thus, mem­
bers of this majority may advance their careers through keeping constitu­
ents' needs addressed, while a central agent, the president, can be charged
with and held accountable for policymaking. As we saw in Chapter Ts
discussion of delegation, congressional majorities retain ways to ensure
that these presidential powers are used within acceptable limits, through
the ability to cut expenditure requests, pass ordinary legislation, or even
reform the constitution itself. The congress can do so without having either
188 Presidents and assemblies
to sacrifice autonomy in the assembly or else to surrender the advantages of
having a single identifiable agent for general policy, as would be the case in
a parliamentary form of government.
Even if the foregoing characterization is correct, one might wonder, why
a separately elected presidency? Perhaps the assembly-independent form of
government discussed in Chapter 8 would be more appropriate for such
politicians, since it would create an executive more dependent for its origin
on the congress than a directly elected president, but would still avoid the
vote of no-confidence and the resulting party whips (or cabinet instability).
Indeed, some of the early Latin American constitutions did provide for
approximations of such regimes, if we include an executive elected by an
electoral college composed of other elected officials (and therefore con­
trolled by the local politicians in question) to be more appropriately charac­
terized as an assembly rather than a presidential regime. Also, as we have
seen, the Bolivian system remains as a legacy of this type of executive
selection process. 10 In Chile, too, the constitution provided for congres­
sional ratification of alliances stemming from electoral success in prior,
locally contested elections. Even where presidents came to be separately
elected and not subject to congressional ratification, as in Colombia in 1910
(Gibson 1948), the electoral process often remained one in which the local
elites and their followers rather than the state retained control over the
distribution of ballots, as indeed was the case in Brazil in the 1950s and in
Colombia until the election of 1990. Where regional authorities and party
workers control even presidential elections, the emergence of efficiency is
retarded at this level nearly as much as in locally contested congressional
elections. Our argument is that this was exactly the intention of those who
designed, developed, and sustained these regimes.
Most of these cases also employed nonconcurrent electoral cycles. Given
that nonconcurrent elections afford greater independence of locally based
politicians from presidential coalitions (see Chapter 10), we should expect
these politicians to prefer such an electoral cycle. Indeed, in Brazil's consti­
tution of 1946 elections for congress were initially concurrent, but the
congressional majority shifted to separate dates after 1950. In Colombia,
too, before the National Front, congressional elections were regularly held
nonconcurrently. During the National Front, they were held with the presi­
dential election, while from 1974 until 1990, they again were held before
the congressional election.

Binding congressional majorities

We have now discussed several ways in which the creation of the presi­
dency, and a powerful one at that, served to free congressional majorities
10 A difference in Bolivia is that, since the 1952 revolution, Bolivia's electoral system has
featured closed lists ( Gamarra 1987).
 Efficiency and inefficiency 189
to attend to c onstituent interests on which their electoral success depended.
N ow it is w orth c onsidering ways in which this instituti onal arrangement
served t o bind the c ongress. The creati on of a p owerful presidency w ould
keep spending fr om g oing out of c ontr ol. Servicing c onstituencies through
the all ocati on of budgetary res ources that pr ovide subsidies, pr otection,
j obs, and other specific benefits is an expensive undertaking. Left un­
checked, such a pr ocess c ould easily kill the g o ose that laid the g olden egg.
Thus the g o ose is pr otected by creating an instituti on that exercises p owers
such as limiting the maximum expenditure f or particular items or for the
wh ole budget and a partial veto. We argue that this instituti onal arrange­
ment offers the c ongressi onal maj ority a credibly enforceable c ommitment
t o s ome degree of fiscal restraint.
W hen a president p ossesses a partial vet o, the ability of c ongress mem­
bers t o l ogroll within congress is undercut. Instead, their ability to d o s o
depends o n having ties directly t o the president, wh o allows p ork on the
basis of transacti ons with c ongressi onal c oaliti ons. Thus the l ocus of l og­
r olling shifts much m ore decisively t oward the president than it d oes in
systems in which the president lacks a partial veto. The restricti ons on
increasing the president's pr op osed expenditures w ork much the same way.
Since members c ontinued t o w ork within an electoral system that encour­
aged them to cultivate pers onal f oll owings, the instituti onal incentives were
f or them t o be be little m ore than l obbyists f or regi onal particularistic
interests, which, we argue, was what they intended f or themselves t o be.
It c ould be objected that fiscal restraint is a public g o od, s o why w ould
members g o along with it? In sh ort, what are the incentives f or obeying?
W hat levers exist? M ost imp ortant is that the ability to pr ovide the private
g o ods on which electi on depends is itself dependent on deals struck with
the president, because the partial vet o gives the president the ability t o set
limits t o expenditures on particular items.

Origins of the inefficient systems

If this acc ount of why s ome presidential regimes exhibit b oth a congress
elected al ong clientelistic lines and an executive with great legislative
strength is c orrect, then there sh ould be levers by which the c ongress w ould
retain ultimate auth ority even over such a str ong executive, as well as ways
by which c ongressi onal interests w ould ensure the continuation of their
elect oral perquisites. Otherwise we indeed w ould have examples of abdi­
cated auth ority - and we are arguing that this is n ot the case. Since Brazil
(1988) and C ol ombia (1886) are tw o of the str ongest presidencies that we
have seen, aside fr om the three-year-l ong experience in Chile after 1970,
these are the two that sh ould c orrespond m ost cl osely t o the the oretical
discussion of this secti on. After c onsidering these cases, we return t o the
Chilean case.
190 Presidents and assemblies
Both the Brazilian constitution of 1988 and the Colombia of 1886 (and to
a lesser extent, as we shall see, Brazil's of 1946) were products of constitu­
ent assemblies whose representation ran along regional, clientelistic lines.
Thus both constitutions should be expected to represent the interests of
these regional authorities as they were portrayed above: freeing congress
members to service their constituents rather than deal with broader policy
concerns, and binding themselves from spending extremes. If this is so,
wherever there are extraordinary powers in the hands of the executive we
should expect some countervailing check by the congress to make sure that
the executive does not wield this power in ways that threaten regional
interests. If the strong president is the guardian of the system, who guards
the guard himself? W hat mechanisms exist to bind the president, in other
words, to use the powers responsibly? The first is in the electoral system.
Because members of the same party compete against one another as well as
against other parties to provide constituent services, and because the presi­
dent still needs legislative majorities to use the powers effectively, the
president's ability to have such support depends on being able to get pre­
ferred congress members elected. The low barriers to election (given some
minimal access to funds) 1 1 and especially the cultivation of clienteles in
both countries mean that the president's ability to see one preferred mem­
ber or factional group elected in some district instead of another, depends
on providing budgetary resources to benefit that member's or faction's
constituents. W ith so little central leadership control in either case over the
specific members who are elected, the presidency, despite all its great
strength, is dependent for congressional majorities on who has the vote­
delivering ability. Vote-delivering ability, in turn, is greatly enhanced by
having central budgetary resources. The result is a considerable degree of
mutual dependency (cf. Shugart and Fernandez 1991). W ith such a decen­
tralized congress one's ability to get access to funds by dealing wholly
within the congressional labyrinth would be low, but by making deals with a
powerful central figure, one's prospects for getting funds are improved.
For the president, such deals are essential for getting legislative support,
hence funds are likely to be directed where they have the greatest electoral
payoff.
We have been speaking of the need for legislative support. But why is
this critical for such a strong president? One major reason is that ordinary
legislation passes by simple majority. This, of course, is hardly unusual; 12
what is unusual in these two cases is that a veto may be overridden by a
majority of the congressional membership, rather than by an extraordinary
11 In Colombia, members each had a small share of the national budget (known as auxilios
parlamentarios) to spend in their districts as they saw fit. Hartlyn (1987:99) refers to these
funds appropriately as "pork-barrel funds."
12 The first ten years of the National Front were an exception, during which legislation
required a two-thirds vote for passage.
 Efficiency and inefficiency 191
(for example, two-thirds) majority. Thus the congressional majority is sov­
ereign over legislation, subject to the president's ability to delete congres­
sional amendments through the exercise of the partial veto. Only during
the Liberal "Revolution on the March" in the late 1930s was a veto requir­
ing an extraordinary majority to override enacted in Colombia, and then
only on expenditure bills (see Gibson 1948:403). In the earlier Brazilian
regime, which contained strong legacies of the populist Vargas movement
(Mainwaring 1992b) , there likewise was an extraordinary majority to over­
ride vetoes. In both instances, the existence of nationally focused parties or
movements that controlled the presidency at the time led to the adoption of
stronger veto authority, which, as a way to enforce consensus, could
weaken the sway in congress of clientelistic majorities in favor of the more
national oriented (often populist) presidential coalition. When clientelistic
majorities prevailed, as in Colombia as well as Brazil after military rule,
vetoes were weaker, being easier for the congress to override. 13 The great
presidential powers that were provided were in areas other than ordinary
legislation, including budgeting.
In the inefficient systems, budgetary powers are, to varying degrees,
important. The ability of the congress in both countries to reduce the
amounts allocated in the president's budget to particular items also is a
meaningful weapon for congress, even if congress does not have the author­
ity to expand the overall budget. By the threat of reducing items that the
president wants, the congress can keep the president from using even his
decree powers rampantly. The members of the congressional majority re­
tain the means to hurt the president in areas important to him, even while
simultaneously retaining the limit on their own ability to increase items (or
overall spending) as a means to bind themselves. Finally, prohibiting the
reelection of the president would prevent this official, once in office, from
developing a popular constituency with which he could turn around and
campaign against the "wasteful" spending by which the congress members
serviced their constituents.
Perhaps the clincher for this theory of a congress ultimately able to get
what its constituents want out of a strong presidency is congress's own
sovereignty over the constitution. In neither case (nor in Chile and many
other presidential systems) is the constitutional amendment process even
remotely as cumbersome as in the United States. Congress retains the
authority to reform these constitutions. In Colombia, as well as under the
Brazilian constitution of 1946, congress could do so by absolute majorities
13 Other differences in the two Brazilian regimes relate to elections. Under the 1946 constitu-
tion, elections at first were concurrent, meaning they were more likely to be contested
around national (at the time, populist) policy concerns. Moreover, in 1945 the electoral
rules favored the largest party. Subsequently, including the new regime, elections have
been nonconcurrent and electoral rules have favored fragmentation. See Mainwaring
(1991, 1992b).
192 Presidents and assemblies
of the whole membership in two consecutive sessions. It thus would take
time, but it would not take an extraordinary (for example, two-thirds)
majority. A president would thus have to exercise some degree of caution
when contemplating making law by decree: congress could amend away
these powers if they were abused; theoretically, it could even amend away
the presidential institution itself or at least bind the executive to congress
more closely, perhaps by instituting censure of cabinet members. Our point
is not that there were not any agency losses as a result of this substantial
delegation of powers to the executive (see Chapter 7). Rather, our point is
that the benefits of delegation discussed here for these congressional politi­
cians must not have exceeded the cost entailed in agency loss, or else
reforms to the process would be anticipated. The calculation that costs had
outweighed benefits would have been most likely had a reformist or "out­
sider" become president and attempted to use his powers to the detriment
of the congressional interests. Such a result did obtain in Brazil in 1961, as
we discuss below. 14
If the constitution that granted legislative faculties to presidents was left
easily reformed by congress, the electoral law in Colombia on which the
whole system depended was put beyond the reach of mere majorities. The
electoral law required a two-thirds majority to reform under the 1886
constitution. Thus we have powerful testimony to what was more impor­
tant to the designers of this constitution. The strong presidency was little
more than "rules" by which to achieve congressional desires, and the
strength could be modified at relatively low cost if necessary; the electoral
law, on which their ability to service constituents and be free of party
discipline ultimately depended, was made more difficult to change.
In Brazil, as well, despite having a federal system that might have encour­
aged a constitutional amending process similar to that of the United States,
reforms to the basic institutions are relatively easy to make. The majority
required in the 1988 constitution is three-fifths of each house, making the
barrier higher than in Colombia or the previous Brazilian democracy, but
lower than in many systems. There was one very prominent precedent for
the use of this amending capability to curtail the powers of a president in
whom much of the congressional politicians did not trust. In 1961, in re­
sponse to the ascension to the presidency of V ice-President Joao Goulart
upon the resignation of J ania Quadros, congress passed an amendment
creating cabinet responsibility before the congress and weakening the veto
override from two-thirds to three-fifths. Moreover, in future constitutional
terms, the president would have been elected by congress, had the system
survived. Such a reform underscores the potential for a congress to check
14 An ultimate irony of the Colombian case is that when such a reformist president did win
office, he used his State of Siege powers to invoke a constitutional reform process that
sought to replace the system with a more efficient one. See, in addition to the brief
discussion below, Shugart (1992a) and Shugart and Fernandez ( 1991).
 Efficiency and inefficiency 193
even a president who is constitutionally very strong, or eliminate the inde­
pendent executive altogether - provided, of course, that the constitution
has left it relatively easy for congress to do so.
How does the story just told about congressional wishes in Colombia and
Brazil apply to the case that we have suggested so typifies the presidential
archetype? The Chilean regime before 1970 featured a congress at least
marginally more capable of addressing national policy matters and a corre­
spondingly more ideological party system that represented more diverse
interests than in Colombia and Brazil. This implies that regional interests,
which might have favored the highly inefficient type of regime seen in those
other two countries, either were not present or were forced to compromise
with other interests. Either way, they were weaker and the resulting consti­
tutional structure placed less legislative burden on the presidency, making
it more of the veto gate that it is in "classic" separation of powers theory,
rather than the "supreme legislator" (Vasquez Carrizosa 1979) for national
policy. The origin of the Chilean constitutions through 1970 was much less
based in congress or regionally controlled constituent assemblies than was
so in Colombia or Brazil, particularly in 1988.
The historical antecedent of the constitution that was in effect in Chile
through the major reforms of 1970 may be found in a presidential commis­
sion after a brief civil war in the 1830s (Sigler, Blaustein, and Flanz 1981).
This constitution granted the president an absolute veto but not the exten­
sive legislative decree or budgetary powers that were to typify the constitu­
tions of Colombia and Brazil. The absolute veto was reduced to one
requiring a two-thirds override when congressionally supported forces
defeated those of the president in the civil war of 1891, ushering in the so­
called parliamentary republic. The constitution of 1925 resulted from a
military-presidential intervention against congressional control over cabi­
nets in the previous periods, but did not alter the basic legislative powers
of presidents from what they had been since 1891. The irony, then, is that
in Chile, greater presidential control over the constitution drafting in 1925
(with the legacy of the 1831 document) resulted by 1925 in a president
with stronger package vetoes but significantly weaker budget and decree
powers. This is not so surprising however, as long as it is recognized that
the legislative powers other than the package veto are presidential facul­
ties more desired by the regional interests represented in the assembly
itself. Thus, where the constitution originated in a process that already
checked the ambitions of these (typically landowning) congressional inter­
ests, as with the mid-eighteenth-century military leaders in Chile and
Vargas in Brazil, those institutional features most favorable to them were
attenuated. The result was that the Chilean Congress became a highly
dynamic legislature. Those of Brazil after military rule and Colombia, on
the other hand, did not, as the regionally based politicians were better
positioned to craft a regime that would serve their interests.
194 Presidents and assemblies

E FFI C I E N T P R E S I D E N T I A L I S M : S T R O N G P A R T I E S A N D
WEAK PRESIDENTS

We have been discussing a cluster of presidential regimes in which congres­

sional elections are or were conducted mainly around the provision of
locally targeted expenditures promised and delivered by candidates and
congress members who have considerable autonomy from national leader­
ship, a form of politics we have branded as inefficient. While we have been
careful to point out important distinctions between Brazil and Colombia,
on the one hand, and Chile, on the other, all of these regimes are differenti­
ated in Figure 9. 1 from another cluster. In the lower right of that figure, we
find a series of regimes in which party control over local rank and file
means that individual candidates and members of congress cannot outbid
one another in providing benefits because their ability to be elected de­
pends on the collective performance of the party rather than on their own
ability to generate resources. We have branded this form of politics as
efficient. We shall focus here on one case, Venezuela, that is almost the
opposite extreme of the Brazilian and Colombian cases.
The method of electing the assembly is fundamental to understanding
why Venezuela and some similar systems do not resemble the archetypal
form of presidentialism. The Venezuelan Congress is not elected with locali­
ties as the basis of representation, as was Chile's during the period of the
inefficient secret, or as is that of the United States. Instead it is elected
from closed party lists, 15 in which voters select only a party and are not
permitted to select a specific candidate within the party. Since being placed
sufficiently high on the party list is the key to being elected, central party
leaders are tremendously powerful ( Coppedge 1988).
While election of the congress from closed party lists might make the
articulation of preeminently local concerns more difficult than the election
of the congress on the basis of individuals and what they can bring to their
constituents, such an electoral system might be expected to make possible
the other form of constituent representation discussed in Chapter 1: that of
ideological or partisan interest. Yet in Venezuela since the early 1970s,
there have been two main parties, thus limiting the degree of diversity that
can be represented while enhancing electoral efficiency. Each major party
15 Regional multi-seat districts are used, but several features of the electoral system in use
from 1958 to 1988 sharply limited the opportunity for the articulation of local interests.
Most prominent among these was the totally closed-list format, meaning that individual
candidates lacked the incentive to cultivate a personal vote. A vote for the list was a vote
for everyone on the list, with the order - and thus the electability - of candidates predeter­
mined by the leadership. Additionally, the concurrence of these elections with the national
presidential election and the existence of nationwide seats to compensate minority parties
for district-level deviations have further restricted localism. Reforms enacted after the
1988 elections are expected to increase the articulation of local interests, although perhaps
only marginally. See Shugart 1992a.
 Efficiency and inefficiency 195
is formed ar ound a principal c ompetitor for the presidency. As we shall see
in Chapter 10 and 1 1, when presidential and congressi onal electi ons are
c oncurrent, as they are in Venezuela, a party system d ominated by tw o
maj or parties is the n orm, pr ovided that the president is elected by plural­
ity. But for the first three electi ons under the new regime f ounded in 1958,
there was a multiparty system. The functi oning of the Venezuelan system,
in which the president's legislative p owers are far weaker than in Chile,
Brazil, or C ol ombia, sh ows h ow the strength of the presidency, the elec­
t oral system, and the electoral cycle w ork t ogether t o impact the degree of
efficiency of the system.

Venezuela's transitional multiparty system

In Venezuela, the presidency was originally c onsciously devalued by the
leader of the parties as part of a p olitical pact t o ensure the stability of the
y oung dem ocracy. The idea was that if one party c ould gain s ole c ontrol
over a p owerful office, the others w ould likely fail to respect the new rules
of p olitical c ontestati on. In the system designed, the president lacks a vet o,
making the c ongress superior in terms of legislation (see Chapter 7).
The practice of Venezuelan dem ocracy up t o 1968 permitted multi­
partism t o fl ourish in the 1963 and 1968 electi ons, but a tw o-party f ormat
subsequently emerged. Am ong all the electi ons for plurality presidencies
and c oncurrently elected PR c ongresses f or which we have data, there are
strikingly few in which the number of "seri ous" presidential c ontenders
exceeds tw o. 16 Only three cases clearly stand out thereby as excepti ons to
the rule: the D ominican Republic in 1990 and Venezuela in 1963 and 1968.
Such electi ons are certainly aberrant, 17 but the case of the Venezuelan
multiparty electi ons deserves s ome further c omment.
Fr om 1958, the date of the founding of the Venezuelan dem ocratic re­
gime and of its first electi ons, until the inaugurati on of the president
elected in 1968, Venezuela was g overned by coalitions. The Pact of Punt o
Fij o, in which the principles of elite c onciliati on were laid d own, pr ovided
that the winter of the 1958 presidential election w ould share p ower with
other parties regardless of who won. W hile the parties could n ot agree on a
c omm on presidential candidate, neither was there any incentive f or the
opp ositi on t o the maj or party (Acci6n Dem ocratica) to c oalesce, given the
limited stakes of the presidential electi on. Indeed, the president was meant
t o be above politics (Coppedge 1992), and with the c ongressi onal electi ons
serving t o determine the r ough makeup of c oaliti on cabinets, the incentives
were much m ore in the direction of fragmentation than c ons olidati on.
While s ome parties left and others j oined c oaliti ons between 1958 and
16 A party system with two big parties, each around 40% of the votes, plus one or two
secondary parties will have a value of N around 2.9.
17 Their occasional occurrence is explained in Appendix A.
196 Presidents and assemblies
1968, the f ormal practice of c oaliti on g overnment c ontinued until the first
n on-AD vict ory in 1968 (with only 29% of the v ote) confirmed that alterna­
ti on was p ossible and led t o the end of the practice of f ormal executive
c oaliti ons. President Rafael Caldera exercised his c onstituti onal prer oga­
tive t o ch o ose the ministers he wanted and g overned with single-party
cabinets despite his narr ow plurality. Thereafter, the party system c ons oli­
dated int o the two-party d ominant system seen after 1973.

Efficiency and representativeness in Venezuela

Before the tw o-party d ominant system emerged, Venezuela had a highly
representative c ongress but inefficient executive elections, owing t o multi­
partism. W ith n o maj ority party ( or even a near-maj ority party) in 1963
and 1968, the ability of v oters t o identify and end orse a package of execu­
tive p olicy opti ons was very weak.
Once the party system c oalesced ar ound tw o maj or parties that offered
meaningful c ompetiti on f or the presidency, identifiability increased. Thus
the system became m ore efficient, but only t o the extent that the presi­
dent's party c ould c ommand maj orities in the assembly, since the president
lacks f ormal legislative p owers. On the other hand, if the president's party
is merely the plurality party, but the maj or opp ositi on party can c oalesce
with min or parties t o c ontr ol the assembly, the p olicy opti on end orsed in
the presidential electi on may be rendered feckless. At times such a coali­
tion has f ormed t o "gang up" on the president and t o legislate over his
head. Having a c ongress c omp osed of representatives wh o have been
elected principally on presidential candidates' c oattails and are n ot dis­
p osed t o c ompr omise individually, plus a president without a vet o, is not a
recipe f or either faithful representati on or sm o oth executive-c ongressi onal
relations. W hat it d oes is make f or an unusually maj oritarian form of
presidentialism, since the presidential check on c ongressi onal maj orities is
minimal. By being majoritarian and having such strong party discipline,
this system bears c onsiderable resemblance t o the efficient secret of Brit­
ain, but with an imp ortant exception: the p ossibility that a president and
the c ongressi onal maj orities might be inc ompatible.18
Many Venezuelans have been widely critical of their system f or its c on­
centrati on of p ower in the party leadership, even c oining the term "party­
ocracy" (C oppedge 1988; see als o Shugart 1992a). Yet, despite these criti­
cisms, it remains true that this regime and tw o others that share with it the
same regi on of Figure 9. 1 (C osta Rica and the D ominican Republic, b oth
of which have slightly str onger presidents) are am ong the m ost stable and
18 It has never happened that one party held the presidency while another held a majority of
the congress. As mentioned above, however, there have been times when a coalition of
parties in opposition to the president has legislated over the objections of the president.
 Efficiency and inefficiency 197
crisis-free examples of presidentialism. Let us consider why that might be
the case and whether we can generalize from the experience.

Origins of efficient presidentialism

Our preceding discussion of the cases that combine strong presidents and
weak parties led us to the conclusion that such a combination was desired
by the very politicians who designed and maintained those regimes. Per­
haps a similar story can be told about the regimes that combine compara­
tively weak presidents and strong parties. We theorized that the inefficient
regimes had their origin in regionally based constituent assemblies that
brought together politicians who sought a system that would allow them to
continue to provide constituent services to their regions. The efficient re­
gimes, on the other hand, have had their origin in revolutions or civil­
military coups that detroyed traditional dictatorships and brought to power
rather cohesive, reformist, and nationally oriented leaderships. In Venezu­
ela, the Acci6n Democratica took power in a conspiracy with part of the
military in 1945 and again in 1958 - this time in a pact with other parties.
The AD was originally Leninist in its internal structure. More significant is
what AD was not: It was not a collection of regional politicians. It was a
social-democratic party of intellectuals and advocates for the urban middle
class and the peasantry (Martz 1966) . Similar origins typify the parties that
dominated the constitution writing in Costa Rica after Jose Figueres's insur­
rection, the Dominican Republic after the fall of the dictator Rafael
Trujillo, and Nicaragua after the Sandinista insurrection. But why would
the particular combination of strong parties and weak presidents emerge?
Each system just mentioned has used closed party lists, which is hardly
surprising when the leadership is in as commanding a position as in a post­
insurrectionary period. Rather than permit local rank and file to contest
elections and demand resources independently of the leaders' agenda, the
party leadership could ensure control over its activists even after the initia­
tion of competitive elections by means of control over lists of candidates.
Logically, it also follows that party leaders who expected to be able ordi­
narily to dominate their rank and file, and who were relatively unconcerned
with servicing particularistic geographic interests, would have less need to
delegate legislative and budgetary powers to an external agent. The coordi­
nation and binding of members, which we discussed with reference to the
Brazilian and Colombian cases, could be accomplished through leadership
control over electability instead of through extraordinary legislative powers
in the hands of a president. But that is not the full story; by maintaining an
assembly more powerful than the president, the party (or parties) could
ensure its continued viability as a policy-making body rather than risk becom­
ing marginalized by a powerful president. Yet each of these founding parties,
198 Presidents and assemblies
or revolutionary movements, was identified with a particular dominant
leader. Thus it is not terribly surprising that they should create a presidency
to allow that leader a fairly autonomous position from which to exercise
limited executive power. We discussed in Chapter 5 the variations of presi­
dential reelection; here we need only recall that these efficient regimes
include the Dominican Republic and Nicaragua, two of the three Latin
American systems that do not restrict presidential reelection.
The result of this process of a regime emerging from a reformist party's
hold on power rather than from regionally based politicians is an institu­
tional system that permits efficient voting on the basis of competing parties
and encourages the nationalization of politics. It does so especially if con­
gressional and presidential elections run concurrently and plurality rule is
used for the presidency, as in Costa Rica, the Dominican Republic, Nicara­
gua, and Venezuela. In placing such a premium on national policy instead
of local and more particularistic interests, the regime does not conform to
the archetype we identified earlier. Indeed, it is the excessive deviation
from the archetype - too little constituent responsiveness - that has led to
calls for electoral reform in Venezuela to reduce the grip of leadership.

An optimal presidentialism?
W hile our primary purpose in this book is not to prescribe a particular
configuration of variables as being best according to some universal set of
criteria, we do from time to time find that some ways of organizing a polity
are preferable to others. If efficiency in the translation of voter preferences
as expressed in elections into actual policy outcome is desirable, but we do
not want to go to the extremes of Venezuelan-style partyocracy, we can find
from the preceding discussion possible means of achieving such goals. It is
reasonable to conclude that there might be potential institutional designs
that grant only minimal legislative powers (greater than O but no more than
3) to the president while retaining considerable but not overwhelming lead­
ership control over parties (perhaps what we would score in the range of 5
or 6) and that such a system would be efficient, but not to the utter exclu­
sion of regional articulation.
What is questionable, however, is how such a regime would be brought
about politically, given that powerful interests whose assent would be
needed would likely exercise their vetoes to prevent reform. In the case of
Venezuela, the well-esconced leaderships of the two dominant parties had
much to lose if they were to weaken their influence over congressional
candidacies. As a result, while the reformed congressional electoral sys­
tem, passed for the 1993 elections, creates single-member districts for
about half the seats and thus raises the potential for more constituent
responsiveness, the remaining seats will continue to be elected from closed
lists (Shugart 1992a). It remains to be seen, of course, what the effects of
 Efficiency and inefficiency 199
this system will be on enhancing representativeness within the parties. The
experience does suggest the difficulty of bringing forth a potentially "opti­
mal" configuration of variables already suggested. The Colombian constitu­
ent assembly of 1991, in which both national-reformist and regional inter­
ests obtained representation in a single nationwide district, might have
suggested a process out of which a more efficient regime could emerge.
However, the weight of the local interests that constitute the traditional
parties proved too strong for dramatic changes to result (Shugart and Fer­
nandez 1991; Shugart 1992a). Thus, what the Colombian experience of
1991 does is confirm again that constituent assemblies, as in 1886 in Colom­
bia and 1988 in Brazil, are reflective of the interests of a majority. In an
already particularistic political system, that majority, if elected through an
essentially unchanged electoral system, is likely to want to maintain weak
parties and a strong presidency.
We can relate one experience in which a reformist party emerged within
an inefficient system and sought to bring about increased efficiency in a
way that only increased conflict. We thus return to the Chilean case and its
demise, for that case is yet further confirmation for one of our more gen­
eral points: Powerful presidencies are problematic for the stability of presi­
dential regimes. The quest for efficiency, where desired, should lead re­
formers toward a combination like that suggested as "optimal," at least by
reducing presidential powers rather than enhancing executive faculties.

THE DEMISE OF THE INEF FICIENT SECRET IN CHILE

Changes made in Chile in the final years before 1973 demonstrate that the
inefficiency of congressional elections combined with moderately strong
presidential legislative powers were the secret to that regime's previous
success. The first of these changes that began to undermine the inefficient
secret concerned the electoral laws. In 1958 a center-left coalition abol­
ished joint electoral lists in an effort to eliminate the "corrupt" practice of
structuring pacts among candidates of different parties at the local level for
mutual electoral benefit. Valenzuela (1989b:182) says that "while this re­
form succeeded in making pre-election arrangements less 'political,' it also
eliminated an important tool for cross-party bargaining. " This reform coin­
cided with the period in which an ideological Christian Democratic party
was emerging alongside the already ideological left and the increasingly
defensive conservative parties. It was inspired by a desire for efficiency, but
undermined the basis for accommodation that, while certainly inefficient,
was the glue of an otherwise polarized political system. The resulting re­
gime was well off the axis that runs between efficient and inefficient re­
gimes in Figure 9.1, foreshadowing its difficult combination of a strong
presidency and parties increasingly attuned to national rather than mainly
local concerns.
200 Presidents and assemblies
But it was the constitutional reform of 1970 that went farthest in under­
mining the consensual basis of Chile's presidential system, placing the
regime in a category all its own: exceptionally strong president and moder­
ately strong parties. A constitutional amendment secured by the Christian
Democratic administration of Eduardo Frei with support from the right
required that no amendments could be attached to a bill unless they were
directly germane to the bill itself. This was tantamount to enhancing the
president's partial veto, especially because only the president could initiate
many items of expenditure. The partial veto, even with an extraordinary
majority override, allows congress to engage in attaching amendments to
bills in order to secure passage, but delegates to the president the authority
to decide which logrolls will be accepted. While, as we noted above, this
means that much logrolling is transferred from congress to congressional­
presidential negotiations, it at least permits such trade-offs to be made. If
congress cannot make amendments that are not relevant to the main bill, 19
then the space for logrolling is not simply reduced and transferred to a new
arena, as with a partial veto; it is eliminated. Thus a process begun in 1958
accelerated: The role of legislative bargaining across parties was being
sharply reduced throughout this period.
These reforms, cases of political engineering designed to "rationalize"
both local-level electoral politics and national policymaking, led to incipi­
ent efficiency in congressional elections. That is, congressional elections
came to be statements on preferred national policy direction. But to func­
tion with a true efficient secret, elections for seats in the assembly would
have to have been capable of determining the direction of the executive or
at least the policy output of the policial process. However, by 1970 the
presidency has been granted such great legislative powers as to characterize
the regime by the most dominant of presidencies, as seen in Figure 8.1.
Thus congressional elections took on far less import in the actual making of
policy, making conflicts less manageable, as the effect was to place ever
greater weight on the presidency. By doing so, Chilean politicians had
made losing the presidency all the more costly to any of the three principal
ideological blocs in the Chilean party system. It was during this period,
therefore, that coalition building for that national executive became more
complicated. The paradoxical result, then, was that reforms enacted in the
name of efficiency for the polity as a whole utterly failed. As shown in
Figure 9.2, only in the last congressional election were voters afforded an

19 The amendment also created a five-member Constitutional Court, three of whose mem­
bers were appointed by the president, subject to assembly confirmation, and eligible for
reappointment at the end of a fixed four-year term (shorter than the president's term by
two years). Two other members were appointed by the Supreme Court from among its
own members. Among the new court's functions was to pass judgment on whether bills
were confined to one subject. Note that the new court's composition was such as to make it
rather responsive to presidential preferences.
 Efficiency and inefficiency 201
efficient ch oice - m onths before the c oup that rem oved a c onstituti onally
p owerful president wh o represented the p olicy preferences of the l osers of
that c ongressi onal electi on.

P residential elections: to coalesce or not to coalesce?

We have seen that, despite a very high number of parties, in the age of the
inefficient secret parties n ormally c oalesced sufficiently bef ore presidential
electi ons t o offer c ompeting nati onal candidates capable of obtaining a
maj ority ( or cl ose t o it) of the v otes. H owever, as c ongressi onal electi ons
came t o be m ore p olicy- oriented, only one of the three presidential elec­
tions offered a tw o-way race. Tw o others were w on by candidates receiving
less than 37% of the v ote.
The results of the one tw o-candidate race suggest the c osts of c oaliti on
building when parties and ide ol ogical bl ocs are sharply defined and p olar­
ized. F or even if a president is elected thanks t o a cr oss-bl oc pact made
before the electi on, the president himself or herself still must come from
one of the bl ocs. With such high stakes in h olding the presidency, simply
helping elect a president via a preelecti on c oaliti on may n ot be sufficient.
Indeed, if the stakes are sufficiently high, it may even become irrati onal t o
f orge cr oss-bl oc c oaliti ons. The dilemma, of c ourse, is that by failing t o
form a c oaliti on, one bl oc's or party's "greater evil" may be elected in a
close three-way race. But in a p olarized envir onment with a d ominant
presidency, even the pr obability of the greater evil am ong one's opp onents
might n ot be as evil as the certainty of l osing the presidency t o any other
p olitical tendency. Thus the dilemma: The m ore that c oalitions are needed,
given ide ol ogical cleavages needing t o be bridged, the m ore irrational it
may be t o f orm them. 20 The Chilean presidential election of 1964, placed in
the c ontext of the preceding and f ollowing electi ons, serves t o illustrate this
p oint.
In 1964 the Christian Dem ocratic candidate, Eduard o Frei, received the
supp ort of the C onservative and Liberal parties, which did n ot present
candidates of their own. While the Radicals ran their own candidate, he
end orsed Frei pri or t o the electi on, as all n ons ocialists c oalesced t o bl ock
Salvad or Allende, the candidate of a leftist c oaliti on. Thus v oters were
presented with a clear ch oice, s ocialism or no, conforming once again t o
the archetype of efficiency in the selecti on of the executive. H owever, the
c osts were apparent in the aftermath. The winner of the electi on himself
represented an ide ol ogical option, and, having had a maj ority electi on,
Frei interpreted his vict ory as a mandate f or Christian Dem ocracy. In an
event unprecedented f or any party in Chile, the Christian Dem ocrats w on a
20 For reasons developed in Appendix A, this dilemma may be especially acute under three­
bloc competition, as in Chile.
202 Presidents and assemblies
majority of seats in a congressional election held shortly after Frei's inaugu­
ration. Frei and his party then pursued a partido unico (single party) strat­
egy, shunning the traditional multipartisan logrolling that had long been
the system's "glue." Each step in reducing the role of logrolling and increas­
ing the powers of the executive raised the stakes of elections, both presiden­
tial (to determine who would have these prerogatives) and congressional
(to determine whether the congress would be disposed to exercise even its
increasingly feeble checks on the president).
The coalition's formation in 1964, while demonstrating the costs of turn­
ing over the head of a coalition to a single representative, especially an
ideological one, was a response to the perceived dangers of not forming
such a coalition in the presence of ideological polarization. In 1958, Arturo
Alessandri won the presidency narrowly against Allende. At 31.6%, this
was the smallest plurality ever in Chile's presidential elections. Allende
received 28.9%. Thus the inability of the parties to forge a common front
on either the left or the right resulted in a narrowly endorsed executive.
Alessandri's presidency was less risky to the established system than either
of the next two, which represented ideological alternatives to the tradi­
tional politics. Indeed, the "nightmare" that the traditional parties and the
Christian Democrats were lucky to escape in 1958 and coalesced to avert in
1964, became reality in 1970, when a socialist executive was elected with
36.6% of the vote in a three-way race.

Efficiency and the rationality of conflict

W hat had happened to the Chilean archetype as the reforms took effect,
concurrent with the emergence of new ideological forces (most particularly
the Christian Democrats, but also extremism on the left)? From the stand­
point of parties as well as of individual legislators, it now became rational
to dispute elections to congress on the basis of support or opposition to the
national goals being articulated in the presidential coalition. For parties it
was rational because the ability of party blocs in congress to exact conces­
sions through legislative logrolling was being diminished. The goals at the
level of locally conducted congressional elections instead would have to be
to facilitate or stymie the president's ability to enact his own goals.
For individual legislators, it was now rational to stress national-level and
ideological conflicts to a greater extent than before simply because there
was little else to offer in exchange for votes. W ith the constitutional re­
forms cutting congressional involvement in budgetary trade-offs and the
ability to engage in locally targeted spending, the payoff for stressing par­
ticularistic concerns in elections was likewise diminished. The nationaliza­
tion of politics was occurring apace in the 1960s and 1970s, and with it
electoral efficiency was increasing. Indeed, in the 1973 midterm elections
for congress, which preceded the coup by only three months, for the first
 Efficiency and inefficiency 203
time all of the parties coalesced into two blocs supporting and opposing the
presidential coalition. Thus a congressional election, as presidential elec­
tions had been earlier, became a great national vote on two general visions
of the direction for Chilean society.
Chile might have been experiencing an incipient two-bloc party system
replacing what had long been a party system divided into three "camps."
Each party maintained its separate ballot identity even in 1973, so we can
not speak of a two-party system, but the parties presented themselves in
two broad and sharply opposed coalitions: for socialism or for capitalism.
Thus, for the first time in modern Chilean history, a congressional election
had provided as efficient a choice as an ideal presidential election. Indeed,
the 1973 congressional election fits the ideal of efficiency better than had
the presidential election in which Allende became president. Allende had
been endorsed by only 36% of the electorate, while his supporters received
44% in 1973 in what was essentially a two-way race (the main opposition
received 52% ).
To be truly efficient, the 1973 election would have to have been capable
of reorienting national policy. Alas, with a president opposed to the con­
gressional majority, holding vastly superior legislative powers, and with
three years remaining in his term, the midterm election could only intensify
polarization and stalemate. W hat Chile needed in 1973, given that Allende
would remain the constitutional president whatever happened in the elec­
tions, was not a great national "referendum" on two competing sociopoliti­
cal visions. That would await the 1976 presidential election. What Chile
needed was an institutional way to resolve a deadlock that had reached
insurmountable proportions. Lacking an effective arena of compromise or
a constitutional "exit" from the crisis, the formerly stable but now chroni­
cally polarized Chilean democracy was destroyed by a military intervention
(to which the Christian Democrats lent initial support) and ensuing dictator­
ship. Thus were prevented any further referenda on policy direction for
more than fifteen years.
The Chilean experience suggests some conclusions about presidential sys­
tems, each of which will be elaborated on in subsequent chapters. First, the
Chilean system was archetypal only in the sense that the logic of separation
of powers suggests entirely separate constituencies for the two branches.
Second, multiparty presidentialism is likely to encounter difficulties if the
basis of these parties is ideological rather than localistic and if the presidency
is very powerful. After the electoral reform of 1958, localistic interests were
undercut. Chile's nonsynchronized terms further compounded the strains on
the regime, as they meant that some presidents (e.g., Frei) had the opportu­
nity to increase their bloc's share of the seats. At other times (e.g., Allende
in 1973) the president's coalition could suffer a rather decisive defeat, yet the
president remained in office. Then, with a powerful president in full control
of the executive, there is a serious risk of deadlock.
204 Presidents and assemblies
As already suggested, the more promising way to bring about efficiency
instead of the way that was attempted in Chile is to weaken the legislative
powers of the presidency while making parties more nationally focused.
The Chilean Christian Democrats, along with the conservative parties,
could have sought such a reform when the deadlock of the Allende years
reached crisis proportions. After all, this was the coalition - and, until
1973, the congress emanating from the same election - that had passed the
1970 reform enhancing presidential powers. 21 Instead, these parties pre­
ferred a military coup, which they expected would right the ship and return
power to them (Valenzuela 1978). It may be hoped that after the experi­
ence of dictatorship in countries such as Brazil, Chile, and the Philippines,
future attempts to address inefficiencies or deadlock will take the approach
of constitutional reform to weaken presidents' powers over legislation (as
well as to strengthen parties), rather than regime breakdown.

C O N C L U S I O N S : I N T E R A C T I O N OF T H E
L E G I S L A T I V E P O W E R S A N D T H E M E T H ODS O F
ELECTING THE TWO BRANCHES

The contrast of the Chilean and Venezuelan systems, on which we have

based much of this chapter's discussion, suggests that the methods of elect­
ing the congress - the electoral system itself, as well as the relative timing
of elections to the two branches - are important to whether or not the
branches are consistent in their outlook. W hile consistency in outlook im­
plies a violation of the presidential archetype, an important dilemma of
presidential democracy presents itself: How can we ensure either that there
is harmony between the branches or else that, when the branches differ,
there are institutional mechanisms for reaching compromise in order to
avert interbranch crises? If the presidency and congress are controlled by
different parties or blocs, how can the efficiency of presidential elections be
maintained without giving the president such powers as to render the con­
gress ineffective as a legislature? Such are the choices of constitutional
design for presidential systems.
The Venezuelan system at first seems to suggest one answer to this
question: Endow the congress with a national focus by making party and
national (rather than local) policy the basis of its own election. The purest
way to do this would be to make the congress, like the president, elected
21 It is, of course, also the congress that selected Allende instead of the traditional conserva­
tive (and ex-president) Alessandri, since the popular vote was not decisive. Because the
Christian Democrats, whose candidate placed third, and the right had been engaged for
years by 1970 in a battle for hegemony over the nonsocialist portion of the electorate, the
Christian Democrats ironically considered a rightist presidency as worse for their political
interests than the alternative of a socialist executive to which they could - and did - lead
the principal congressional and electoral opposition.
 Efficiency and inefficiency 205
from a single, nationwide, constituency. No presidential system has ever
gone to this extreme,22 but one, Nicaragua in 1984, has done the functional
equivalent, using districts but allocating seats left over after district alloca­
tion on a nationwide basis (Taagepera and Shugart 1989). 23
A presidential system in which the congress would be elected from one
national constituency is far from the inefficient secret of the Chilean case at
mid-century. It also would violate much of what Madison saw as the advan­
tage of the separation of powers - the encouragement of checks on one
branch by the other by having different branches respond to different
constituencies. The same can be said for the Venezuelan and other similar
systems, in which, as a result of the party lists and concurrence of presiden­
tial and congressional elections, the parties that provide the principal com­
petition in congressional elections are primarily vehicles for expressing the
coattails of the presidential candidates.
Denying the president constitutionally granted legislative powers, as in
Venezuela, obviates much of the advantage of presidentialism, as originally
intended. In such a system the policy options endorsed in congressional
elections are sure to be translated into policy output. Of course, if presi­
dent and congress are of different political makeup, presidential elections
are then rendered inefficient, because the one office that is responsible to
the whole nation is unable to implement or credibly bargain on behalf of a
policy program. In Venezuela, efficiency has prevailed only because the
electoral system (party lists) and the concurrence of elections have gener­
ally meant that the president's party has also been the largest party in
congress. Still, the greater success of Venezuela (and Costa Rica) relative
to Brazil, Chile, and Colombia, suggests the following combination of
institutional features may be most conducive to effective presidentialism:
• A president elected by plurality, having limited legislative powers. (We suggest a
veto that requires more than a simple majority to override. )
• A party-list electoral system (but not with such tremendous control b y leadership
over elections as that of Venezuela through 1988).
• Concurrent elections, or, at least, synchronized terms such that there are no
midterm elections as there were in Chile.
We elaborate on these themes in the next chapters, considering espe­
cially variations on the method of presidential election and going into much
greater depth on the subject of the consequences of different electoral
cycles.
22 At least two Latin American presidential systems, Ecuador and Guatemala, use a rather
halfhearted way of providing some members of the congress with a nationwide constitu­
ency by allowing voters to cast both a ballot for district representative and one for the
national district, with seats in local and national districts allocated separately.
23 In Nicaragua, the congress was so elected in 1984 concurrently with the president, but this
electoral system was changed before the 1990 election. Namibia uses a nationwide PR
allocation and concurrent elections but it is a president-parliamentary regime.
 10
Elector al rules and the p arty system

In Chapter 9, we n oted that Chile f or many years c ombined a multiparty

c ongress and, ordinarily, tw o-candidate c ompetiti on f or the presidency. In
this chapter we generalize this observation with analysis of elect oral rules.
Students of elect oral systems have l ong n oted that plurality electi ons in one­
seat districts tend t o p olarize electoral ch oices ar ound tw o principal contes­
tants, while systems of pr op ortional representation are generally ass ociated
with multiparty systems. This chapter c onsiders the implications of different
electoral rules for party systems in presidential dem ocracies. We have indi­
cated already (in Chapter 2) that presidentialism d oes n ot by definiti on
require that there be only a one-seat executive; h owever, the prevailing
practice is t o use districts of magnitude one (M = l) for the presidency. A
district of M = l requires that the elect oral system be s ome f orm of plurality
or maj ority. Empirically plurality rule is m ore c omm on, alth ough maj ority
runoff systems have been widely ad opted among newer dem ocracies since
ar ound 1980 in Latin America, and 1989 in Europe and Africa.
As indicated in Chapter 9, systems of prop orti onal representati on, in
which the district magnitude must by definiti on be greater than one (M >
1), are typical f or assemblies even in presidential and premier-presidential
systems. Before presenting empirical data on the effects of these elect oral
systems, we sh ould review the m ost widely discussed statement on this
matter, kn own as "Duverger's law" or "Duverger's rule."

D U VERGER' S R U LE AND THE NU MBER OF PARTIES

When Riker (1982b) wanted t o dem onstrate that p olitical science had a
history of cumulati on of kn owledge like the "hard" sciences, he turned t o
the set of ref ormulations kn own as Duverger's law. Alth ough Riker identi­
fied formulations pri or t o Duverger's own, it is Duverger (1951, 1954) wh o
n ormally gets the credit f or having made questi ons ab out the influence of
elect oral rules on party systems a major concern of p olitical science.
Through repeated ref ormulati ons, a pair of prop ositions kn own as Duver­
ger's law and Duverger's hyp othesis and c ollectively referred t o herein as
Duverger's rule, have emerged. We shall briefly trace the origin of these
prop ositi ons, which may be stated thus:
 Electoral rules and the party system 207
1 . One-seat districts with plurality rule tend to reduce the number of parties to two
(law); and
2. multi-seat districts with proportional representation tend to be associated with
more than two parties (hypothesis).

The wording here is close to that employed in two recent discussions of

the Duverger effects, Taagepera and Grofman ( 1985) and Taagepara and
Shugart ( 1989). 1 Recently Duverger has emphasized that he never meant to
imply that proportional electoral systems inevitably produce multiparty
systems, but he still indicates that there is a "multiplying effect" of PR that
manifests itself mainly in a proliferation of very small parties. National and
social factors are more important, he says; electoral rules act "usually only
as an accelerator or a brake" (Duverger 1986:73). This less forceful (per­
haps even too weak) wording of the "multiplying" effects of PR, as op­
posed to the reductive effects of one-seat plurality, is what for Riker justi­
fied the use of the term Duverger's hypothesis, as distinct from the law.
The usual understanding of Duverger's rule has suffered from a narrow­
ness of scope in the sense that studies of electoral systems have tended to
assume implicitly that the only institutions that mattered were electoral
rules for assemblies, taken in isolation from other institutions. Such isola­
tion of the assembly from the process of executive formation may distort
our understanding of the number of parties even in parliamentary systems;
it especially is certain to do so in the case of presidential systems. Particu­
larly when presidential and congressional elections are held at the same
time, the presidential election imposes a single-seat nationwide district
over the assembly districts. If the president is elected by plurality, the
presidential election falls under Duverger's law, which suggests a reduction
in the number of parties to two. The congressional election, if it is held in
multi-seat districts using proportional representation (PR) , suggests the
possibility of a multiparty system, according to Duverger's hypothesis. As
we shall see, the outcome may go either way, depending to a considerable
degree on the relative timing of congressional elections with respect to
presidential elections. Thus this chapter and the next one develop the
notion that two variables are critical in shaping the nature of the party
system: electoral rules (particularly district magnitude) and electoral cycles.
District magnitude is the decisive feature in electoral rules (Rae 1967;
Sartori 1968; Taagepera and Shugart 1989). District magnitude, M, has
1 The two sources just cited add a clause to the hypothesis suggesting that the correlation of
multipartism with M > 1 is conditioned on there being political issues sufficiently numerous
to provide a basis for such parties. We no longer see this qualifier as necessary, as the
"exceptions" no longer seem so exceptional. For example, Austria's party system has been
dominated by two parties despite PR; yet Austria's effective magnitude before 1970 was
very low. Thus, while one might have expected around three parties instead of almost
exactly two, a low number of parties despite PR should not be seen as deviant when M is
very low. Since 1970, when effective M was significantly increased (Taagepera and Shugart
1989), the dominance of the two big parties has been weakening, making Austria fairly
consistent with Duverger's hypothesis.
208 Presidents and assemblies
been recognized at least since Hogan (1945) as the most important factor in
proportionality. Since disproportionality is what Duverger hypothesized to
be the factor that would induce underrepresented parties to go out of
business, the number of parties in a district must bear some relation to
district magnitude. 2
More recent quantitative analysis has looked at the relation between
the number of parties and the number of seats across the whole range of
observed district magnitudes. Taagepera and Shugart (1989) have done
this using mainly nationwide data, while Shugart (1988) and Taagepera
and Shugart (1992a, b) analyzed district-level data both within and across
nations.
The other principal variable that we shall explore is the electoral cycle,
simply meaning the timing of elections to the two branches, executive and
assembly, in relation to one another. In some systems we find that all
members of the assembly are elected at the same time as the president, that
is, concurrently. Other systems employ terms of different length for the two
branches and do not let election dates for president and congress coincide.
We simply call such an electoral cycle nonconcurrent.
Do these and other institutional features of presidential and premier­
presidential democracy make a difference? How do their effects compare
with results of parliamentary democracies? Since one of our primary con­
cerns in analyzing representative democracies is with the degree of repre­
sentativeness, which in turn may be expected to be fairly closely related to
the number of parties, we begin in this chapter with a survey of the relation
between different institutional designs and the "effective" number of par­
ties. We then proceed to a more systematic examination of data on the
assembly party system that conforms to various electoral cycles. Chapter 1 1
then will seek t o develop models of the number of parties in presidential
systems, drawing upon earlier studies that concerned themselves princi­
pally with parliamentary systems.

M E T H O D OF P R E S I D E N T I A L E L E C T I O N

The first matter o f institutional choice with which we concern ourselves is

how to elect the president. One basic breakdown, mentioned already in
Chapter 1 , is between those systems in which there is an electoral college
and those in which the voters vote directly for presidential candidates.
2 This means not that the particular type of allocation formula (such as simple quota or any of
the various "highest average" formulas) does not matter, but that the relative emphasis on
allocation formula versus district magnitude has all too often been reversed. Especially
among reformers and designers of electoral systems, allocation rules have often attracted
far more attention than they deserve, given their more limited impact on proportionality
and the number of parties, compared to district magnitude. With M = I, as we shall see, the
precise electoral rule (plurality versus majority, for example) does indeed make an apprecia­
ble difference. In general, formula matters more with low M, less as M increases.
 Electoral rules and the party system 209
However, in the modern era there is a more important distinction: between
those direct elections that are by plurality and those that are by majority
runoff, plus those in which the popular vote is not necessarily decisive. As
we shall see, electoral methods in the latter category may function as if they
were plurality, unless partisan (as opposed to presidential-candidate) vot­
ing is the primary criterion.

Plurality rule
In several presidential systems the president is elected by plurality, in which
the winner is simply the candidate with the most votes (sometimes referred
to as the relative majority). Among the countries in which plurality is used
to elect the president are the Dominican Republic, Nicaragua, the Philip­
pines, and Venezuela. The average share of votes for the winner in straight­
forward plurality elections is 49.7%, and for the runner-up it is 34.9%, as
seen in Table 10.1. Thus plurality elections have a tendency toward build­
ing two large blocs that together encompass around 85% of the vote, on
average. Such a finding is consistent with Duverger's law. The following
two features must then be taken as characteristic of plurality presidential
elections:
1 . the tendency for the formation of a broad coalition behind the front-runner;
2. the tendency to coalesce the opposition behind one principal challenger.

In Costa Rica, the constitution prescribes a runoff in the event that no

candidate obtains 40% of the popular vote. We shall, however, consider
this system to be effectively plurality, for two reasons. First, no runoff has
ever been required. Second, the incentive to build a coalition large enough
to surpass 40% is not so different from the incentive to build a coalition
that is simply the largest, as in straight plurality, for reasons developed
below. That this is so may be seen in Table 10.1, in which the shares of the
vote for the first and second candidates are displayed for several presiden­
tial election formats. We shall have more to say about this 40% rule and its
implications later on.
In Costa Rica the results look like pure plurality, or perhaps even
stronger on the two tendencies of plurality just mentioned. The median
vote share in Costa Rica is actually 50.3%, while that of the runner-up is
itself more than 40%. As discussed in Appendix A, the 40% provision may
be seen as a threshold that must be attained for a first-round victory. Pure
plurality sets no threshold, meaning victory to the largest vote-winner,
regardless of the actual percentage obtained. As the threshold is raised, so
are the costs of failing to coalesce, unless the threshold is raised so high as
to encourage many "spoiler" candidacies in order to influence the runoff.
As we shall see next, the 50% threshold - the majority runoff system - is
an example of a threshold too high to induce coalescence in most cases.
210 Presidents and assemblies

Majority rule
Most of the remaining presidential and premier-presidential systems use
majority elections in two rounds; if no candidate receives a majority of
votes (in some systems, a majority of all votes cast, including null or blank
ballots), a second round is held to determine the winner. 3 Table 10.1 shows
that the average share of votes for the candidate placing first in the prelimi­
nary round is 39.5% , about 10% lower than for plurality systems, including
the Costa Rican variant. The second-place candidate, on average, falls
considerably farther behind the first finisher (in the first round), with only
25.1%. More than 35% of the vote, on average, is divided among other
candidates. Thus majority runoff can actually discourage coalescence of the
opposing forces, even though only a single seat is at stake. These forces
instead run separately with the intention of either (1) placing second in
order to make the runoff themselves and thus attract the support of others
who failed, or (2) enhancing their own bargaining position with one of the
two candidates in the runoff, perhaps exchanging support for policy or
office concessions. W ith the lower threshold, either the same leading candi­
date who, on average, receives just under 40% , or else another candidate,
would be likely to win in the first round. The more attainable 40% thresh­
old, for example, would ordinarily give supporters of potential minor candi­
dates the incentive to join with a larger coalition rather than seek to influ­
ence an expected runoff.

Nondecisive popular vote

Some other systems, including those of Chile (before 1973), Argentina,
and the United States, function much like plurality in their encouragement
of large coalitions. However, each of these systems uses a different method
in which the popular vote is mediated through an electoral college or the
congress, at least in the absence of a decisive majority of the popular vote.
We also shall see that one case, Finland, has an electoral college that is
elected on a partisan basis rather than on a candidate basis. In that system,
much as in the Bolivian case discussed in Chapter 4, we do not see coales­
cence behind two front-runners.

Congressional ratification. In the system of presidential election in Chile

before 1973, as discussed in Chapter 5, the congress chose the president in
the event that no candidate received a majority of popular votes. However,
3 There are two exceptions to this rule. The presidents of Sri Lanka and Ireland are elected
by a method in which voters may rank the candidates on the ballot. If there is no majority of
first preferences, second (and, if necessary, later) preferences are added, until a candidate
obtains a majority of votes. The rule is usually called the "alternative vote" and is identical
to the single transferable vote in one-seat districts.
 Electoral rules and the party system 211

Table 10. 1 . Median percentage of votes for two highest-finishing candidates in

presidential elections, according to different electoral methods (first round in case
of majority system)

Number of % votes % votes

Coun!!l'. and time eeriod* elections 1st 2nd
Pure plurality
Brazil (1945-60) 4 48.5 3 1 .6
Colombia ( 1930, 1942-46, 8 47.4 34.2
1974-90)
Dominican Republic
( 1962, 1982, 1990) 3 46.5 33.9
Nicaragua (1990) 1 54.7 40.8
Philippines ( 1946-69) 7 55.0 39.5
Venezuela (1958-88) 7 49.2 34.6
Mean of all plurality elections 30 49.7 34.9
Effective plurality
Rynoff if Isiss than 40% 11Iurnlit:t
Costa Rica (1953-86) 9 50.3 4 1 .2
Majority (two rounds)
Brazil (1989) I 3 1 .5 17.7
Chile ( 1989) 1 5 1.5 26. 1
Ecuador (1978-88) 3 28.7 23.9
France ( 1965-88) 5 36. 1 29.4
Peru (1980-90) 3 45.4 24.7
Poland (1990) 1 40.0 23. 1
Portugal (1976-86) 3 56.4 25.4
Mean of all majority elections 17 39.5 25. 1
Nondecisive popular vote
C2n i:rsissi2nal ssilsi�liQll if DQ l!Q£ll.lli!.I !Ililiori1)'.
Chile (193 1 -70) 9 50.5 34.8
EIWm:al CQll!ll:!l
Argentina (1983, 1989) 2 50.6 36.9
United States (1824-1988) 42 5 1 .6 43.3

• Excluded are elections that were boycotted by a potentially significant political force and
those for which presidential and assembly votes were fused (straight-ticket).

in most elections, coalitions had been successfully structured before the

election such that one candidate received a majority or very close to it.
W hen congress entered the selection process, it was constitutionally bound
to make its choice from among the top two contenders. The median shares
of those two candidates were 50.5% and 34.8% (Table 10.1), hardly differ­
ent from the pure plurality cases.
212 Presidents and assemblies
Electoral college. In two presidential systems there is an electoral college
that functions more or less like plurality: Argentina and the United States.
These institutions were expressly designed to be nondeliberative - indeed,
the U.S. electoral college does not actually even meet as a body. In the
United States if there is no candidate obtaining a majority of electors at the
first count, the decision goes to the House of Representatives. If this were
the usual occurrence, the United States might be institutionally similar to
either Bolivia or Chile, except that the U.S. House votes for president on
the basis of one vote for each state's delegation. However, the allocation of
electors by plurality rule in each state (and therefore in multi-seat districts,
which with list plurality are especially favorable to bigness), and the emer­
gence of political parties as vehicles for electing presidents has meant that
the United States has functioned effectively as plurality throughout nearly
all of its history, encouraging broad coalitions that, on average , have gar­
nered 51.6% for the front-runner and 43.3% for the runner-up. 4 In Argen­
tina, similarly, if there is no majority of the popular vote , the decision rests
with an electoral college that is chosen concurrently with the presidential
race and represents the states of the Argentine federation. There, too,
winning candidates have thus far won majorities of the popular vote.

The Finnish case. The Finnish case demonstrates that an electoral college
can be designed in such a way as to make presidential elections more
closely resemble the assembly selection method of systems like Bolivia's
rather than popular presidential elections. Finland, a premier-presidential
system, had not used truly popular elections for president until 1988. In­
stead, voters voted for parties (or more precisely, for one candidate within
a party list) by a means similar to that used to elect the parliament. This
system gave priority to parties. Besides having an electoral system based on
parties, the Finnish electoral college differs from those of the United States
and Argentina in that it is deliberative, taking multiple ballots when there
is no majority on the first count. Additionally, before 1988, electors and
parties did not necessarily pledge themselves to a particular presidential
4 There are some potential problems with the U.S. electoral college that could diminish its
plurality character. The best known is the possibility that the electoral college might reverse
the plurality verdict of the popular vote. However, the cause of this, which has happened
only twice, lies in the state-by-state nature of the counting of electoral pluralities. Taking
pluralities on this basis, and weighting them by states' populations, is hardly inconsistent
with U.S. federalism and thus does not contradict the plurality character of the overall
system. Potentially more serious, but within the realm of the theoretical, are the possibili­
ties of sufficient "faithless electors" to overturn the popular vote , the lack of a mandate in
federal law for either "faithful" electors or their allocation by plurality rule (it is the choice
of the states), and the possibility of victory by unpledged electors in some states in order to
force the brokering of an election in the House of Representatives, sitting as a second
electoral college in which each state has only one vote. See Stedman (1968) and Peirce and
Longley ( 1981).
 Electoral rules and the party system 213
candidate; now they must do so (Mackie and Rose 1991:109). The proce­
dure, established in 1925, was a compromise between those who favored
parliamentary selection of the president (as in the present-day systems of
Greece and Israel) and those who favored a direct, popular election. Given
its party-centered character, it was not much different from election in
parliament; presidential candidates ordinarily did not even tour the country
and campaign, according to Markku Suksi (personal communication).
Even so, the restriction of the body chosen in these elections to the sole
function of selecting a president did encourage electoral outcomes that
differed somewhat from the normal pattern in Finnish parliamentary elec­
tions. Parties could still perform better if identified with a popular candi­
date, in order to be in a position to "deliver" a sizable share of votes to the
eventual winner. This process of a party's being electorally rewarded for
having a popular candidate may be demonstrated by the last election under
this format, that of 1982. In that election, the Social Democrats received
43% of the vote when Mauno Koivisto was their candidate, even though
this party's share of the vote in parliamentary elections has normally been
less than 30%.
In 1988, there was a modification of the Finnish system. In that election
voters cast a vote directly for a presidential candidate in addition to the party
vote. The law provides that if no candidate receives a majority of personal
votes, the electoral college will make the decision. In 1988, Koivisto received
47.9% of the personal votes, even though his Social Democratic party re­
ceived only 39.3% in the concurrent contest for electors. (Three other par­
ties endorsing Koivisto combined for 7. 3 % . ) The electoral college ratified
the electorate's plurality choice. W ith this new personal vote, Koivisto thus
ran ahead of his own party in the concurrent party vote and even farther
ahead of the party's showing in the previous parliamentary elections (24. 1 %
in 1987). Thus the election took on one basic characteristic of plurality: the
formation of a broad following behind a candidate in order to ensure a first
place finish, although the opposition remained fragmented, resembling
more the field of majority runoff systems.

Which method to use?

We find the plurality rule appealing because it is more likely than majority
runoff to give voters an efficient choice between the candidates of two broad
coalitions. That this is so is supported by the data in Table 10. 1. Majority
elections, on the other hand, may discourage long-term consolidation of a
party system. They may even encourage fragmentation. The appeal of major­
ity runoff to recent constitution designers has been enhanced by its results in
France, where a fragmented and multidimensional party system under the
Fourth Republic rather quickly was transformed into stable two-bloc (left­
right) competition after the initiation of presidential elections by majority
214 Presidents and assemblies
runoff in 1965 under the Fifth Republic. Let us consider the experience of
France and whether or not it might be generalizable.

Two-bloc competition in France. One should be careful about generalizing

from the French experience. In France the two-bloc format of the party
system has been aided by the use of majority runoff not only in presidential
elections but also in elections for the assembly (except in 1986). Addition­
ally, it has been aided by premier-presidentialism. Neither factor is present
in many systems elsewhere, especially in Latin America, where majority
runoff has recently been adopted for presidential elections.
The use of the majority runoff for French assembly as well as presidential
elections aided the development of the two-bloc system because it allowed
parties to make cross-district deals between rounds. Parties have been
encouraged to withdraw candidates from some districts and support the
candidates of other parties in exchange for reciprocal deals in other dis­
tricts. Obviously this king of bargaining is not feasible where only one seat
is at stake in the election, as with a presidential race. In many other cases,
proportional representation is used for assembly elections, eliminating the
French incentive for alliance formation.
The two-bloc system in France has been further aided by premier­
presidentialism. Since the cabinet in France is dependent on the confidence
of a majority of the assembly and not on the president, alliances are impor­
tant in order to sustain a government or a credible governing alternative.
Where the cabinet serves at the pleasure of the president and cannot be
removed and reconstituted by the parliamentary majority through a vote of
no-confidence, parties contesting congressional seats are not compelled to
offer a formula for putting together an executive.
For these reasons, the majority runoff presidential electoral rule alone
should not be viewed as being conducive to stable two-bloc competition.
We shall leave aside the question of whether presidential systems in Latin
America and elsewhere should adopt majority runoff for their assemblies
as well as for president in an effort to institutionalize a two-bloc party
system. One wonders whether party leaders in those systems would be
willing to sustain the substantial deviations from proportionality that the
system has entailed in France. 5 We wish to caution that both majority
runoff assembly elections and premier-presidentialism have been crucial to
the pattern of competition in France.

Proliferation of candidates under majority runoff Peru is an example of a

party system in which the use of majority runoff for presidential elections
may be blamed for a proliferation of candidates. What had been a two-bloc
5 One of the newer systems in the Americas has adopted majority runoff for assembly as well
as for president, albeit a constitutionally powerless president: Haiti.
 Electoral rules and the party system 215
party system before military intervention in 19686 led to two consecutive
presidential elections in the 1980s that were won by first-round majorities
of valid votes cast. However, by 1990 the majority runoff rules encouraged
two "outsiders" to enter the first round as challengers to the established
party system. W ith a decisive plurality election, such proliferation would
have been far less likely. An "outsider" like Alberto Fujimori, the eventual
winner, could not have appeared as a viable candidate under plurality rule.
Fujimori's strategy centered around scoring an upset second-place finish in
the first round in order to appear as a second-round alternative to the
acknowledged front-runner, Mario Vargas Llosa, also an "outsider" to the
established party system (albeit one who, unlike Fujimori, had forged a
pact before the first round with some of the established parties). 7
A similar first-round story may be told about the 1990 presidential elec­
tion in Poland. A virtual unknown, Stanislaw Tyminski, finished a surpris­
ing second, with 23. 1 % , to Lech Walesa's 40.0%. As with Fujimori,
Tyminski's strategy for winning centered around scoring an upset second
place in the first round and then combining the anti-Walesa vote in order to
win in the second round. Walesa, however, won an overwhelming victory in
the runoff. W ith a plurality election, it is nearly inconceivable that Tymin­
ski could have been anything but a fringe also-ran, while Walesa would
likely have won with a bare majority (or perhaps slightly less than an
absolute majority) instead of the large "mandate" he could claim on the
basis of the second round.

Majority runoff and efficiency. If we desire presidential elections for their

ability to allow voters to identify the most likely governing options being
presented to them in the initial round and make an efficient choice, then
majority runoff rules are less desirable than plurality rule. The basic weak­
ness of the majority format is that its outcome depends on first-round
contingencies. The principal contingency is the number of candidates en­
couraged to run in the first round by the fact that the first round is not
decisive. Many such candidates would not run in the event of a plurality
election, or would receive fewer votes if they did. The formation of a
6 The electoral system for president under the 1933 constitution was that the popular vote
would be decisive if the leading candidate won at least one third of the vote. Otherwise the
congress, elected concurrently, would choose the president from among the top three
contenders.
7 See Cameron (1991) for a discussion of Vargas Llosa's "blackmailing" of his partisan allies
by threatening to run his presidential campaign as an independent. The threats to resign
from the alliance that Cameron describes would not have been so credible had the presiden­
tial election not been a majority runoff system. Because majority runoff opens up the
possibility that a candidate finishing second initially can then go on to win, Vargas Llosa was
able credibly to point to the possibility that he could win without a preelection coalition with
established parties. Indeed, that is precisely how the election was won, but not by Vargas
Llosa!
216 Presidents and assemblies
widely endorsed executive, and one with a sizable bloc of congressional
supporters, may be compromised if many candidates enter the field in
hopes of forcing one of the two who pass to the second round to make a
deal in exchange for support.
The distinction between plurality and majority concerns in part the level
at which strategic bargaining occurs. Some strategic bargaining is desirable
when a single seat is being filled, in order to maximize the ability of voters
to identify the alternative potential executives. To maximize their chances
of victory, party leaders contesting a plurality presidential election in a
multiparty environment must seek to build coalitions before the election.
In most cases this means that the principal contenders will be two relatively
moderate candidates. On the other hand, because majority runoff leaves
much of the bargaining among parties to the period between rounds, the
efficiency of presidential elections is reduced.
W ith majority runoff, some candidates will enter primarily to force a
second round, as with Fujimori and Tyminski. That is, parties under this
electoral method specialize (Schlesinger and Schlesinger 1991). Some seek
mainly first-choice votes, while others seek to win as second-choice candi­
dates. If the largest party (or most popular candidate) also is widely dis­
liked, such a candidate may be defeated in a second round by whatever
candidate finishes second. The most striking example of this effect was in
Portugal in 1986, when Diego Freitas do Amaral won 46.6% to Mario
Soares's 25.4% (the next candidate had 20.9% ). Despite his wide and
nearly decisive first-round margin, Freitas lost the runoff, 51.2% to 48.8% .
Thus the outcome of majority runoff elections may frequently hinge upon
which candidate is capable of beating out other contenders for the second
slot and on whether or not this candidate can form alliances with defeated
candidates between rounds. Unless the second-place candidate in a typi­
cally divided field can be predicted reliably beforehand, identifiability and
efficiency suffer, compared to a plurality race in which parties have the
incentive to offer voters broader preelection coalitions.

A compromise: the double complement rule. In making a case for plurality,

however, we also understand that the results of Table 10.1 show tenden­
cies, not absolutes. It is the occasional president endorsed by less than 40%
of the electorate in direct plurality elections that stirs fears in those who
have opted recently for majority runoff rules instead of plurality. One such
occasion might be a regime's last, as in Chile, although one should not
exaggerate the danger; several regimes have survived such presidencies. If
one wants to avoid the second stage of an electoral college, 8 the Costa
8 Perhaps because of their inherent risks of reversing the nationwide plurality choice of the
electorate, electoral colleges enjoy few favorable mentions nowadays. Other than Argentina,
which reestablished its electoral college with the return to democracy in 1983, we know of no
new regime in the twentieth century that has adopted such a format since Finland in 1925.
 Electoral rules and the party system 217
Rican rule looks like a very good compromise. The 40% threshold may be
likened to a "safety valve," encouraging broad coalitions, like plurality, but
with a built-in mechanism that prevents a winner with very limited endorse­
ment in a multicandidate race.
Given the successful experience in Costa Rica, it is thus surprising that
other systems have not incorporated such a rule. The 40% provision is
often advocated by those in the United States who wish to abolish the
electoral college (Peirce and Longley 1981). Generally, however, the prefer­
ence in revised electoral procedures - principally in the new democracies
of Latin America9 - has been for a runoff in the event that no candidate
obtains more than 50% of the vote. As we saw in Table 10.1, majority
formats, while appearing to prevent their designers' fears of narrowly en­
dorsed presidencies like Allende's, actually result in considerably lower
average shares in the first round for eventual winners owing to a greater
number of initial competitors. In Appendix A we discuss the dynamics of
coalition building under plurality and why 40% emerges as a compromise
between straight plurality and majority.
The 40% rule does have one potential drawback, however: A winner
could still emerge from a very closely fought contest with a questionable
mandate, for example, with 45% against a runner-up with 44%. Perhaps
there might be a better compromise that is halfway between plurality and
majority runoff. The derivation of such a rule is shown briefly here. An
extended discussion may be found in Shugart and Taagepera (1992). The
condition that defines plurality rule is that the vote share of the leading
party (v 1 ) must be greater than the vote share of the second largest party
(v2 ):

while the first-round condition that defines majority runoff is:

V1 > 50%.
The arithmetic average of the two conditions is
V1 > (1/2)V2 + 25%.
We can rearrange this criterion and express it in terms of each party's
difference from majority. We then have a double complement rule, which
requires the following:
9 The trend toward majority runoff is particularly noticeable among cases of redemocrati­
zation. In their returns to competitive elections since around 1980, the following countries
have adopted majority runoff: Brazil, Chile, Ecuador, El Salvador, Guatemala, and Peru.
Most previously used plurality or required congressional confirmation in the event of no
electoral majority. In Finland, too, majority runoff rules are under consideration, while
most of the new systems with directly elected presidents in East Europe and Africa have
similarly opted for majority runoff.
218 Presidents and assemblies
50% - Vz > 2 (50% - V 1),
This rule thus stipulates that the front-runner wins at the first round if the
shortfall of the runner-up from a majority of votes is more than double the
leading candidate's shortfall from a majority. If the front-runner does not
meet this requirement (or win a first-round majority), then there is a runoff
between the top two contenders. We find this rule appealing for its simplic­
ity and its discouragement of fragmentation. 10 We suspect that the double
complement rule would be appropriate for most situations, except perhaps
for electing a unitary presidency in polities that are characterized by signifi­
cant ethnic or other ascriptive cleavages. For such societies, one rule that
has been attempted, although it is very complex, is the rule of concurrent
pluralities used in Nigeria.

Concurrent pluralities and the problem ofplural societies. There is one final
variant of plurality rule that encourages the construction of broad coali­
tions before a presidential election. A presidency may be filled by a system
in which pluralities must be won concurrently in several regions of the
country rather than in the country as a whole. The U.S. electoral college is
effectively a variant of this form, since to be elected without recourse to the
House of Representatives, a candidate must win pluralities in states that
amount to a majority of total number of electoral votes. Since the electoral
college overrepresents small states compared to a nationwide direct elec­
tion, the accumulation of pluralities in a series of relatively smaller states is
one way to win the presidency.
Another form of concurrent pluralities has been used in Nigeria. The
winner must win a nationwide plurality plus at least 25% of the vote in no
fewer than two-third of the states, which were themselves drawn mainly
along ethnic lines. If no candidate met this requirement, a second election
would take place between the one with the overall plurality and the one
among the remaining candidates with the most votes in the highest number
of states (Nigerian constitution, Article 126). In this way the designers
intended to ensure the election as president someone who could win broad
appeal across ethnic groups. In this task, according to Horowitz (1990),
"they succeeded." The complexities of this rule and the possibility that no
candidate would meet its criteria (in which case a new round of voting must
begin, with new candidates) leave us skeptical about its viability. Indeed, in
his discussion of the desirability of presidentialism for South Africa, Horo-
10 Since the rule is derived as an average of plurality and majority criteria, we might expect
the vote shares of the typical two leading candidates also to be averages of the plurality and
majority cases: 44.6% and 30.4%. If so, this would suggest that a runoff would be needed
about half the time, rather than in nearly all elections, as is the case with majority runoff.
Moreover, the figures suggest that if a runoff were needed, ordinarily the two candidates
squaring off would be considerably farther ahead of the rest of the field than is often the
case under majority runoff.
 Electoral rules and the party system 219
witz (1991) advocates the Irish and Sri Lankan alternative vote rather than
the Nigerian concurrent pluralities. W hile we sympathize with Horowitz's
claim that a presidency, given a proper method of election, can be an
appropriate conflict-regulating institution for an ethnically divided society,
we are not convinced of the utility of either alternative vote or concurrent
pluralities. The alternative vote is not in practice much different from
plurality (see Lijphart 1991), while the rule of concurrent pluralities may
be unwieldy, whatever its intellectual merits. For these reasons, we prefer
for plural societies the two-person or three-person collegial presidency that
we proposed in Chapter 5, with the double complement rule being used to
determine the winning ticket.

T H E N U M B E R OF COMPETITORS

We have seen in Table 10.1 some variation among electoral methods in the
vote shares of the two largest presidential contenders. How is the method
connected to the number of contenders? Table 10.2 shows the effective
number of presidential candidates (Np) by country, grouped according to
general categories of plurality and majority, as well as subcategories con­
cerning the rules and timing of assembly elections. The table also shows the
effective number of parties in assembly elections (NJ. For presidential
candidates the dividing line between plurality and majority is clear, but of
even more interest for our purposes are the differences within plurality
countries.
Among the empirical configurations we find three major categories: (1)
those in which the president is elected by plurality and the assembly by
proportional representation in multi-seat (M> l) districts; (2) those in
which the president and congress are both elected by plurality; and (3)
those in which the president is elected by majority. Each group may also be
further subdivided according to electoral cycle. In the third category, only
France uses majority runoff for the assembly as well as for president, while
the others use PR. 1 1 Because both majority runoff and PR are methods
associated with multiparty systems (Duverger 1954), this distinction will
not concern us. Indeed, of greatest interest are the countries that use both
plurality and PR, as these are employing contradictory electoral systems,
one that (as confirmed in Table 10.2) is associated with bipartism and one
that is associated with multipartism.

Plurality and PR. In this first category, the variable that most clearly sepa­
rates our cases is the electoral cycle. Table 10.2 breaks down these cases by
those in which there are concurrent elections and those in which elections
are ..!! nonconcurrent.
1 1 France itself used PR in 1986.
220 Presidents and assemblies

Table 10.2. Median effective number of presidential candidates (Np) and median
effective number of parties obtaining votes in congressional elections (NJ, accord­
ing to different institutional formats

Country N N

Regimes using (effective) plurality for president

Assembl)'. PR, concurr�nt
Argentina 2.6 3.0
midterm 3.2
Brazil ( 1945-50) 2.5 3.4*
Costa Rica 2.2 2.6
Dominican Republic 2.7 2.9
Nicaragua 2.2 2.2
Venezuela 2.6 3.4
Mean of all (excluding
midterm elections) 2.5 3.1
Ass�mbl)'. PR, all noncon1.;1.1m:nt
Brazil ( 1955-62) 2.7 4.5*
Chile (1933-73) 2.7 6.9
Colombia 2.7 2. 1
Mean of all 2.7 5.0
Assembl)'. 12Iuralit)'., rnn!.1.11Ent
Philippines 2.0 2. 1
United States 2.1 2. 1
Regimes using majority for president
Assembl)'. concyrrent with first round
Chile (1 989) 2.6 2.6
Ecuador 5.0 7.4
midterm 10.8
Peru 3.4 3.4
Mean of all (excluding
midterm elections) 4.1 5.8
Assembl)'. nonconcurrent
Brazil (1 989-90) 5.3 8.7*
France 4.0 4.8
Portugal 2.3 4.0
Mean of all 3.5 4.3

*Seat-winning parties; figures for votes (which would be higher) were unobtainable.

When PR congressional elections are concurrent with the plurality presi­

dential election, average NP = 2.5, while average Nv = 3. 1. The typical
assembly party system for these cases has two parties sharing at least 75%
of the vote, with the third largest party having about 5% to 10% . The range
of national values for Nv is 2.2 to 3.4. Not surprisingly, when we consider
the effects of magnitude, M, this range is considerably greater than that
 Electoral rules and the party system 221
seen for Np, for which M is always 1. Venezuela has the highest N/2 and is
also the case with the highest effective M, owing to a provision that makes
representation for very small parties quite easy. A party needs only about
0.55% of the nationwide votes, regardless of whether or not it wins any
district-level seats (Taagepera and Shugart 1989:269; Shugart 1992a). 1 3
The one case that uses midterm in addition to concurrent elections,
Argentina, does not differ in any observable way from the others, in which
only concurrent elections are used for the congress. Apparently, despite
the holding of some elections at a different time, the holding of others
along with the presidential is sufficient to induce the maintenance of the
two-party dominance. However, this remains a new regime, and the num­
ber of parties has been showing a tendency to grow throughout the period
(from 2.6 in 1983 to 3.4 in 1989), raising the possibility that the existence of
two midterm elections for every concurrent one could enhance the pros­
pects for smaller parties.
W hen elections are nonconcurrent, average NP = 2.7 and average Nv =
5.0. For Brazil (before 1964) and Chile (before 1973), the gap between NP
and Nv is enormous, as the holding of congressional elections apart from
presidential elections encouraged parties to run independently of presi­
dents' parties. As noted in Chapter 9 in our discussion of archetypal presi­
dentialism, the nonconcurrent format allows party systems for the two
branches that are almost entirely different from one another.
Brazil changed its electoral cycle during the democratic regime that ex­
isted between 1945 and 1964. The first two elections, 1945 and 1950, were
concurrent presidential-congressional elections. Thereafter, elections to
the two branches ran nonconcurrently. In both periods, Brazil conforms to
the norm for the appropriate category. In 1945 especially, the two candi­
dates and their parties dominated the field. In 1950, the fragmentation of
the party system that we have discussed began to assert itself. Perhaps
12 However, the value of Nv is deceptively high. Besides including two unusually fragmented
elections (1963 and 1968, which were addressed in Chapter 9), it also fails to indicate the
degree to which two parties have dominated since 1973, as numerous very small parties
push up the value of N. Consider this tabulation of vote shares for Acci6n Democratica
(AD), the Christian Democratic COPEi, and the socialist MAS in Venezuela, 1973-88:
1973 1973 1978 1978 1983 1983 1988 1988
Pres. Cong. Pres. Cong. Pres. Cong. Pres. Cong.
AD 48.6 44.4 43.4 38.5 55.3 50.0 52.9 43.3
COPEi 35.2 30.2 45.3 38.6 32.6 28.7 40.4 31.1
MAS 3.7 5.3 4.7 6.0 3.4 5.7 2.7 10.2
N, 3.4 3. 1 3.0 3.4
13 Taagepera and Shugart (1989:269) estimate Venezuela's effective magnitude to be 27,
which is simply the geometric average of the extreme values: M = 90 for the very small
parties, owing to the low threshold, and M = 8 on average in the districts where typically
only the three largest parties win seats. The other PR systems in Table 10.2 are all in the
range of 7 to 10, lacking the complicated minority-party provisions of Venezuela.
222 Presidents and assemblies
because this was desirable to the locally based politicians, they subse­
quently moved their congressional elections to different dates from the
presidential and the fragmentation of the congressional party system in­
creased substantially. At the presidential level, however, only in the 1955
election did the front-runner fail to draw as much as 48%, while the median
shares for all presidential elections in this period are typical for plurality
elections more generally, as we see in both Table 10.1 and Table 10.2.
Colombia is the only case for which, counterintuitively, NP > Nv. This is
explicable by the unusual electoral system used in Colombia, where fac­
tions of parties or individual candidates can present their own lists without
necessarily having been endorsed by the party leadership and with no
pooling of votes among lists (see Shugart 1992a). One or both of Colom­
bia's parties have frequently presented more than one presidential candi­
date each in certain elections, thus NP is significantly greater than 2.0. In
congressional elections, on the other hand, the same factions and more are
counted with Liberals or Conservatives, hence low Nv. If we used factions
rather than parties as the units of analysis, assembly N for Colombia would
be in the same range as for Brazil, or quite possibly higher.
These findings confirm empirically what was sketched out in Chapter 9,
that holding congressional elections on their own terms allows such elec­
tions to represent diversity that is bound to be subsumed under broader
coalitions - for assembly as well as president - when the elections are at
the same time. More empirical data as well as theoretical justification for
the observed patterns will be presented in Chapter 11.

Plurality for both branches. We have two cases in which both president and
congress are elected by plurality in one-seat districts. In both the Philip­
pines and the United States, the values of assembly N are lower than in any
of the cases using PR for their assemblies. Indeed, in the United States, at
both presidential and congressional levels, N = 2. l , while for the Philip­
pines the figures are NP = 2.0 and Nv = 2.1. These two cases are thus the
purest two-party systems that we observe. Over time these two countries
have had practically no important third parties or presidential candidates,
so high are the barriers that plurality rule imposes to minor parties. At least
where PR is used, a third party can hope to win some seats on its own; such
a party may then also run its own presidential candidate, thus raising NP as
well as Nv.
The reason for the virtual absence of third parties stems from more than
the simple barriers imposed by M = l and plurality, however. After all,
regionally based third parties could compete for (and win) seats in some
districts. The two-party configuration stems as well from an electoral sys­
tem that facilitates members' cultivation of ties with local constituents
rather than dependence on party (Cain et al., 1987), as would be the case if
list PR were used or if the system were parliamentary. Because there is no
 Electoral rules and the party system 223
vote of confidence in a presidential system, parties do not need to control
who is elected under the party label. Because there is also no party list,
given one-seat districts, parties can afford to accommodate some diversity
that in either a parliamentary system or a party-list electoral system would
be expressed by multiple parties (cf. Epstein 1967).
Presidentialism and concurrent M = 1 elections allow congressional candi­
dates to enjoy two advantages simultaneously: identification with a national
presidential candidate, and the freedom to pursue local particularism. Such
parties should, obviously, be very diffuse and internally diverse, as well as
involved to a great degree in the provision of constituent services. Such a
characterization clearly fits the parties of both the United States and the
Philippines. 14

Majority systems. W hen the president is elected by majority runoff, there is

not much obvious difference between concurrent and nonconcurrent elec­
toral cycles in terms of the number of parties. The majority runoff exhibits
considerably less consolidation of forces behind presidential candidates
(Table 10. 1), meaning that there is a multiparty system already at the
presidential level.
In systems in which there is an assembly election concurrent with the first
round of the presidential election, 15 the combination of PR (especially with
high M) and the majority presidential election might encourage party lead­
ers to present presidential candidates with the primary aim of attracting
votes to their congressional slates. By doing so they could enhance their
bargaining leverage with the candidates who will square off in the second
round. W hen there is the expectation that there will be a second round and
many slates enter the first round, there is a slim likelihood that the eventual
winner of the presidential election will have a majority in congress. Thus
the concurrence of a (high-magnitude) PR congressional election and ma­
jority presidential first-round election might encourage considerable frag­
mentation, as the experiences of Ecuador and Peru (especially in 1990)
both suggest. 16
14 Such a characterization also fits our other party system that is closest to being "pure" two­
party: Colombia.
15 In Ecuador, the first congressional election (1979) was concurrent with the second round
of the presidential election. Since then, however, congressional elections have been concur­
rent with the first round. There was no obvious difference in fragmentation in that first
congressional election, although if such a practice were made permanent, we might expect
Jess fragmentation over time.
16 In Ecuador, this fragmentation at the presidential level is much exacerbated because the
electoral law prohibits a party from maintaining its own congressional list while simulta­
neously entering into an electoral coalition with another party behind a common presiden­
tial candidate (Conaghan 1989). Even without such a law, however, the incentive for
unlikely winners might be to maximize their congressional representation for the sake of
presenting the eventual presidential winner with a fait accompli something like the follow-
224 Presidents and assemblies
Peru's highly fragmented presidential election in 1990 (N = 4.4) followed
previous elections that had been much less fragmented. As already sug­
gested, the majority runoff encouraged this fractionalization of a party
system once characterized by at most three important partisan blocs
(APRA, ex-president Belaunde Terry's AP, and, after 1978, the left). In
Ecuador, there is no obvious trend either toward or away from fragmenta­
tion. The Ecuadoran party system is extremely fragmented to start, but as
shown in Table 10.2, it is even more fragmented in midterm elections (Nv =
10.8) than in those assembly elections that are concurrent with the presiden­
tial first round (Nv = 7.4). Thus in Ecuador, which (unlike Peru) started out
with extreme fragmentation, the concurrence of the presidential election
induces some consolidation, despite majority runoff. However, this consoli­
dation is both feeble, owing to the majority system, and temporary, since
there is always a midterm election two years away.
Among our cases, there is one in which there was consolidation of poten­
tially competitive forces, in spite of the majority presidential election; it
was also the one case in which a very low district magnitude was used for
the congressional election that was concurrent with the presidential first
round. In Chile in 1989, the center and left presented a common candidate
who won in the first round. W hy did they form a preelection pact, as we
would expect with plurality, when one of the two would have been certain
of advancing to the runoff in which it could have counted on the other anti­
Pinochet candidate's supporters? The key rests in the low M for the con­
gress. W ith M = 2 and party lists (but with voters choosing a candidate
within a list), the center and left could be assured of sufficient votes com­
bined to win both seats in many districts. Had they run separately, in many
of those same districts only one tendency, either a centrist or a leftist, might
have won, with the other seat going to the right (Brager 1990). Running
together, they could pool their congressional votes in order to maximize
their representation. Thus, low congressional magnitude with a party-list
allocation rule concurrent with the first round of a majority presidential
election discourages the fragmentation that is otherwise encouraged by
majority runoff. 17

CONCLUSION

We have seen empirically the effect on the party system associated with
different electoral rules. Our first observation was that plurality election of
ing: "You'll have to deal with us once you're elected, so make a pact with us before the
second round." Bolivia, where the "second round" is in the congress and among the top
three candidates, thereby even further encourages fragmentation of the concurrently
elected legislative slates (all the more so given that ticket-splitting is prohibited). In Bo­
livia, NP = N, = 5, approximately.
17 The new Chilean system may, at the discretion of a president exercising dissolution pow­
ers, have nonconcurrent elections as well.
 Electoral rules and the party system 225
the president tends to induce the formation of two blocs to contest the
presidency, while majority elections are often considerably more frag­
mented. Even when there is a single large party under majority runoff, the
opposition tends to be more fragmented than when plurality rule is used.
Thus we have argued that the goal of gaining a broad, popularly endorsed,
coalition behind a potential executive is better served by plurality rule than
by majority runoff rule. Ironically, the plurality method is more likely to
produce victory by a genuine electoral majority than is the method of
majority runoff. The majority-endorsed candidate who must eventually
emerge from a second round in a two-round majority system may have
been endorsed in the first round by far less than a majority of voters and
often less than a third. A candidate's ability to win the second round is
largely contingent on which two happened to survive the first round. Thus
first-round contingencies are especially important to determining who wins
the second round. A key contingency is the number of candidates running,
which tends to be quite high.
We admit a strong preference for election of the president by a method
other than majority runoff. Almost any other method that we have dis­
cussed and that has been in use in some country or countries - plurality, a
U.S.-style electoral college, the Costa Rican 40% rule, or the Nigerian
concurrent pluralities - is more conducive to the forging by parties of
preelection coalitions broad enough to approach closely, if not reach, an
absolute majority. We find especially appealing the double complement
rule. As a variant of plurality, it would encourage broad preelection coali­
tions, but it does not carry the risk of very narrowly endorsed plurality
winners. The avoidance of such winners, as has occurred in some cases
(Chile in 1970 being the most famous), is the purpose behind the recent
move toward majority runoff rules in many countries. But the double
complement rule has the further advantage of not encouraging the fragmen­
tation of the field of competitors. For now, however, plurality rule remains
the most common method for electing presidents. In the next chapter we
discuss and analyze in more detail the effects of combining plurality rule for
the presidency with PR for congress. We see no reason to doubt that the
general conclusions derived there would apply just as robustly to any sys­
tem that employed the variant of plurality we have called the "double
complement rule."
 11
Elector al cycles and the p arty system

As mentioned in Chapter 10, m ost studies of the relati onship between

v otes and seats have n ot taken int o account the p ossibility that presidential
and parliamentary systems might differ on the relati ons being investigated.
Indeed, m ost have been based overwhelmingly, if n ot exclusively, on data
from parliamentary systems. The empirical results of Chapter 10 suggest
that presidential systems might make a difference in tw o ways. The first,
and m ost obvious, is that, unlike parliamentary systems, presidential sys­
tems have separate elections f or tw o separate agents of the electorate. The
second, which fl ows directly fr om the first, is that these electi ons need not
be at the same time; we have seen that the timing of electi ons, the elect oral
cycle, makes a difference in the number of parties. If the number of parties
varies, s o perhaps d o such outc omes as the extent and quality of representa­
tion and the stability and effectiveness of the system. If c oncurrent elec­
tions by PR are empl oyed al ong with plurality f or the president, th ose
wishing t o articulate min ority viewp oints might find themselves f orced ei­
ther t o j oin a large presidential c ontender's party or else suffer electorally
as a third party. Then, perhaps, the degree t o which the system represents
diversity may be undercut. On the other hand, if n onc oncurrent electi ons
are used, presidents might find they rarely have maj orities in the congress.

R O L E O F A S S E M B L I E S A N D T H E N U M B E R OF P A R T I E S

A principal theme o f this b o ok has been that assemblies d o n ot exist in

instituti onal is olati on within their p olitical systems. Instead they are
elected either t o form an executive (the cabinet in a parliamentary system)
or t o c oexist with an executive (the president in a presidential system) or
b oth (premier-presidentialism). This fundamental difference in the means
of c onstituting the executive can be expected t o affect the number of par­
ties in the assembly inasmuch as the purp ose of the assembly differs acc ord­
ing t o the type of executive.
Where the only means that v oters and, during electoral campaigns, party
leaders have t o influence the c omp ositi on of the executive is thr ough parlia­
mentary elections, the decisi on on whether or n ot t o v ote for or even to run
 Electoral cycles and the party system 227
candidates is affected by calculations on government formation. This is the
process discussed in Chapter 9, and it is tied to the "efficient secret" (Cox
1987). As a result, even a party that has little chance of winning in a given
district may present a candidate, or list of candidates, in that district in
order to institutionalize itself in the voters' minds as a potential future
government or partner in a governing coalition. Such a party, by perhaps
playing a spoiler role, also serves a "blackmail" (Sartori 1976) purpose
against parties in parliament.
In a presidential system, on the other hand, the election of the assembly
need not bear any relation to the choice of the executive. Separation of
powers carries with it separation of elections. This basic feature of presiden­
tial systems has been invoked to explain why congressmen in the United
States spend so much time courting personal followings in their own dis­
tricts in contrast to the United Kingdom, where qualities and skills of the
candidate are found to have much less effect on the likelihood of election
(Cain et al. 1987). In a parliamentary system like the United Kingdom's,
voters select a party for policy purposes (Cox 1987). In a presidential
system, on the other hand, the selection of a candidate or party for con­
gress can be motivated differently from that for a preferred executive.
Even so, we have noted (Table 10.2) that, despite the formal separation of
elections and powers that define presidentialism, when plurality presiden­
tial and PR congressional elections occur at the same time, two major
parties emerge.
In disentangling this issue let us take the case of M = 1 and plurality for
the assembly races. If the system is parliamentary, the ability to offer a
long-run collective alternative government and perhaps force a hung parlia­
ment is an incentive for third parties to run and for voters to cast "protest"
votes for such parties, even in districts where their chances of victory are
slim or nil. In a presidential system, the advantages of a few districts won
by a minor party are not such an incentive when said party cannot play any
meaningful role in selecting (or subsequently voting no confidence in) the
executive. This reasoning accounts for why the U.K. and other parliamen­
tary systems with M = 1 have multiparty races for MPs even if two parties
are more important than the others (Nv = 2.3, according to Table 11.1),
while the United States and the Philippines have almost perfect two-party
systems (N = 2.1, according to Table 10.2).
Increase M with PR and what happens? The incentive for multipartism
grows. It was already higher in parliamentary systems even at M = 1, as
Table 11.1 shows. Now, given higher M, even in presidential systems, the
"prize" available to minor parties increases in value. Rather than, at best, a
small number of token seats and lots of defeats with no ability to play a role
in executive formation, a minor party in a presidential system with PR
congressional elections has the opportunity of winning a proportional share
of seats and thus institutionalizing itself. If elections are concurrent and the
228 Presidents and assemblies

Table 1 1 . 1 . Effective number of parties in parliamentary systems, according to

effective magnitude

Effective
Effective number of
magnitude parties
(M) (Nv)
1 2.3
2-5 2.7
>5 3.7
Source: Taagepera and Shugart ( 1989).

rule for the president is plurality, then unless a very large number of voters
split their tickets, a basically bipolar system results; but an institutionalized
presence of minor parties leads to the potential for coalition formation
before future presidential elections. And, assuming that the congress has
even minimal lawmaking powers, even a small share of seats won via PR
might allow a minor party to induce the president to negotiate for legisla­
tive support. Thus the effective number of parties in the congressional
elections of presidential systems with concurrent elections is higher (3.1 on
average, according to Table 10.2) when M is high than when M = 1, but not
as high as in parliamentary systems that use large-magnitude PR. Note that
Table 11.1 shows that, with parliamentarism and large M, average N., = 3. 7.
In considering the interaction of executive form, assembly M, and the
electoral cycle, the basic combinations of institutional variables can be
arrayed as in Table 11.2.
The institutional package that will concern us in greater detail in the next
section of this chapter, plurality president elected concurrently with a PR
(M > 1) congress, thus occupies an intermediate position. So does that
which has provided much of our contrast and for which the phrase "effi­
cient secret" was developed, the British-style parliamentary system with M
= 1. As discussed, the British system occupies an intermediate position
because, with parliamentary elections serving as the basis for choosing an
executive, parties seeking to provide an alternative to the major parties will
tend to run in many districts even outside of their own regions of strength.
Examples of such parties are the Liberal Party and the Social Democrats in
Britain, the Social Credit Party in New Zealand, and the Canadian New
Democratic Party.
The presidential systems that are intermediate are encountered here
because of the simultaneous action of the Duverger effects tending in both
directions: the plurality presidency pushing the number of competitors
toward two, and the PR congress pushing the number of parties upward.
We have seen that the presidential effect does push the effective number of
 Electoral cycles and the party system 229

Table 1 1 .2. Type of assembly party system and associated institutional arrangements

Type of assembly Associated institutional

party system arrangements and examples
Multiparty Parliamentary with M> 1
(Continental Europe)
Plurality president with
nonconcurrent M> 1 congress
(Chile pre- 1973)
Majority president
(Ecuador, l?eru)
Two-party, but with minor Parliamentar'J with M=l
parties (New Zea!and, U.K.)
Plurality president with
concurrent M> 1 congress
(Costa Rica, Venezuela)
Almost pure two-party Plurality president with M=l
congress
(Philippines, U.S.)

parties in the congressional elections down from what is observed in parlia­

mentary systems when M > I. A model to be shown in the next section
suggests that there is also some effect the other way, pushing up the num­
ber of presidential contenders. Let us consider this type of electoral cycle.

CONCURRENT ELECTIO N S : N UM B E R O F PARTIES

We have seen that the institutional design of plurality presidency and con­
current assembly by PR tends to support two major parties, plus some
minor parties. Thus far our discussion of this association between a particu­
lar configuration of institutions and a type of party system has been primar­
ily empirical. In this section we turn to a theoretical discussion of this
particular type of presidential regime, placing it in the context of parliamen­
tary regimes. 1
1 In contrasting this type of system with parliamentary regimes, we shall also be effectively
contrasting it with those presidential systems that use only nonconcurrent elections or elect
their presidents by majority runoff. As we have seen, these latter systems tend to behave
more like large-M parliamentary systems in the sense that they are associated with
multipartism. We must add, however, that N for such presidential systems tends to exceed
by a considerable degree N for parliamentary systems using comparable magnitude. While
the institution of parliamentary confidence is seen to increase N at low M, at higher values
of M, the desire of parties in parliamentary systems to be able to present themselves as
viable governing parties must temper the increase in N. This finding relates directly to the
theoretical argument about inefficiency in regionally based party systems: These politicians
would have lost much of their ability to service local interests had parliamentarism been
adopted, since some fusion of diaparate parties would have been encouraged. Thus at high
M, presidential systems using nonconcurrent elections or majority runoff rules have higher
values of N than parliamentary systems.
230 Presidents and assemblies
(This discussion of the "presidential difference" requires the use of
mathematics, but for the sake of less mathematically inclined readers, we
have left the more complex matters out of this chapter. Readers who want
to see how the model presented herein was derived, as well as more in­
depth discussion of its implications, should refer to Appendix B. Others
may ignore that appendix without loss of continuity.)

Theory of the number ofparties

The model for the number of parties in presidential systems is part of a
broader family of models of the relationship between district magnitude
and number of parties. The model starts with the most simple assumptions
about the relationship between assembly M and seat-winning parties, then
adds necessary complexity to take account of multiple districts. Only then
is it possible to estimate the effects of presidentialism on the basic relation­
ship. The agreement to empirical data, as we shall see, is quite good,
considering the several steps involved.
The first step, suggested originally by Taagepera (1989), is to estimate a
mathematical relationship between the number of seats available in a dis­
trict (M) and the number of parties that can win those seats, which is
designated by p,, the number of actual (not effective) parties winning at
least one seat each. Taagepera suggested that the relationship, in the ab­
sence of other information, might be expected to be the geometric average
of 1 and M; i.e.,
Ps = M·s.
Shugart (1988) tested this expression and others derived from it against
empirical data from several systems and found that the data tended to
agree with the estimate. Taagepera and Shugart (1992a) made further ad­
justments to the reasoning, indicating that such a simple model could apply
only in the simplest district. Such a simple district would be the only district
in the political system, such that the politics of other districts, perhaps with
a smaller or greater M or different political cleavages, could not "contami­
nate" the effects of M in our "test" district. Fortunately there are such
systems in which this simplest of individual districts exists: those in which
seats are allocated in a single nationwide district, as in Israel and the
Netherlands. For such districts, the above equation provides an excellent fit
to empirical data (Taagepera and Shugart 1992a).
We are confronted with a question as to why the relationship should take
this form. W hy would party leaders and candidates assemble their lists and
campaigns, and why would voters distribute their votes, in such a way that
this relationship between p and M would prevail, on average, instead of
some other. To begin untangling this question, let us start by considering the
range of possible outcomes. In a district of M seats, obviously there are clear
 Electoral cycles and the party system 231
boundaries to the range of outcomes. On the one hand, one party may win all
the seats, in which case p = l = M°. On the other hand, each seat may be won
by a different party, in which case p = M = M1 • We seek a value, k, for the
general equation p = Mk , in which O < k < l. But why should k = .5?
Obviously k = .5 means that it is simply the average of the extreme cases,
implying a normal distribution. However, this assumes that both extremes
are equally likely, or equally unlikely, to occur. Is this justified?
Assume the simplest of parties and allocation procedures in the simplest
of districts with a value of M considerably greater than 1. We shall not
concern ourselves with the precise allocation rule - what kind of quota or
divisor is used to allocate seats to parties - as long as it is some form of PR.
We should, however, assume a closed-list form of PR, as this will allow the
model to depend only on parties and not on intraparty vote competition.
Thus the determinants of whether a candidate wins a seat are simply the
interaction of some number of parties within a district of M seats taken in
isolation from any other districts and from the process of executive forma­
tion. Candidates in this model are assumed only to seek a seat in the district
for its own sake. They do not care about executive formation or policy
debates. W hat would we expect an aspirant for such office to do?
If policy and executive formation did not matter, candidates could group
together on a list of M candidates that would maximize each aspirant's
chance of winning. In other words, they would strive to form a list capable
of winning all M seats, giving us p = l.
But what if there are more than M aspiring candidates? If there are, what
happens to the candidate who would merely be ranked by this unique party
as candidate number M + l? His chances of winning would be zero, until
some future election when a member higher on the list died or retired. So
he or she would do just as well to form his or her own list and seek to win
without the dominant party. As soon as this is done, the certainty of win­
ning for the candidate ranked in position M on the dominant party's list is
diminished. The party may still win all M seats or it may now win only M -
1 seats. Candidate M now has to weigh the expectation of winning on the
dominant list versus the possibility of winning by joining with the candidate
who defected from the M + l position. Alternatively, candidate number M
might also present his or her own list. Each time someone makes such a
calculation in favor of defecting, the exponent in the equation p = Mk
increases from zero and toward one. If all candidates except the one in the
number one list position defect, we have p = M 1 • The point is, therefore,
that as long as there are more than M aspirants, one list that wins all seats is
not expected to be an equilibrium, except possibly at the lowest values of
M and, of course, when M = l. Someone who is left out of the first M list
positions will set up a rival list, thereby inducing defections from the domi­
nant list.
But why start with the assumption that M aspirants initially seek to form
232 Presidents and assemblies
one list? We need not. Let us try starting with the assumption that each
aspirant presents his or her own list. The seats now go to the first M vote
winners in what becomes a system of the Japanese single-nontransferable­
vote variant of plurality. We would then have p = M 1 , or as many "parties"
winning seats as there are seats to be won.
But given the option of forming lists in a PR system, candidates who
expect to fall lower than M in the voters' rankings may calculate they are
better off joining with a candidate likely to finish higher. Any such candi­
date thus joins a list of a candidate who is more resourceful, more charis­
matic, or otherwise more likely to secure the election of the candidate
whose chances of winning independently are not as great. Again, as long as
there are more than M aspirants, an extreme value of p, in this case p = M,
is out of equilibrium. Some aspirant will always seek to run in tandem with
some other, in order to increase her or his own chances of election. Such an
arrangement should in turn be desirable to the stronger candidates, by
making other politicians dependent on them. That is, by "sharing" their
electoral strength, they can increase their influence in the assembly. They
are, in effect, avoiding "wasting" their excess electoral strength.
If either extreme is equally unlikely, the pulls toward separate candida­
cies, on the one hand, and toward unity, on the other, must converge in the
center, unless we assume that one pull is stronger than the other. Lacking any
reason to assume a greater pull from either end, the exponent on M for this
simplest of all isolated districts would be the average of O and 1 : p = M5 •
The reasoning thus far has assumed no political issues, no executive, and
no other districts. Deviations from the norm, represented by p = M s , must
be accounted for by any one or combination of these additional factors.
Taagepera and Shugart (1992a) account for one: the existence of other
districts in most systems. Their means of doing so are reviewed briefly in
Appendix B. Here we concern ourselves with the effects of one particular
form of executive formation, the election of a president (with M = 1 and
plurality) at the same time as the election of members to fill assembly seats
(with M > 1 and PR) .
Once again, the complexities of this mathematical exercise are devel­
oped in Appendix B. If p = M s is the model for a basic district and provides
a reasonable fit to data drawn from parliamentary systems, as it does
(Shugart 1988) , what exponent would be appropriate for the assembly of a
presidential system? Recall that our concern here is with systems in which
president and congress are elected concurrently. Therefore the congres­
sional election is occurring at the same time as another election for which p
must be 1 , or p = M°.
What if there could be no ticket-splitting and the assembly and executive
were of equal importance electorally? Then our estimate should be an
average value between the exponent .5 (for assembly) and O (for presi­
dent) . This would predict an exponent of .25, or a fourth-root relationship.
 Electoral cycles and the party system 233
However, voters need not cast a single ballot, so the effects of the presi­
dential election on the congressional (and vice versa) must be tempered.
Taking averages again (lacking a basis on which to make some other proce­
dural assumption), we get the following estimate for p for the congress of
such a system:
p = M ·31s = M 18_
3

This relationship is tested in Figure 11.1 for data from three presidential
systems that meet the criteria: concurrent elections, president elected by
plurality, and congress by (closed-list) PR. These countries are Costa Rica,
the Dominican Republic, and Venezuela. The figure shows a good agree­
ment with data. The agreement is improved by the equation that takes into
account the effects of multiple districts (instead of assuming an isolated
district). That equation is represented by the dashed curve in Figure 11.1,
and its derivation and form are shown in Appendix B.

The effective number of congressional parties. By a procedure elaborated

upon in Appendix B, it is possible to move from this expression of the
number of actual seat-winning parties to the number of effective parties
winning seats. This relationship, which also gives a very good fit to data, is
derived and presented in Appendix B. Here we shall show only the final
step, an expression of the relationship between M and the effective number
of electoral (vote-winning) parties (NJ. This relationship is shown in Fig­
ure 11.2. The equation for the curve derived from the theoretical model is
NV = 1.sM21 •
Also shown in the figure is a dashed curve conforming to the model's
prediction for parliamentary systems, N,, = 2M 25 • The derivation of this
expression and the data that confirm its validity are provided in Taagepera
and Shugart (1992b), and a brief overview may be found in Appendix B.
We discuss in Appendix B the degree of fit of the equation above to the
data in the figure. Recall that this expression is meant to be a theoretical
statement; it was not derived through empirical curve-fitting, as would be a
regression of Nv against M. That this equation should express the right
tendency in the data relative to the parliamentary data set gives us cause to
believe in the validity of the general theoretical reasoning.

The effective number of presidential contenders. The expression for the

effective number of presidential contenders in concurrent elections accord­
ing to congressional district magnitude is derived in Appendix B and is
NP = l.68M09 •
This expression tells us that minor parties fare slightly better in larger­
magnitude districts than in smaller districts, not only for congressional
 10
• Costa Rica, 1978-86
o Dominican Republic, 1962, 1982
c Venezuela, 1973-88

Forbidden
5 Zone

=
Cl.
�
="!i
4
·a
3 ..
..
.! 2

2 3 4 5 10 20 30 40 50

District magnitude (M)

Figure 1 1 . 1 . District magnitude and number of seat-winning parties, Costa Rica, Dominican Republic,
and Venezuela

Downloaded from Cambridge Books Online from within the IP domain of the University of California on Fri Jan 02 01:39:54 GMT 2015.
 10

• Costa Rica, 1978-86

--zrl
.-.;,. 0 Dominican Republic, 1962, 1982

c Venezuela, 1973-88
"'!a 5

e
c:i..

I..."'
4

3
Q

�CII
=
§
2
"'
CII
;,.

�
r.l 0

2 3 4 5 10 20 30 40 50
District magnitude (M)
Figure 1 1 .2. District magnitude and effective number of electoral parties, Costa Rica, Dominican Republic,
and Venezuela

Downloaded from Cambridge Books Online from within the IP domain of the University of California on Fri Jan 02 01:39:54 GMT 2015.
 10
,..._
r::..
Z
._, • Costa Rica, 1978-86
!" o Dominican Republic, 1962, 1982
"O c Venezuela, 1973-88
;a
5 5

=
-a
·-=
ci,
"O
4

eClo
.....
"lil C
3 0
C
0 0 C
B C
� 0
B
8
0 C 0
0
c3i B ... 00
C C

J a· � a Np :: 1 .6SM·
e=
0
2
=
C

0 0
ci, 0 0
>
.:
0
�
i:.:i
I I I I I I
I I I I

2 3 4 5 10 20 30 40 50

District magnitude (M)

Figure 1 1 .3 . District magnitude and effective number of presidential candidates, Costa Rica, Dominican
Republic, and Venezuela

Downloaded from Cambridge Books Online from within the IP domain of the University of California on Fri Jan 02 01:39:54 GMT 2015.
 Electoral cycles and the party system 237
elections but also for presidential. They do so despite plurality rule for the
presidency and despite the clear dominance of the two biggest parties, as
indicated in Figure 11.3.

CONC U RRENT E L ECTIONS : POLITICAL STRATEGIES

Appendix B develops a theoretical model of the number of parties in

presidential systems with concurrent elections. That model is a mac­
rostructural model that concerns itself with systemic-level variables. How
do the systemic level independent variables (electoral rules) lead to the
party system dependent variable? Can we suggest ways in which electoral
rules affect the behavior of individual political actors - party leaders and
voters? The following section is an attempt to consider how rational actors'
goals and strategies, within a given institutional context, lead to observed
outcomes. The microlevel explanation below produces results consistent
with the macrolevel explanation of the preceding section and Appendix B.

Major parties and concurrent elections

We have noted that it is possible, given the mix of plurality and PR in many
presidential systems, to have two-candidate competition for the presidency
but multiparty competition for congress. Yet, in systems employing this mix
of electoral rules, such a configuration of competition has been observed
primarily in systems using only nonconcurrent elections (Chile before 1973
and Brazil in the 1950s). A qualified exception is Costa Rica until the
1970s, where a coalition of conservative parties maintained separate con­
gressional identity even while running a common presidential candidate at
each election. More recently, however, even Costa Rica has conformed to
the two-party dominant format in congressional as well as presidential
elections. W hy do parties not more commonly take advantage of the mixed
electoral rules and maximize their individual organizational interests by
maintaining separate ballot identities, as they do when all elections are
nonconcurrent?
To approach an explanation, we have to look at the incentives of individ­
ual members of parties within the bloc endorsing a particular presidential
candidate, as well as at the voters. First, let us assume that most politicians
are motivated both by power seeking (winning office or the spoils of office)
and by policy goals. This means, especially in a presidential system, that
they seek to maximize their chances of being in a team that wins the
executive.
Moreover, the great visibility of the presidential candidate in the race
suggests that voters are likely to focus on the candidate's own party as the
vehicle by which to endorse their preferred presidential bloc congression­
ally as well as in the executive. The Costa Rican experience confirms the
238 Presidents and assemblies
strength of these coattails, as even when there was an explicit coalition, the
presidential candidate's own party significantly outpolled the other parties
in the coalition, even beating those parties that had been important in
earlier periods. 2 Thus, rational office-seeking politicians - whether or not
they also have important policy goals - would be better off being on the
president's common slate of proponents in the congressional race. A candi­
date might be expected to prefer the near certainty of being elected on such
a common slate, with the list position being negotiated before the election,
over the risk of being overlooked on a separate list as a result of the
president's coattails. 3
This discussion has considered why a presidential candidate's party
would dominate over other possible parties supporting that same candi­
date, and is borne out by electoral dynamics in the Costa Rican case. But
why wouldn't parties seek to endorse a common candidate who did not
have his or her own party? After all, nonparty heads of government are far
more common in presidential systems than in parliamentary ones (Blonde!
and Suarez 1981). Aside from the difficulty of convincing presidential candi­
dates to have separate congressional slates supporting them - an irrational
situation for them because it muddles the message of their own campaign
and proposals for government - there is a straightforward problem of cam­
paign dynamics. How can leaders of separate parties maintain the neces­
sary unity to elect a president, if at the congressional level they are busy
pointing out to voters why they should support this slate over those others?
Such a multiple-slate approach is likely to undermine the common themes
and sense of purpose - and hence the efficiency - of the whole campaign.
Thus, even V ioleta Chamorro, although unaffiliated with any party in her
successful bid for the Nicaraguan presidency in 1990, had a common
"UNO" slate behind her. This common slate was crafted even though there
were clear centrist, rightist, and even leftist tendencies within the anti-
2 When the Union Nacional candidate was the standard bearer for the conservative parties in
1958, its congressional slate received 19.2%, compared to 6.1% in the previous (1953)
election. In 1962, when it did not have the conservatives' presidential candidate, its congres­
sional vote declined to 12.8%. In that election the presidential candidate was from the
Partido Republicano, whose slates for congress obtained 32.5% , compared to 20.1% in
1958. In 1966, without a presidential candidate, the party's only surviving independent
faction obtained only 0.4%.
3 This observation can also account for why, especially in Venezuela with its easily cleared
threshold for minority party representation, only very small parties tend to participate
separately for congress while endorsing a major party presidential candidate. Individual
political entrepreneurs who can place themselves number one on their own lists can stand a
chance of getting elected in a high-magnitude (or, in Venezuela, nationwide) district. How­
ever, if one is going to be farther down on some list, it is better to be number eight in a
twenty-five-seat district (for example) on the list for a presidential contender's party that, as
the unified expression of the presidential candidate's supporters, might win 40% of the
vote, than it is to be number four on a secondary supporting party that might not even win
15%.
 Electoral cycles and the party system 239
Sandinista front. The common slate was a collective solution to what other­
wise would have been a highly individualistic congressional campaign.
Closed lists made it even more collective, eliminating any rationality of
pointing out differences among members of the coalition - until after the
election of course. Conveniently for the opposition, as it turned out, the
Sandinista's electoral law already mandated closed-list PR.

Voters: Why so little ticket-splitting?

The models and empirical data suggest that there is indeed some ticket­
splitting in the concurrent elections format. Yet there is not enough to make
for a parliamentary-style multiparty system even when a high magnitude is
used. W hy not? Surely most of the answer has already been provided. Most
politicians want to enhance both their office-seeking and their policymaking
potential. To do so means lining up behind a party capable of winning the
executive, that is, the presidency. Thus there is a substantial gap in size
between those two parties that contest the presidency and the remaining
parties that do not seriously do so, at least in the present election. This means
there are not many options for voters besides the two big parties.
Another institutional reason reinforces straight-ticket voting by the great
majority of voters in the systems under consideration: the list form of PR.
Especially with the closed list, but presumably with most variants of open
list, the party can present itself as a team. 4 Given the concurrence of the
congressional election with the presidential, voters are likely to perceive
that voting for the president's party is a means of confirming the endorse­
ment given to the presidential candidate against his or her opponent. Vot­
ing for third parties might increase the probability that the party represent­
ing the team opposed to one's preferred president will be able to stymie the
president. Therefore, most voters who really prefer that presidential candi­
date are also likely to endorse her or his team. 5
4 Some variants of open list clearly undercut this "team" tendency, as in Colombia and Brazil.
The multiple list system used in Colombia makes that country's elections at least as
candidate-centered as in the United States. Thus, we may be able to point to an institutional
reason why Colombia's two parties remain so dominant: the combination of plurality­
presidentialism with a congressional electoral system that emphasizes personal over policy­
oriented voting, as in the United States. The Brazilian form, which is combined with
numerous ancillary features that weaken partisan identity, is an exception (Mainwaring
1991). Even if this electoral system in Brazil were used concurrently with a plurality presi­
dential election, it might remain as individualistic as in the United States, where ticket­
splitting is indeed high. Less extreme forms of open-list systems should be somewhere in the
middle range. Currently no presidential system has had long experience with concurrent
elections and an open list variant of PR, although Venezuelan voters beginning in 1993 will
have one-seat districts for half the seats. Such a format should weaken the virtual "team
ticket" nature of parties found under presidentialism and closed lists. See Shugart (1992a).
5 This assumes, of course, that voters want the same things from their representatives at both
levels, president and congress. Jacobson (1990) has recently suggested that this may not be
240 Presidents and assemblies

The interaction of presidential and congressional elections

The reasoning, presented above and in Appendix B, that leads to expres­
sions for both congressional and presidential parties suggests that the con­
currence of elections for an assembly using M > 1 and PR with an executive
using M = 1 and plurality leads to a two-way "contamination." The effec­
tive number of parties winning votes (and seats) for the congress tends to
be lower at a given magnitude than what we find both theoretically and
empirically is the case among parliamentary systems. Thus, even though
high M could accommodate many more effective parties, the plurality
presidential election dampens the number of parties. The above equation
that showed that NP rises with M tells us that the congressional election also
contaminates the presidential election with more effective parties than
would be expected in the absence of the congressional election.
We do see evidence of some ticket-splitting, yet its extent is limited for the
reasons suggested above. Minor parties do exist, however, because of PR,
and even though they do not expect to win the presidency, they may nonethe­
less present presidential candidates. They may do so in part to present a
long-run alternative political force that, they hope, may win the presidency
at some future election. However, such long-run goals need not be present to
encourage "third" parties to present candidates for the presidency. It is likely
that a party would expect that its ability to maximize its potential congres­
sional representation would be aided by having a full slate, including a
presidential candidate, in order to attract attention and votes. At the end of
this chapter, the special dilemma of third parties in concurrent elections is
contrasted with their position in nonconcurrent elections.
How does this idea of "two-way contamination" in concurrent elections
compare to other institutional formats? We, of course, do have cases of
elections for assemblies where there is no popular election - concurrent or
otherwise - for an executive. These are the parliamentary systems, al­
though one theme in this book is that we can never completely divorce an
election for an assembly from the executive-formation process. Indeed, we
have seen that parliamentarism does affect the number of parties by encour­
aging parties to act as potential "spoilers" or alternative governments in M
= 1 countries and to seek to enter coalition cabinets when M > l. There
are two regimes that come close to divorcing the concurrent presidential
so in the United States; enough voters regularly split their tickets to make divided control
the rule, despite concurrent elections. The critical difference in the U.S., compared to other
presidential systems discussed in this chapter, is the lack of a party list, meaning that party
leadership controls the election of members far less than in, say, Venezula. Under U.S.
circumstances, candidates tailor their messages to maximum effect for the districts, a fact
that Jacobson acknowledges even while emphasizing that party matters far more than is
allowed by those who have decried an "incumbency advantage" based upon constituent
servicing and campaign finance imbalance.
 Electoral cycles and the party system 241
election from voting for the assembly; it is not surprising that they are cases
of exceptionally weak presidents. One such case is Bolivia, where, as dis­
cussed in preceding chapters, the congress makes the ultimate selection of
the chief executive in a way that takes this case outside the class of presiden­
tial regimes. In Venezuela through the 1968 election, an interparty pact
mandated power-sharing cabinets and a president intended to be above
politics. In both these cases, Nv exceeded five. However, as soon as a
president in Venezuela chose to govern with a single-party cabinet - as he
was constitutionally permitted to do - Venezuela developed a two-party
system. This process was described in more detail in Chapter 9.
In considering cases where the interaction of executive and assembly
elections is weakened, we obviously cannot find the logical opposite of the
above cases, which would be an election for executive in which there is no
assembly - again, concurrent or otherwise. The desire of minor parties to
contest presidential elections, even though they have no chance of winning,
in order to attract support to a congressional slate is potentially present in
any presidential system in which M for the congress is greater than M for
the presidency. However, the contamination effect described here is
weaker when the presidential elections are nonconcurrent.

"Contamination" in concurrent versus nonconcurrent formats

Indeed, there is some support for the notion that the effective number of
presidential candidates is more likely to be as low as 2.0 when the congres­
sional election is nonconcurrent and the contamination effect of minor
parties from a concurrent congressional election is thus absent. In the data
presented in Chapter 10, we have fifteen nonconcurrent elections by (effec­
tive) plurality rule; four of those, or 27% , had NP of 2.0 or less. We have
twenty-five concurrent presidential-congressional elections; four of those,
or only 16% , had such low NP. 6 We thus have a bit of an irony: The greater
frequency of low values of effective number of presidential contenders is
found in precisely those cases in which the effective number of congres­
sional parties tends to be highest. Nonconcurrent elections are best able to
support separate party systems, mutliparty for congress and two-party for
president, as we suggested in Chapter 9 would be the case for an archetype
of presidentialism. Concurrent elections, on the other hand, bring about a
dampening of the number of parties in congress and, simultaneously, a
6 Recall, however, that the average effective number of presidential candidates was observed
to be higher under nonconcurrent elections (2. 7) than under concurrent elections (2.5). The
reason is that while fewer presidential elections in the cases of Brazil and Chile had the
pattern of several minor parties that occurs regularly under concurrent elections (hence
average N of 2.5), some had N of 3 or higher. In other words, the range of N values
observed under nonconcurrent elections is greater, given the greater frequency of elections
with three principal candidates, resulting from the occasional failure of party leaders in a
multiparty system to coalesce for presidential elections.
242 Presidents and assemblies
slight increase in the number of presidential candidates. In so doing, concur­
rent elections run some risk of lessening the ability of elections to both
branches to perform their functions best, as defined at the outset of this
book. The congress, despite PR, may be less representative, while there
may be additional presidential competitors, despite plurality. 7 Still, concur­
rent elections seem to strike a balance between efficiency and representa­
tiveness and are certainly superior in providing efficiency compared to a
system with all off-cycle congressional elections.

N O N C O N C URRE N T E L E C TORAL CYCLES

The preceding section, concerning concurrent elections, concluded with

observations about the interaction of elections to congress and president. In
cases of nonconcurrent elections it is feasible that there might be little or no
interaction. Such an expectation would be consistent with the archetype of a
presidential system that was presented in Chapter 9, in which the principle of
separation of powers is carried to the logical end of separation of campaigns
and party systems in presidential and congressional elections. Such extreme
separation could prevail only to the extent that congressional elections were
not conducted as referenda on the coalition or party controlling the presi­
dency at the time of the congressional election. As we shall see, noncon­
current elections often do appear to serve the function of referenda. There­
fore, with this format, too, there is an interaction between presidential and
congressional elections. How this interaction works depends to a great de­
gree on matters of electoral timing more specific than the simple dichotomy
between concurrent and nonconcurrent elections. We must also consider
honeymoon, counterhoneymoon, and midterm elections.

Honeymoon elections. We define any election that occurs within one year
of the president's inauguration to be a honeymoon election. Such elections
tend to be particularly beneficial electorally to the newly elected presi­
dent's party. A few regimes, including especially premier-presidential re­
gimes, allow a honeymoon election at the discretion of the president. Thus,
Frarn;ois Mitterrand twice dissolved parliament immediately after his own
election to call a honeymoon election. The new Chilean constitution gives
the president a similar prerogative.
7 At an extreme, this process might make the presidential election less efficient, since the
votes being cast for minor presidential contenders would cut into the share of the larger
parties. However, in the cases seen in Table 10. 1 , the normal vote shares of the two largest
parties are about 50% and 35%. The case with the highest magnitude favoring small parties
is Venezuela; yet the nationwide vote shares for the two major presidential contenders do
not tend to be on the low side in Venezuela. The equation just given, expressing NP in terms
of M, implies that with extremely high magnitude - and in the absence of smaller regional
districts as exist in Venezuela - these shares would drop appreciably; if so, we have further
reason to use the double complement rule we discussed in Chaper 10.
 Electoral cycles and the party system 243
Counterhoneymoon elections. If honeymoon elections are those that occur
within one year after the president's inauguration, it follows that an elec­
tion that occurs within the year before the presidential elections would be a
counterhoneymoon one. Such elections might be used by parties to provide
information on electoral trends that they can use in upcoming negotiations
over the formation of coalitions before presidential elections. This was
certainly the case in Chile (Valenzuela 1989a), and in Colombia, although
in the latter instance the effect works mainly on internal party factions that
present separate congressional lists (Hartlyn 1987). We shall have little to
say about this form in this chapter, because of its limited empirical scope.
However, we shall return, in Chapter 12, to counterhoneymoon elections
in our discussion of the role of electoral cycles in designing constitutions.

Midterm elections. Any election that does not occur within a year before a
presidential election or within a year after a new president has taken office
will be considered a "midterm" election. Thus some midterm elections do
not occur at precisely the middle of the president's term, although in many
countries they do. The United States and Ecuador, for example, have half
their congressional elections concurrent with presidential elections and half
at the exact midpoint between presidential elections. Argentina has concur­
rent elections and two midterm elections in each six-year presidential term,
one about two years after the presidential elections, the other two more
years on. Some countries, including Chile before 1973, France, and El
Salvador, may have assembly elections at varying points in the presidential
term, owing to terms for the two branches that are of different length.

Mixed cycles. Many systems have been quite haphazard in the occurrence
of elections of these various types. Nonsynchronized terms in some cases
have meant that some presidents experience a honeymoon election while
others do not. Some coalition-building negotiations in advance of presiden­
tial elections have been informed by parties' fates in counterhoneymoon
elections, while others have not. Our own view is that any of these formats
may be valid for maximizing particular political goals, but that consistency
from one presidential term to the next is desirable, in order not to treat
different presidents differently owing to a "luck of the draw." We return to
these political issues in Chapter 12, after we have completed our empirical
observations. Our empirical investigation of nonconcurrent elections con­
tinues now with Chile, a case that especially demonstrates the conse­
quences of haphazard electoral timing.

The Chilean experience

Presidential elections in Chile before the coup of 1973 were held every six
years. Elections for the congress were held every four years. All municipal
244 Presidents and assemblies
councils nationwide were also elected on a separate cycle, originally every
three years, later at four-year intervals. Never did elections to two or more
types of office coincide.
There was a honeymoon election in most presidential terms, but not
always for the congress. Some were for municipal councils. Furthermore,
Chile's varied election calendar meant that, in addition to the honeymoon
elections, either a congressional or municipal election would occur at mid­
term and that some kind of election would occur near the end of the
president's term (a counterhoneymoon). Thus, Chile offers an especially
rich environment for examining the impact of the timing of elections. Mu­
nicipal elections in Chile may be treated as equivalent to congressional
elections for our purposes, since they were contested on essentially the
same terms as the congressional elections. Before the emergence of "na­
tionalized" politics, localism prevailed in both congressional and municipal
elections. Later on, municipal elections became sort of "referenda" on the
ideological and partisan forces that disputed all other elections in Chile. 8
Also, for municipal and congressional elections nearly identical forms of
PR were used.
For Chile, Table 11.3 shows the percentage of votes obtained by the
president's party or coalition of parties in each presidential term for an
entire electoral cycle: the last election before the president was elected and
the three elections that would be held before this president completed his
term. The corresponding figure for the president himself is also shown.
This figure is not directly comparable to the congressional and municipal
data because it is partly dependent on the number of candidates running
and is a true personal vote, not a party vote. It is especially not comparable
for the height of the period of the inefficient secret (particularly in the
1930s and 1940s), when majority or near-majority presidential coalitions
were common and were clearly separate from congressional lists.
Indeed, before the emergence of nationalized politics in legislative races,
presidential elections had less effect on congressional support of the presi­
dent's party or coalition than in the 1960s and 1970s. There are two excep­
tions to this observation. The first is the Popular Front, a time of unprece­
dented reformism and labor organization, sponsored by the administration
of President Cerda (1938-42). Unfortunately, we lack data for the honey­
moon municipal election of 1939. W hat we can see is that the leftist and
Radical portion of the electorate grew from 1937 to 1941. In the other
presidential terms, the usual pattern was for supporters of the president to
experience a decline in support by the end of the president's term. Such
declines are especially evident in the terms of Ibanez and Frei, all the more
so because those presidents experienced major surges in support in their
honeymoons.
8 Indeed, even elections in schools and private associations were contested on partisan lines
(Valenzuela 1989b) .
Table 1 1 .3. Percentage of votes for parties supporting president during each term, Chile, 1938-1973

Cerda Rios Ibanez Alessandri Frei Allende

1938-42 1942-46 1952-58 1958-64 1964-70 1970-73
(Socialist, (Radical, (Agrarian Labor, (Conservative, (Christian (Popular Unity)
Radical, Socialist, "Ibailistas") Liberal) Democrat)
Communist) Falange)

Base (prior to Presidential 34.1 51.8 13.1 33.3 22.0 43.9

election) (1937) (1941) (1949) (1957) (1963) (1969)

Presidential election 50.5 56.0 46.6 31.2 56. 1 36.6

Honeymoon NA NA 45.4 none 42.3 48.6

(1953) (1965) (1971)

Midterm or later 54.7 45.0 7.9 31.4 35.6 44.0

(1941) (1945) (1957) (1961) (1967) (1973)

29.8
(1969)

Note: The presidential tenn running from 1946 to 1952 is omitted because the Communists, who had
supported the president's election, were subsequently proscribed.
Source: Urzua ( 1986).

Downloaded from Cambridge Books Online from within the IP domain of the University of California on Fri Jan 02 01:39:54 GMT 2015.
246 Presidents and assemblies
Using the last election of the preceding presidential term as a base, the
mean increase in votes for the president's supporters in the honeymoon
election was very large, 19 .1 % . In those same terms the percentage of votes
going to the party or parties supporting the president fell at the midterm
election by an average of 16.2% from its honeymoon peak. As Table 11.3
shows, the percentage of votes for the relevant party or parties fell back to a
level close to or even below its base point by the end of the president's term
in the cases of Rios, Alessandri, Ibanez, Frei, and Allende. Ifwe include the
Cerda administration in our calculations, the average net change from the
base election to the midterm is + 3.4%, but without that administration, the
figure is - 0. 1 % . So we could say that, excepting the period of rapid social
transformation and mobilization of the Popular Front, Chile's pattern was
for there to be, on average, no long-term change in the representation of the
several partisan blocs as a result of electing a president. Electing a president
would mean a temporary surge ( + 19% on average) if there were a honey­
moon election in the given term, but over the longer run, normal political
trends reasserted themselves.

The special effects of honeymoon elections. We now turn specifically to

honeymoon elections. Such elections, as we have already noted, tend to be
beneficial to the largest party or coalition from the presidential election,
but at whose expense? Here we shall see that it is the second largest party
that suffers the greatest losses. Another finding is that the third largest
party also gains in honeymoon elections. The explanation for this is that if
the usual tendency is for plurality presidential elections to polarize around
two major contenders, then some of the votes that go to one of the two
largest parties in the presidential election must have come from those who
would vote for smaller parties in elections for which PR is used. The
presidency, because of the single-seat district, imposes a restraint on third
parties when plurality rule is used. A nonconcurrent congressional or mu­
nicipal election with multi-seat districts releases this restraint (cf. Shugart
1988). The effect presumably could apply to fourth or lower-placed parties,
but in the party systems under examination here, parties ranking below
third either do not exist in the presidential election or are so small that the
effect is negligible.
Chile represents a "paradigmatic" case because of its multiparty system
and frequent elections, some of which were honeymoon. We can even in­
spect the effects of honeymoon elections on two different party configura­
tions. The presidential elections studied are 1964, when only two major can­
didates ran, and 1970, when there were three major contenders because of a
failure of party leaders to coalesce before the election. In 1964 the coales­
cence of the center and right restrained voting for the Radical Party's candi­
date such that the vote shares of the top (and only) three contenders were:
 Electoral cycles and the party system 247
Eduardo Frei 56. 1 %
Salvador Allende 38.9%
Julio Duran 5.0%

Many voters who usually would have voted for the Radical candidate,
Duran, instead voted for one of the two probable winners, Frei or Allende,
presumably most for Frei. The Radicals were the only one of Chile's tradi­
tional center-right parties (including the Liberals and Conservatives) not to
join the candidacy of Frei, the Christian Democrat, in order to block the
leftist Allende. The result was that many Radical voters deserted Duran in
the polarized campaign between Allende and Frei.
In the honeymoon congressional election of 1965, Frei's Christian Demo­
cratic party gained significantly over its level in the previous congressional
election, as shown in Table 11.3. However, because this party was only one
of the parties backing Frei, Frei's vote share cannot be attributed to the
Christian Democrats alone (as discussed in Chapter 9).
W hen we look at the other parties, we really see the differential effects of
the (effectively) plurality presidential election, in which there were only
two "serious" contenders, and the PR honeymoon congressional election.
The supporters of the runner-up Allende, the Communists and Socialists,
received a combined 22.7% in 1965, down from their 39% in the presiden­
tial election. On the other hand, Duran's radicals rebounded substantially.
While the party could only muster 5% in the polarized contest in the
presidential election, it managed 13.3% in 1965. We could even say that the
traditional Conservative and Liberal parties enjoyed the most spectacular
honeymoon surge of all, from 0% in 1964, when they presented no candi­
date of their own, to 12.5% in 1965.
The situation was rather different in 1970. The constitutional ineligibility
of Frei for reelection and the by-election successes of some right-wing
candidates late in Frei's term encouraged the Conservatives and Liberals,
now united in a new National Party, to go it alone (see Valenzuela 1978).
The Christian Democrats, emboldened by Frei's absolute majority in 1964,
refused to endorse a right-wing candidate, and Allende ran again, this time
with backing from a declining Radical party. Thus, with all three ideologi­
cal blocs presenting their own candidates and no cross-bloc endorsements,
each of the three candidates was perceived as having a real chance at
obtaining a plurality. Given the expectation, already discussed, that the
Chilean congress would confirm as president the candidate with the popu­
lar plurality, only slightly more than 33% of the votes would be needed to
elect the president. The actual results were:

Allende (Left) 36.6%

Alessandri (Right) 35 . 2%
Tomic (Christian Democrat) 28. 1 %
248 Presidents and assemblies
W hat happened in the 1971 honeymoon election? Allende's leftist sup­
porters experienced a honeymoon surge to nearly half the vote in the
ensuing municipal election (see Table 1 1.3), while the runner-up National
Party declined to 18%. The third-place Christian Democrats also declined,
but only slightly, winning 26% of the vote.
It is interesting to compare Allende's administration to the previous one
in Chile that was committed to major social transformation, the Popular
Front of Cerda. That Allende's supporters could obtain close to half the
vote in 1971 after Allende's having been endorsed by only 36% may appear
to represent a significant boost in support for Allende and perhaps for
socialism. However, the comparison to the previous election before Al­
lende's own reveals one of the smallest honeymoon surges shown in Table
11.3. Perhaps such a small surge is not surprising, given that Allende's
election was made possible only by the split in the nonsocialist forces in
1970. A combined Christian Democratic and rightist candidate would
surely have made 1970 a virtual replay of 1964.
We have now seen how nonconcurrent elections in Chile allowed for
fluctuating strengths for parties in a multiparty system. The reason why
there was no consolidation of the party system around the supporters of the
major presidential contenders was that in each Chilean presidential term
there were subsequent elections in which parties could test their strengths
independently. Later elections in Chile allowed parties outside the presi­
dent's coalition to reassert themselves in elections that used proportional
representation and thus had no bipolarizing tendencies. Valenzuela (1989a)
has observed that the frequent elections encouraged parties even within the
president's original coalition to distance themselves from the president.
Thus the electoral cycle in which there were honeymoon, midterm, and
counterhoneymoon elections allowed parties to demonstrate their potential
worth as future coalition partners in subsequent presidential contests. The
resulting fluctuations might even mask an underlying stability in voter opin­
ion, as Prothro and Chapparo (1974) argue was so in Chile between 1952
and 1973.

Honeymoon elections in Venezuela's revised electoral cycle

From 1958 to 1973, Venezuelan municipal elections as well as congressional
elections were held concurrently with the presidential election. 9 In recent
electoral cycles, however, separate municipal elections have been held
within the president's honeymoon period. This reform provides a valuable
test of the effects of honeymoon elections. Do the patterns observed in

9 Elections for all offices except president were fused, meaning there was no ticket-splitting at
these levels. Through 1988, this continued to be true for congress, senate, and state assem­
blies. For an analysis of the effects of these fused elections, see Shugart 1985.
 Electoral cycles and the party system 249

Table 1 1 .4. Percentage of votes for three largest parties in concurrent and honey­
moon elections, Venezuela, 1978-1984

Difference,
Congress Municipal concurrent
Party Pres. (concurrent) (honeymoon) to honeymoon
1978 1978 1979
AD 43.4 38.5 30.2 -8.3
COPEi 45.3 3 8.6 49. 1 +9.4
IU* 9.0 14.5 16.3 +1.8
1983 1983 1 984
AD 55.3 50.0 52.6 +2.6
COPEi 32.6 28.7 2 1 .7 -7.0
MAS 3.4 5.7 7.2 + 1 .5

Party abbreviations: AD, Acci6n Democratica (Democratic Action); COPEi, Christian

Democrats; IU, lzquierda Unida (United Left); MAS, Movimiento a Socialismo (Movement
to Socialism).
* Common list in the 1979 elections. For 1978 the component parties, the most important
of which was MAS, ran separately. The figures listed are the sums for the member parties.
Source: Official documents of the Consejo Supremo Electoral, Caracas.

Chile hold in a country that had been previously unaccustomed to such

elections?
Table 1 1 .4 shows the percent of votes obtained by the three largest
parties in the Venezuelan 1978-9 and 1983-4 electoral cycles. 1° Following
the Chilean example, we should expect to see in 1979 and 1984 a surge in
votes for the president's party in the municipal elections. In 1979 the new
president's party, COPEi, received 49. 1 %, a surge of 9.4% from the con­
gressional election (which was held concurrently with the presidential elec­
tions). The 1 .8% increase in votes for the combined left also is consistent
with results in Chile. W ithout a presidential contest to polarize the cam­
paign, minor parties fare better, just as the separate elections in Chile
helped support a multiparty system there.
In 1984-5 the party that won the presidency, Acci6n Democratica, had its
share of votes increase 2.6% from the congressional to the municipal elec­
tion and the left increased its share slightly. The AD vote in 1984 was actually
less than what the party's presidential candidate had received in 1983. This
10 For the 1989 elections, an analysis is more problematic because for the first time in
Venezuela, mayors and governors were directly elected along with municipal councilors.
Thus regional coalitions emerged for the first time since the crystallization of the two­
party-dominant system in 1973. With localism playing a significant role in the 1989
elections - a phenomenon not observed in 1979 and 1984 any more than in the concurrent
elections - comparisons of the 1988 presidential-congressional and 1989 regional elections
are difficult. See Shugart 1992a.
250 Presidents and assemblies

Table 1 1 .5 . Percentage of votes for party or parties supporting president in systems

using majority runoff for president

El Salvador France

Duarte Mitterrand I Mitterrand II

(Christian (Socialist) ( Socialist/ (Socialist)
Democrat) Communist)

Base 40.2 22.6 43.2 3 1 .2

(1982) (1976) (1978) ( 1986)

President 43.4 25.8 41.1 34.1

(1984) (1981) (1981) (1988)

Honeymoon 52.4 37.5 53.7 37.5

(1985) (1981) (1981) (1988)

Midterm 35. 1 3 1 .2 41 .0
(1988) (1986) (1986)

may seem anomalous, but it is not inconsistent with the combined effects of
concurrent and honeymoon elections. We expect that with plurality being
used for the presidency, a major party presidential candidate will receive a
higher vote share than its congressional slates. In the honeymoon election,
AD did gain again over its share in the previous election for which plurality
was not used (the congress), consistent with expectations. This result is
similar to that seen in Chile in 1964-5. Frei's Christian Democrats gained in
regard to the previous PR congressional election, while the party's candidate
ran far above the party's own usual share in the presidential race because
plurality had led to a two-bloc configuration.

Honeymoons after a majority presidential election

For systems using majority runoff and honeymoon elections, we have few
cases, but those cases fully support the results seen with plurality presiden­
tial elections, as shown in Table 11.5. Thus in the premier-presidential
system of France, as well as the presidential system in El Salvador, the
supporters of the president gain in honeymoon elections and lose in later
elections. W hether the president was elected by plurality or majority thus
does not affect the basic tendency of honeymoon elections. Some of these
 Electoral cycles and the party system 251
gains, and particularly that of the French Socialists in 1981, are especially
great.
After Frarn;ois Mitterrand was elected president in 1981 with 25.6% of
the first-round votes, his party's share of the votes for assembly increased
to 37.5% in the honeymoon election. It had been only 22.6% in the assem­
bly election before the presidential election. If we include the Communists'
votes, Mitterrand's electoral coalition achieved an absolute majority of
votes in the honeymoon election, despite previously having been in the low
40s, a level to which the coalition would return at the midterm. The pattern
for Jose Napoleon Duarte's Christian Democrats in El Salvador is very
similar.
These examples, as well as that of Chile in 1970-1, suggest that where
the presidential election is contested by multiple candidates, the honey­
moon surge is the greatest. Or, viewed from a slightly different perspec­
tive, the honeymoon surge after a two-candidate presidential election
might be said to have a lower "ceiling." The reason is that the election of
a president from among only two "serious" choices, as is usual in plurality
elections, means that some voters who normally support other parties in
congressional (or municipal) elections have contributed to the winning
margin in the presidential elections. Many if not most of these voters
might be expected to return to these parties in the next nonpresidential
election.
To sum up the effects of honeymoon elections, we can say that the
following tendencies are observed:
• The president's party gains over its share in the previous nonpresidential election;
• if the presidential election was itself a multicandidate affair (including the first
round of a majority runoff election), the president's party may gain even over its
own presidential candidate's showing;
• the vote share of the party whose candidate was the runner-up in the presidential
election declines;
• third parties regain their strength after having been suppressed in the case of a
two-candidate presidential race.

Long-term consequences of nonconcurrent election cycles

A question that might present itself at this point is whether there were any
long-term effects on the party system made possible by the election of a
president from a particular party. To investigate this question, we have
computed ratios that indicate the fluctuating vote shares for the parties
supporting presidents over time. These data are presented in Table 11.6,
while the raw data come from Tables 11.3 and 11.5. Obviously for this
exercise we can use only those presidential terms in which there were both
a honeymoon and a midterm election. Each ratio was computed by using as
the denominator the percent of votes won by the president's own party or
252 Presidents and assemblies

Table 1 1 .6. Ratios of votes in honeymoon and midterm election to votes before
presidential election, president's own party and whole coalition

(Base = 1 .00 for vote share in last election prior to presidential election)

President Honeymoon Midterm

Ibanez
Party
Coalition 1.44 0.60

Frei
Party 1 .92 1 .62
Coalition 1 .05 0.96

Allende
Party 1 .86 1 .53
Coalition 1.11 1 .00

Duarte
Party 1 .30 0.87
Coalition

Mitterrand
Party 1 .66 1 .38
Coalition 1 .24 0.95

coalition in the last election before the presidential election. What Table
11.6 makes clear is that, in the cases where it is possible to identify both a
particular party of the president and a coalition contributing to his election,
the honeymoon gains by the president's own party are greater than for the
coalition as a whole. Additionally, in the three cases for which we can
identify both a single party and a coalition (Frei, Allende, and Mitterrand's
first term), the president's party remained larger at the midterm than at the
base-point (although smaller than at the honeymoon). In each case , how-
 Electoral cycles and the party system 253
ever, the total coalition's support at the midterm returned to no more than
the level at which it had been before the president's election.
The overall effect is depicted graphically in Figure 11.4. Here are pre­
sented medians of the ratios from Table 11.6, but with the support of the
coalition additionally expressed as a ratio to the presidential party's base­
point support. The complete picture, shown in the figure, is that the elec­
tion of a president in a multiparty system leads to a realignment within the
bloc of parties supporting the president, but does not aid the overall coali­
tion. Thus, for example, the Socialists in both Chile and France benefited
electorally from having held the presidency, but they did so primarily at the
expense of their own coalition partners (Communists and Radicals in
Chile, Communists in France). Put another way, and in terms reflective of
a point we made in Chapter 9, it is costly not to hold the presidency, as
partners in the coalition are unable to reap the rewards of - and indeed
they suffer the costs of - incumbency. It is these incumbency costs - being
in the presidential coalition but not having the presidency oneself - that
make coalition formation in multiparty presidential systems difficult and
now and then lead to multicandidate presidential contests under the format
of nonconcurrent elections even in systems in which the president is elected
by plurality.

M I N O R PARTIES A N D E LECTORAL C YCLES

The foregoing discussions of both concurrent and nonconcurrent elections

suggest that the electoral cycle employed makes quite a difference in
whether minor parties are able to achieve much representation. Such parties
are considerably "more minor" under concurrent elections (assuming a plu­
rality presidency, of course) than under nonconcurrent, including honey­
moon, elections. We conclude our overview of the effects of electoral cycles
by considering the institutional context in which leaders and (potential)
voters of such parties must operate. We shall speak in terms of the payoffs
and strategies for such actors, drawing upon the literature concerning party
behavior (see Strom 1990). We shall assume that payoffs can come in the
form of officeholding and policy. At times the two forms of payoff may
conflict, as when a party's policy preferences are long-term and thus lead it to
adopt a platform that does not make the winning of many assembly seats a
viable course for the party in the short run. For the purpose of this discus­
sion, we shall assume rather efficient (policy-oriented) voting. Such voting
is, as discussed in Chapter 9, hindered to the extent that voting for congress
takes place on a particularistic and personalistic basis, rather than on the
basis of party platforms.
In a concurrent election format in which the president is elected by plural­
ity and the congress by PR, what are the payoffs and strategies for a party
that is not a serious presidential contender, owing to its relatively small
 3 --r-------------------------------------�
President's party

-
- -- - Presidential coalition

2
2.14
-- - - - - -- -- -- -- 2.44

-- -- -- --
2.04
1.76

Base Honeymoon Midterm

Figure 1 1 .4. Median ratio of votes for president's party and coalition in honeymoon and midterm elections to votes
for president's party in last election before presidential election (base = 1 .00)

Downloaded from Cambridge Books Online from within the IP domain of the University of California on Fri Jan 02 01:39:54 GMT 2015.
 Electoral cycles and the party system 255
following? In the presidential race, there can be no immediate office pay­
off, because by definition the party is too small to win. In the congressional
races, there is such a payoff, as PR will allow it to win some share of seats
approximating its share of votes.
W hat about policy payoffs? In the short run, policy requires the ability to
win office, although not necessarily executive office. There is likely to be
some uncertainty about the short-term payoffs. If one of the major parties
does not win a majority of seats, the third party may play a critical role in
forming legislative majorities and may thus obtain some portion of its
policy agenda. If the president's party does win a majority in the assembly,
there may be few opportunities for minor-party influence over policy, espe­
cially if closed lists are used.
On the presidential side, the short-term policy payoffs are as certain as
the office payoffs, as both are zero if the party cannot win the presidency.
However, the long-run payoff is that the party may hope one day to be part
of a presidential coalition. A party may not appear to be a serious con­
tender for national power if it does not present a presidential candidate. It
is hard to say, "Here is what we would do if the people entrusted us with
the presidency," while running only a congressional slate. It is especially
difficult given the different purposes of the two branches, and the more
immediate and perhaps more mundane concerns of voters that congres­
sional candidates must frequently address in lieu of grand policy options.
Because of the greater uncertainty regarding elections years hence, the
long-run payoffs of one day influencing policy via the executive are sure to
be discounted, compared to the short-term benefits of winning congres­
sional seats. The certainties in the short run are that the party will win some
seats in the congressional races and that one of the two big parties will win
the presidency. Thus, rationally, many voters who like the minor party's
program will split their tickets. And party leaders may even encourage this
to some extent, or at least remain indifferent to it.
The short-term cost is that by running and voting for a presidential
candidate, the minor party and its followers risk tipping the presidential
race to their less preferred option among the two serious presidential con­
tenders. 11 Thus there are rational motivations that predict that a minor
11 Thus, ticket-splitting should be greatest when the race between the two big contenders is
expected to be close. In a close race, the short-term costs of not taking sides in the
presidential race are greater (assuming that the party's leaders and voters are not perfectly
indifferent to the choice between the big two). Additionally, given an expectation of
closeness, the potential payoff of holding the congressional balance of power is great.
However, while a close contest between the big two should depress the voting for president
by those who favor the minor party, closeness may make it more difficult for the minor
party to draw votes from marginal supporters in the congressional race. It is possible in
ordinary elections that some voters whose choice of general policy direction is best articu­
lated by a big party might cast a vote for the minor party for congress (perhaps to express a
256 Presidents and assemblies
party will win presidential votes, even when it has no hope of winning the
presidency, and that there will be ticket-splitting. However, based on the
two-party dominance that prevails in the congressional election and the
theoretical reasoning regarding ticket-splitting that we have offered, it ap­
pears that ticket-splitting is practiced more by voters who prefer minor
parties but want to cast an effective vote for president, than it is by voters
whose first preference is for a major party presidential candidate. If the
latter form of ticket-splitting were more common, third parties should
perform better in the congressional election. Third parties' fates in concur­
rent elections are thus more determined by the constraint of a plurality
presidential election than they are by the opportunities of a PR congres­
sional election.
In honeymoon elections ( or midterm elections in systems without concur­
rent elections) , we notice that "third" parties perform better than in concur­
rent elections. This was observed both cross-nationally (Brazil and Chile
compared to Venezuela and Costa Rica) and within one country where
both concurrent and honeymoon elections have been held (Venezuela).
Why should we observe this pattern? For some voters who supported the
runner-up in the presidential election but who are not committed followers
of that party, the third party may be attractive as a way to continue one's
vote against the president while not backing a perceived loser. There could
even be a policy motivation. The likelihood that the president might cooper­
ate in congress with leaders of the party of his or her defeated opponent
may be lower than the possibility of cooperation with the third party, which
is less threatening. Such voters may perceive an opportunity to provide a
more plausible check on the president by voting for a party other than the
second-largest party.
For minor party leaders, honeymoon or midterm elections offer a way
out of a conundrum presented by the concurrent election format. In concur­
rent elections, if they seek to maximize their forward-looking policy pay­
offs by emphasizing national policy differences with the big parties, they
may lose votes for congress. After all, we are speaking of a party whose
fundamental policy alternatives must not be appealing to great numbers of
people, given that it is a small party. So the party also seeks to maximize its
congressional vote. How does it do that? To some degree, its candidates
may seek to address more local, mundane, and immediate concerns of
voters in their regions. The party is thus caught trying to do two things at
once: provide a long-run policy alternative in a race it cannot immediately
more qualified endorsement, or for reasons of local particularism). But in a close race,
these same voters may prefer not to risk that their preferred party fails to control congress
and thus may not vote for the minor party. Only with ballot-by-ballot analysis (to count
actual split ballots) and by survey research ( to find out what party voters really feel closest
to, as done for Spain by Gunther 1989) could we really disentangle this issue.
 Electoral cycles and the party system 257
win, and appeal to a constituency in races it can win and has won in the
past. 12
In a nonconcurrent election, the party can focus its resources solely on
winning congressional seats and speak to voters in terms of what it can do for
them now. Much of that will concern mundane items, but policy themes need
not be excluded. In a honeymoon election, once the congressional election
takes place, it will already be known who is president (and for how long) and
what that individual's policy commitments are. The third party can promise
either cooperation or vigorous opposition, whichever is likely to maximize
its votes and influence. If it gains sufficient congressional representation, it
can then seek to win accommodation from the president. In a counter­
honeymoon election, the minor party can seek to demonstrate its value as a
coalition partner for the upcoming presidential election. These observations
recall those of Chapter 9 that nonconcurrent elections let the campaign for
assembly take place on its own terms. While the presidential outcome influ­
ences a honeymoon election, such an election is at least not overshadowed
by a simultaneous presidential contest. For the minor party with its mixed
and sometimes contradictory short-run and long-run goals, taking the assem­
bly contest out of the immediate influence of a presidential race that it
cannot win lightens its load and reduces its contradictions.

CONCLUSIONS

We have seen that there are special problems raised by the combination of
presidentialism and proportional representation, yet most presidential sys­
tems have elected their congresses by PR. The original presidentialism,
that of the United States, and its one former colony, the Philippines, are the
only two presidential democracies to have used plurality rule for their
congressional elections for a long period of time. 13 It is important to realize
that a format that combines plurality and PR, as do most presidential and
premier-presidential systems, may not be the best way to maximize both of
the goals that we suspect (there apparently being no "Framers' Memoirs"
on this topic) were in the minds of those who opted for such a format:
• a broadly representative congress for which parties can run without the need to
form coalitions first (hence PR) ;
12 We refer here to the addressing of constituent interests by the party as a whole, since we
are assuming closed lists, which do not allow particular candidates to cultivate votes
independently. Many electoral systems are somewhere in the middle of this implied contin­
uum between highly candidate-centered voting (Brazil, Colombia) and party-centered
voting (Venezuela), as suggested by Table 9.1.
13 Nigeria also has used this combination in its brief democratic experience. Mexico, which
may be in a transition to democracy, also uses this combination, but not in a pure form, as
there are seats reserved for minority parties that do not win in single-seat districts.
258 Presidents and assemblies
• an executive under the control of one party and endorsed by a majority (hence
presidentialism, and more particularly, plurality elections whenever there was a
dominant political force at the time of adoption of the rule) . 14
Not surprisingly, if we seek to maximize these goals at once, we are likely
to fall short. We may even get at times the reverse of these goals. For
example, when concurrent elections are used, we may not get a faithful
representation in the congress, owing to the powerful incentives of minor­
ity political tendencies to join with one of the presidential contenders'
parties. Yet because PR encourages a multiparty system, some presidents
under plurality rule may be endorsed by considerably less than a majority.
Ironically, however, majority runoff, the institutional fix frequently chosen
to redress this problem, tends to increase fragmentation rather than to
induce consensus. Worse, it makes lack of presidential-congressional har­
mony highly likely, as do nonconcurrent electoral cycles.
We next take up the question of "engineering" electoral cycles, suggest­
ing alternative institutional designs that may do a better job at serving goals
such as those just mentioned. These alternatives involve being cognizant of
how the choice of a presidential or premier-presidential regime interacts
with the electoral cycle to maximize particular political goals.
14 Where no one party dominates the initial rule-making and each of several parties perceives
a likelihood of winning sole control over the executive, majority runoff may be a more
likely result, as in Ecuador and Peru, and the recent constituent assembly in Colombia.
Or, if plurality is chosen, nonconcurrent election cycles would be expected in order to
enhance parties' independence vis-a-vis the executive. These situations would also seem
most likely to produce restrictions on reelection, in order to ensure alternation. We are
inspired to undertake these suggestions by Geddes (1990), but we recognize a need for
further work on the question of political origins of institutions.
 12
Elector al cycles and comp atibility between
president and assembly

We have seen how presidentialism permits a nationally endorsed execu­

tive while still allowing the representation of minority or regional interests
to be expressed through the assembly. We have also seen that premier­
presidentialism entrusts presidents, with their electoral mandate, to form
a government, but only if the government can hold the confidence of the
assembly. There are, of course, no guarantees that the president's party
or coalition must hold a majority of seats. This chapter will consider
electoral cycles in presidential and premier-presidential regimes and how
they may effect patterns of governance. The longest-lived premier-presi­
dential regimes (e.g. , France and Finland) have had their elections and
terms for president and parliament on wholly different cycles, as have
some presidential systems (e. g . , Chile before 1973 and the current regime
of Korea). Presidents in such systems are elected for longer terms than
are parliaments, although in most of the premier-presidential systems the
president has the option to dissolve parliament and call new elections. In
France, the president also may shorten his own mandate and call a new
presidential election. Even when such options exist, an important charac­
teristic of actual premier-presidential regimes is that there is no constitu­
tionally prescribed harmony in the timing of elections and therefore
terms, despite the importance that electoral cycles were shown in Chapter
10 to have. Under the nonconcurrent cycles and varying-length terms used
in these systems, some assemblies are likely to be elected in the presi­
dent's honeymoon and others at midterm or late in the term. Even if a
newly elected president possesses and uses the power to dissolve parlia­
ment and should happen to achieve a majority to his or her liking in new
(honeymoon) assembly elections, having terms of different length still
means that this president's mandate outlasts that of the assembly (and the
government that may be responsible to it). Compatibility may not always
be necessary or even desired. But we caution that the polar opposite of
compatibility, cohabitation between a president and premier who are one
another's principal political opponents, or "divided control" between
president and congress, may be problematic in many settings. In a presi­
dential system, we suspect, incompatibility between coalitions and parties
260 Presidents and assemblies
holding the executive and assembly may be even more a problem, as
presidentialism offers less clear-cut mechanisms for resolving such differ­
ences. We take up later in this chapter potential scenarios in which incom­
patibility might be acceptable or even preferred in premier-presidential
systems. First, let us consider the logic of compatibility and electoral
cycles likely to produce it.

QU A S I - P R E S I D E N T I A L I S T E L E C T O R A L F O R M A T S

If the presidential election i s to be efficient in terms of leading to actual

policy output akin to what was endorsed in the presidential election, there
should be some institutional means of making likely that president and
assembly will be at least roughly compatible in their partisan persuasion.
Otherwise the attempts of the president to pursue a legislative program are
likely to be checked by an opposition-controlled assembly and, in a
premier-presidential system, also by the cabinet itself. We thus would want
to avoid cohabitation or divided control.
We shall refer here to "quasi-presidentialist" electoral formats. The defin­
ing characteristic of such formats is that the presidential election exerts a
strong "pull" on the assembly elections - the phenomenon that we dis­
cussed at length in Chapters 10 and 11. The reason for this presidential pull
is traced to the electoral cycle. Thus quasi-presidentialist formats would
have either concurrent or honeymoon elections for the assembly as well as
terms for president and assembly that are synchronized. Such formats
make likely an assembly majority (and the cabinet, if it is responsible to the
assembly) in harmony with the partisan alternative endorsed in the presi­
dential election. Midterm elections might seem desirable in order to pro­
vide some check on the presidential majority. However, given the demon­
strated tendency of midterm or later elections to result in defeats for the
presidential coalition, such elections are likely to counteract the effect of
the quasi-presidentialist format, and we do not recommend them for most
circumstances. As seen in France in 1986-8, the presidential majority
might suffer a midterm electoral defeat even when the president herself or
himself proves capable of subsequently being reelected.
In a later section, we suggest an alternative, to be called "quasi­
parliamentarist," in which the assembly election does not concur with or
follow upon the presidential election, and so is not subject to the presiden­
tial "pull." In most quasi-parliamentary formats incompatibility is thus
more likely. Nevertheless we can envision a format in which it can be made
less likely, without giving temporal precedence to the presidential election,
as is done in the quasi-presidentialist formats. We shall also discuss systems
that employ hybrids of these two basic formats.
W hatever the electoral cycle employed, we reiterate that the president
should be elected by plurality or the double complement rule that we
 Compatibility between president and assembly 261
discussed in Chapter 10. Throughout the following discussion, when we
refer to presidential elections, we mean one of these methods, unless
othewise indicated. In this manner we can approximate the ideal that the
executive power should be formed out of a broadly endorsed electoral
program.

Concurrent elections
If the elections were concurrent, we would expect two major parties, but the
party winning the presidency would not necessarily have a majority of seats
in the assembly, especially if a high magnitude were used for assembly elec­
tions. This has been the pattern in the presidential regimes of Venezuela and
Costa Rica. Because we have already discussed this format in considerable
detail with regard to presidential systems (Chapter 11), let us consider the
implications of concurrent elections under premier-presidentialism.
W ith the concurrent elections, the ordinary result would be that the
president could appoint a member of his or her own party as premier. No
doubt that government would sometimes be a majority government, but at
other times it would be either a minority cabinet or a coalition. Is this
advantageous? We have sought, in suggesting this as a possible institutional
design, to create both the ability of the voters to designate a general policy
direction and to have minority viewpoints represented meaningfully. The
first criterion may be met by the electoral method used to elect the presi­
dent. As for the second criterion, minority views must be taken into ac­
count in forming governments whenever the president's own party is short
of a majority in the assembly.
One actual regime that comes close to the model just broached, in some
aspects, is that of Namibia.1 Under the Namibian constitution of 1990, the
president and assembly are elected at the same time, with the president
appointing a premier and cabinet that are subject to parliamentary confi­
dence. There are important aspects in which the actual Namibian system
deviates from our proposal, however. The first concerns presidential pow­
ers. The president as well as the assembly may dismiss cabinet ministers,
and there is also a presidential veto. As suggested earlier, the lack of
ultimate authority by one or the other elected branch over the cabinet is
often problematic, and the veto means that even a government controlled
by the president's opposition can not be sovereign over legislation. An­
other deviation from our suggested regime design is that the Namibian
constitution specifically requires that a majority system be employed for
1 Haiti's 1990 elections were concurrent. However, if the regime stabilizes, concurrent elec-
tions will occur at most only every twenty years, since the constitution provides for five-year
presidential terms and four-year assembly terms (and is silent about concurrence in years
when both types of election occur). Romania's 1991 premier-presidential constitution man­
dates four-year terms for both president and assembly.
262 Presidents and assemblies
the election of the president. If the party formed out of the former guerrilla
movement, the South-West Africa People's Organization (SWAPO), re­
mains dominant (it won 57% of the vote in 1989), this provision will be
unimportant. Indeed, the majority election may make it more difficult for
other parties to present a common alternative than if the electoral method
were plurality, especially given a district magnitude for the assembly that is
very high (72). Moreover, the concurrence of parliamentary elections with
what would be merely the first round of a majority runoff presidential
election could lead to a degree of party fragmentation resembling that seen
in a country such as Peru or Ecuador (Chapter 10). If this is the case, and
SWAPO ceases to be a majority party, then the system could foment con­
flict between the president and assembly over cabinet composition. We
thus could have incompatibility despite concurrent assembly elections be­
cause of the majority runoff election of the president.
The election of the president by majority runoff, as in Namibia, coupled
with the importance of the assembly in forming the government might very
well inhibit the consolidation of an alternative party (or bloc) to compete
against SWAPO. W ithout this consolidation, the presidential election
would fall short of our ideal of efficiency. Is there also such a risk even with
plurality, as long as the assembly plays such a role? It could be argued that
parliamentary confidence in a premier-presidential regime would inhibit
the consolidation of parties around two major contenders, despite the con­
current election and the use of plurality for president. Indeed, we have
seen one presidential system with a plurality presidency in which there was
a period of clear assembly supremacy: Venezuela before 1968.
If the source of the multiparty system in Venezuela through 1968 was
indeed the power that party leaders gave congressional elections to deter­
mine the membership in government coalitions, as suggested in Chapter
11, on what basis can we argue that a premier-presidential regime would
not result in the same? That the Venezuelan congressional elections should
have come to be dominated by two parties despite the superior constitu­
tional position of the assembly over legislation is testimony to the power of
the pull toward bipartism exerted by the plurality presidency, once parties
have focused on winning this office as their major objective. 2 In a premier­
presidential regime, too, given the president's power to propose a premier,
the presidential campaign would be likely to bring about two-bloc domi­
nance of the concurrent assembly election, although important third parties
might be expected to hold the key to coalition making. While assembly
representation takes on great importance here, as it does in basic parlia­
mentarism, the contestation of the presidential election and the concurrent
assembly election around themes of national policy significance would pro-
2 As discussed in Chapter 10, this was not effectively the parties' goal in the earliest years of
the Venezuelan regime, when an interparty power-sharing pact had weakened the role of
the presidency and made being in "opposition" less costly.
 Compatibility between president and assembly 263
vide powerful incentives for most political actors to line up with a coalition
capable of winning the presidency.

Possible disadvantages of the concurrent cycles. The system of concurrent

elections would tend to maximize the effects of the normally two-party
presidential competition on the congressional representation. What if we
do not desire this dominance of two, probably catchall, parties in the
assembly, but still want the presidential coalition and the assembly majority
to be ordinarily in consonance? As we saw in Chapter 11, in nonconcurrent
elections we would anticipate a greater presence for (larger) "third" parties
than with the format of concurrent elections. Such parties, by finding it
easier to maintain their separate identity as significant parties, might be in a
better position to influence the structuring of future presidential coalitions,
even if at some given moment they are outside the government.

Honeymoon elections
Data presented in Chapter 11 showed that honeymoon elections are more
likely than other electoral cycles to produce an assembly majority for the
just-elected president. At times this may even be a very large majority, with
the second-largest party (as determined by the presidential runner-up)
falling well behind. This pattern is seen most clearly in Venezuela's 1979
and 1984 honeymoon elections for municipalities that were elected on the
same party-list basis as the concurrent congressional elections. These hon­
eymoon contests produced a surge in support for the newly elected presi­
dent's party. In the case of the premier-presidential regime in France,
Frarn,ois Mitterrand obtained a majority in 1981 and thus could appoint a
Socialist premier, thanks to the honeymoon election. In 1988, in another
honeymoon election following his own reelection, Mitterrand was again
able to obtain a government consistent with what he proposed to the elec­
torate. Calling for an "opening to the center," Mitterrand was able to
appoint a minority Socialist government dependent for support on centrist
tendencies that had supported him instead of the Gaullist candidate in the
second round of the presidential election (and the cabinet occasionally
relied upon the support of Communists, who also had supported Mitter­
rand in the runoff).
This experience illustrates how a honeymoon election might preserve
sufficient third-party strength to provide alternatives to the president's
party. However, it should be remembered that majorities for the presi­
dent's party frequently do result from honeymoon elections, thus meaning
that the alternatives preserved by this format may be future-oriented, while
governance will tend to be majoritiarian. Unlike the concurrent format,
however, in which the third party's representation may be very limited, the
honeymoon format is likely, as shown in Chapter 11, to reduce the gap
264 Presidents and assemblies
between second and third parties, thereby potentially anticipating a future
coalition to oust the (currently) dominant party.

Q U A S I - P A R L I A M E N T A R I S T FORMA TS

The previous discussion considered electoral cycles likely to give the presi­
dent an assembly majority influenced by the voters' choice for president.
Whether concurrent or honeymoon elections are used, both formats have
in common a potentially dominant role in legislation and cabinet-formation
for the newly elected president's party. In some other settings, it might be
desirable to give primacy to the parties and their ability to garner votes in
assembly elections without the overwhelming importance of the presiden­
tial election. To weaken the presidential effect on assembly elections in this
way implies using either counterhoneymoon or midterm elections. Coun­
terhoneymoon elections may preserve the advantages of presidential­
assembly compatibility without making assembly elections overshadowed
by presidential elections; midterm elections, on the other hand, are almost
certain to limit the prospects for compatibility.

Likely compatibility preserved: counterhoneymoon elections

An election held as a counterhoneymoon is one that closely precedes a
presidential election. In the institutional pattern to be discussed here, the
terms of the president and assembly are synchronized, but the elections
themselves are on separate dates.
Suppose that there were a counterhoneymoon election, perhaps with
rather large M for the assembly. The presidential election then would
follow soon afterward, within a few months, as was formerly the case in
Colombia. With more efficient congressional elections than in Colombia
and without Colombia's factionally based electoral system, a multiparty
system would be likely. However, given the use of plurality or the double
complement rule to elect the president, parties might anticipate an advan­
tage in lining up behind two basic blocs to elect the president. Assembly
elections might be said in such a format to function as almost a "primary"
for the presidential election (much as they have often done in Colombia for
the various internal factions of the two traditional parties). It is even possi­
ble that such a format might produce two rather large blocs among the
parties contesting the congressional elections, thus serving a function simi­
lar to that of the first round of majority presidential elections in France.
However, even if the party system did not exhibit the French two-bloc
pattern, but instead featured two large parties plus a third one needed to
achieve a majority, this institutional package would have potential advan­
tages over either a two-round majority presidential election or a pure
parliamentary system.
 Compatibility between president and assembly 265
Unlike a two-round majority system when there are three or more blocs,
the eventual (second-round) presidential victor would not be so dependent
on first-round contingencies such as the precise number of candidates and
which ones happened to finish first and second in the preliminary round.
Instead, there would be time for negotiations leading to the presentation of
two principal presidential candidates after the voters had been given the
chance to express partisan preferences in a nonpresidential election. The
advantage over parliamentarism is that these negotiations among the parties
would still have to be ratified by the electorate in the presidential election,
unlike the products of coalition-making negotiations in parliamentary sys­
tems. There would thus be an electoral check on the parties' coalition­
building negotiations that would take place after the assembly elections.
The format of counterhoneymoon elections is called quasi-parliamentarist
because of the primacy of the assembly election and parties represented in
the assembly. The direction of the interaction in this format is principally
from assembly election to presidential, rather than vice versa, as in a quasi­
presidentialist format. Primacy is also accorded to parties and assembly
elections in those regimes that employ midterm elections. The difference
between systems using counterhoneymoon elections and those that would
employ exclusively midterm elections is the diminished likelihood of
presidential-assembly compatibility.

Midterm elections
As already suggested, midterm elections go the farthest in weakening the
impact of presidential elections over elections to the assembly. Also, com­
pared to counterhoneymoon elections, midterm elections are Jess likely to
produce a coalition that will be presented to voters in the following presi­
dential election. Midterm elections, then, are conducted "on their own
terms" to the fullest extent possible. Moreover, we have seen in Chapter 11
that the fortunes of the president's party tend to fade in midterm elections.
Thus we would expect a premier-presidential regime after a midterm elec­
tion to be premier-dominated rather than president-dominated, while an
opposition-controlled congress would typify a presidential regime.
Most actual presidential or premier-presidential regimes that have em­
ployed midterm elections have been inconsistent. Presidents in such re­
gimes have sometimes governed with assemblies elected under the influ­
ence of their own presidential elections (e.g. , Mitterrand 1981-6, Frei
1965-9, and Duarte 1985-8) and at other times with an assembly elected
apart from the influence of the presidential election (Mitterrand 1986-8,
Allende 1970-3). In Argentina, Ecuador, and the United States, every
president governs with both types of assembly, given a frequency of con­
gressional elections that is half (or in Argentina, one-third) the length of
the presidential term.
266 Presidents and assemblies
If we want t o ensure that the minimizati on of the presidential effect
applies t o every assembly electi on, all such elections must be midterm.
There is s ome question, h owever, about the value of having a politically
important presidency if an assembly elected at ar ound the midterm were
the b ody entrusted with forming a cabinet, as in a premier-presidential
regime. And in a presidential system, the efficiency of the presidential
electi on is severely c ompr omised when the president is alm ost sure t o l ose
legislative support after the midterm. F or this reas on we dismiss midterm
electi ons in (pure) presidential systems as being undesirable. We maintain
that where the elect oral cycle makes a regime disp osed t o cost the president
the likelih o od of a majority, there sh ould be an institutional means of
rec onstituting the g overnment. Because only premier-presidential regimes
aff ord such means, the remainder of this discussion will be dev oted only t o
premier-presidential systems.
If we are t o rec oncile the c onflicting mandates expressed thr ough off­
cycle elections t o the tw o branches that express different p olicy directions
for the p olitical system, it f ollows that the president's p owers sh ould be
sharply circumscribed. Such a president's r ole w ould c onsist mainly of
being the one wh o designates which party leader shall seek out c oaliti onal
pr ospects within the assembly. A regime of this type with off-cycle electi ons
d oes n ot seem t o us t o be particularly viable where there is a clear
g overnment-opp ositi on f ormat t o the party system, that is, where there are
tw o parties or at least tw o bl ocs. H owever, if there is a multipolar party
system in which it is quite feasible that the president's party w ould be a part
of m ost g overnments, this presidential r ole c ould add a degree of acc ount­
ability and efficiency absent fr om m ost multiparty parliamentary systems.
Indeed, a premier-presidential regime with such a f ormat w ould allow the
regime t o function the way the designers of the French c onstituti on ex­
pected their system w ould functi on, but seld om has: with shifting majori­
ties (Pierce 1990).
It is the ab ove-menti oned presidential r ole that, in our interpretati on,
has been accorded presidents in the premier-presidential regimes of Fin­
land and P ortugal (since 1982). In the French case, on the other hand, the
two-bloc nature of the party system at least thr ough Mitterrand's first term
has made f or a pattern of cabinets either basically c ompatible with the
president, or else diametrically opp osed t o him (as in 1986-8). Because the
l onger-lived premier-presidential regimes have empl oyed midterm elec­
tions, and theref ore at least phases of quasi-parliamentarism, let us briefly
c onsider the implications of this f ormat in Finland and France.

Finland. The Finnish experience, particularly during the presidency of

Urh o Kekk onen, gives us a taste of the r ole of a president in a premier­
presidential regime in which there is neither a tw o-bl oc division of the party
system n or an expectation of presidential-cabinet c ompatibility. H owever,
 Compatibility between president and assembly 267
Finland also cannot be considered a case of cohabitation, since by that term
we mean a president whose party is shut out of the cabinet owing to an
assembly controlled by the president's opposition. Kekkonen was associ­
ated with the Center (originally called Agrarian) Party, even though he
frequently secured the endorsement of most other major parties as well.
During the presidency of Kekkonen the Center Party was in government
for all but 12% of the time (based on data in Strom 1990). Since Mauno
Koivisto became president, his Social Democrats have been in government
all the time from his election in 1982 until mid-1990. W hile both presidents'
parties have regularly been in government, and thus cohabitation has not
been the norm, coalition partners have frequently shifted, such that aver­
age cabinet durability during the period has been just over twelve months
(Strom 1990:251-2).
The role of the presidency in Finland thus might be seen as a primarily
regulative or arbitral one. This is especially visible during the period in
which Kekkonen was supported by all major parties. In as complex and
fractionalized a party system as Finland's, the parties have, in effect, ceded
authority over government formation to the president. To comprehend why
they would do so requires an understanding of both the difficulties of
coalition building in Finland and the peculiar institutional position of the
Finnish president.
Not only have changes of government been more frequent than in most
Western parliamentary systems, but also the government-formation pro­
cess has been to only a very slim degree informed by elections. Strom's
identifiability rating for Finland is zero, meaning that voters are never able
to discern and choose from among alternative cabinet possibilities before a
parliamentary election. In this context, a president fills a need for a figure
to take responsibility for arbitrating among parties. Under Kekkonen, par­
ties regularly delegated to the president initiative in the formation of coali­
tions, effectively giving him powers of confidence over the government that
are not formally part of the constitution.
In pure parliamentary systems, when there is no clear choice stemming
from elections as to which party should make the first formation attempt,
the role of designating a formateur (or proto-premier) falls occasionally to
a monarch or speaker of the house, but more often to a president. The
difference in Finland is that this president is chosen through an electoral
process, rather than by the parties directly in parliament (perhaps with an
extraordinary majority), as are the presidents of Greece and Israel, for
example.
The Finnish procedure, at least up to 1988, made presidents inherently
more dependent on parties than any procedure short of simple selection in
parliament. An electoral college in which voters vote for party along the
same format as in parliamentary elections, and in which brokered presiden­
tial elections are at least theoretically possible, gives parties primacy. (We
268 Presidents and assemblies
might even speak of "quasi-parliamentarist" presidential elections.) Using
this institutional arrangement, parties are able to ensure that no president
who had even limited powers of designation would be unresponsive to
them. Moreover, the lack of limit on reelection has meant that the presi­
dent remains accountable at the end of the term to the parties through the
next election for delegates to the electoral college. Multiparty endorse­
ment, with resulting extraordinary majorities, coupled with the party-list
format of presidential elections thus has given the parties control over the
president to a degree approximating that of many pure parliamentary sys­
tems. If, in the post-Kekkonen era, presidential competition returns, there
may emerge a degree of presidential accountability to the electorate that
was lacking during Kekkonen's long rule. The recent steps toward giving
the president some personal electoral connection to the electorate (see
Chapter 10) should hasten this process. If so, greater electoral efficiency
may emerge in the process of choosing who shall choose - meaning that
voters might be presented with differing alternative themes that the presi­
dent, once elected, would attempt to have reflected in the interparty
government-formation process. Not surprisingly, however, this potential
for more popularly accountable presidents has also ushered in a constitu­
tional debate in which parties may seek to curtail the president's formal
powers further and impose limits on reelection. 3

France. Given the two-bloc nature of the French party system before
Mitterrand and through (at least) his first term, it is somewhat of a surprise
that cohabitation did not occur earlier. Closely fought elections in 1973 and
1978 nearly produced assembly majorities that would have represented the
exact opposite governmental preference from that of the president.
Indeed, were it not for the constitutional right of the French president to
dissolve parliament, Frarn;ois Mitterrand in 1981 would have been required
to appoint a premier acceptable to a conservative-dominated assembly that
had been elected in 1978. 4 There would not have been an assembly election
until 1983. We cannot predict what kind of majority would have been
returned in that election, but it might not have been socialist. The only
socialist governments to date in the French Fifth Republic have resulted
from an institutional structure that permits or even encourages an impact
of presidential elections on assembly elections by permitting the president
to call a honeymoon election, as happened in 1981 and 1988.
3 Markku Suksi (in personal communication) indicates that in the 1988 election, which used a
ballot with presidential candidates' names for the first time, candidates toured the country
to court voters, a practice not ordinarily seen in previous Finnish presidential election years.
On the reform debate, we rely on Berglund and Worlund (1990) and personal communica­
tion with Sten Berglund.
4 This, of course, ignores previous dissolutions as well as shortened presidential terms that
indeed had already altered the electoral calendar from the basic "7-5" plan (seven-year
presidential terms and five-year assembly terms) with which the Fifth Republic began.
 Compatibility between president and assembly 269
In 1986, such a division of the executive against itself finally occurred, and
its relatively smooth functioning has been attributed to a series of contingen­
cies that add up to a best-case scenario for cohabitation in a bipolar party
system (Pierce 1990). However, the old left-right divide may be be declining
in significance, given the virtual demise of the Communist Party, Mitter­
rand's "opening to the center," and the rise of an extreme right, which may
force future conservative presidents to look leftward for coalition partners,
to avoid depending on extremists for an assembly majority. If so, shifting
coalitions may become the norm and, as such, coalitions might come to
depend on party bargaining positions more than on "automatic" majorities
for one potential government or the other, as with the two-bloc party system.
In such phases, the electoral format and pattern of government formation
would resemble what we have called "quasi-parliamentarist." If, however,
presidents retain the constitutional right of dissolution during the honey­
moon, 5 there will also continue to be potential five-year phases of quasi­
presidentialist governance. In this sense of phases alternating between
quasi-presidentialist and quasi-parliamentarist formats, France approxi­
mates a final type: the "true" hybrid.

H Y B R I D F O R MATS

Finally, there is a possible electoral format for a premier-presidential

regime that combines features of both quasi-presidentialist and quasi­
parliamentarist formats. Indeed, the French case during Mitterrand's first
term exhibited this mix of qualities. The pattern in the years of Mitter­
rand may become standard practice in France, with presidents dissolving
the assembly in order to call a honeymoon election in which they can
maximize their own support, then losing their advantage after a midterm
election. This means that cohabitation is likely to become commonplace
and, we agree with Pierce (1990), it might not always take place under
such favorable circumstances as it did in 1986-8. There are alternatives to
the particular institutional design of France that might nonetheless allow
us to maintain advantages of both quasi-presidentialist and quasi-parlia­
mentarist formats.
Duverger ( 1980) called premier-presidentialism (or, for him, "semipresi-
5 The possibility that parties in a post-bipolar party system might seek a constitutional restric­
tion on the president's prerogatives of honeymoon dissolutions can not be ruled out. In
Finland, with the expectation that presidential elections will become more competitive and
personalized for reasons that we have discussed, and with proposals for abolishing the
electoral college outright, some parties are advocating that the president's powers of dissolu­
tion also be abolished (personal communication with Sten Berglund). Such a stance by
parties would be quite consistent with the quasi-parliamentarist format, especially as party
control over presidential selection would be curtailed. Of course, unless the terms of presi­
dent and assembly are also modified so as to be synchronized, honeymoon elections would
still occur in some (but not all) presidential terms.
270 Presidents and assemblies
dentialism") a hybrid form of government. We have argued that the pattern
of president-cabinet relations associated with premier-presidential regimes
depends to a great degree on the electoral cycle. The electoral cycle
strongly influences a polity's pattern of governance, meaning the degree of
checks imposed upon the government, the likelihood of single-party versus
coalition cabinets, and the probability that the cabinet will be politically
compatible with the president. For these reasons we have argued that the
choice of electoral cycle should be made carefully.
There is one pattern we have yet to discuss: A premier-presidential
regime with an electoral cycle that places some elections in the temporal
context of the presidential elections and others at midterms would be a true
hybrid because some cabinets would be formed out of assemblies elected
under the influence of the presidential "pull," while others would not. This
pattern is indeed the best description of the current French regime. We
have suggested that there may be certain advantages to ensuring that all
cabinets result from either quasi-presidentialist or quasi-parliamentarist
electoral formats; however, this does not mean that true hybrid designs
should be ruled out. As always, it depends on the goals sought.
W hat is the primary potential advantage of such a hybrid design? Per­
haps its best feature is the opportunity to provide a "mid-course correc­
tion" on the president's cabinet. Because the support of the president's
party nearly always declines at midterm (see Chapter 11), the quasi­
parliamentarist coalition to be structured after the midterm election might
have to bring in different partners, in accordance with new voter prefer­
ences expressed in the composition of the renewed parliament. Would this
design not make periods of cohabitation, as in France in 1986-8, highly
probable? Our answer is that it depends on the precise electoral cycle used
and on the parliamentary electoral system.
Before proceeding to these considerations in the context of the hybrid
premier-presidentialism, we should point out that we do not, in the first
instance, intend to condemn cohabitation universally. We do, however,
caution that its relatively smooth experience in France might not be the
norm. If presidents have grown accustomed to having cabinets compatible
with them politically, and then suddenly must submit to a cabinet from
across a sharp ideological divide, the experience of cohabitation might be
less happy.
Assuming that in many instances it would be preferable to keep the
president's party in the cabinet after the midterm election, but have the
cabinet shift from single-party to a coalition or from one coalition to an­
other, an appropriate hybrid format can be designed. First of all, we should
rule out concurrent elections, as this cycle is likely to produce a party
system dominated by two parties. In such a party system, the midterm is
likely to see a shift of electoral support to the one principal opposition
party, ushering in cohabitation with confrontation. Second, we should re-
 Compatibility between president and assembly 271
quire that such a system use proportional representation, so that a small
shift in votes at midterm would not be exaggerated into a large swing in
seats that would depose the president's party from the cabinet. It was
exactly such a motivation that led the French Socialists to change the
electoral system from majority runoff to PR in advance of the 1986 mid­
term election. Had the majority runoff been maintained, the Socialists'
decline and the right's gains could have produced a very large majority
opposed to the president instead of the razor-thin one that resulted. For
our purposes, though, what is more interesting is what would have been the
coalition possibilities had PR been used in 1981 as well.
The French Socialist vote in 1986 was 31.2%, a not so dramatic decline
from their vote share in 1981, which was 37.5%. The majority runoff
system used in the 1981 parliamentary election translated that vote share
into 59% of the seats. Suppose a PR system had been in use in 1981. The
percent of seats would have been around 40%, making a minority cabinet
of the president's party feasible. If not a minority cabinet, Mitterrand
either would have been forced to rely openly on Communist support6 -
their support was indeed critical to his own election - or else to form a
center-left government (centrist votes, too, were necessary in his own run­
off victory).
At the midterm, the decline of the Socialists would have seemed less
dramatic. Given dissension within the right-wing bloc, particularly with
both Raymond Barre and Jacques Chirac seeking to be the next president,
a center-left cabinet might have remained a viable possibility. This is par­
ticularly so since the less majoritarian climate fostered by an institutional­
ized PR system would have been in place for the whole period. Indeed,
whether taking the form of formal participation in the cabinet or not,
centrist support would in all probability have been vital for enacting part of
the government's program after 1981. This would have meant clearer differ­
entiation of centrists from the (Gaullist) right and perhaps made viable a
continuing role for the president's party after the midterm.
The point of this counterfactual exploration of the French case has sim­
ply been to demonstrate that a honeymoon-plus-midterm electoral cycle,
using PR, could allow the president's party to remain in office, but with a
midterm correction. But wherever such a system is implemented, its design­
ers should bear in mind that cohabitation or even conflict would be a real
possibility. Parties opposed to the president could always refuse to accept
any premier and cabinet presented by the president that included the presi­
dent's own party.
The hybrid format for a premier-presidential system is probably most
viable in a three-bloc party system, in which the president might obtain a

6 Communists were brought into Mitterrand's first cabinet, but were not needed to maintain
a majority and subsequently departed.
272 Presidents and assemblies
majority government in the honeymoon election and then accede to a
coalition between his or her own party and the third-largest party after the
midterm. In a two-bloc party system, there might not be a party of the
opposition that would be willing to form a coalition with the president's
party rather than with other opposition parties. However, part of the point
of this discussion is that with PR and the electoral cycle suggested here, the
French party system might already by 1981 have appeared three-bloc in­
stead of two-bloc, and now all the more so, given the increasing mar­
ginalization of the Communist Party. 7

E LE C T O RAL FORMA T S : C O N C LUSIONS

The prospects of compatibility between president and assembly may be
shaped by the electoral cycle used. We have spoken of compatibility not
necessarily in the sense of the same party controlling the presidency and a
majority of the assembly, but in the more minimal sense of not being a
situation of either divided control or cohabitation. Divided control refers
to a congress controlled by the opposition party or bloc of parties that form
the principal opposition to the president. Cohabitation refers to a premier­
presidential system in which the premier and president come from opposite
partisan blocs. Either divided control or cohabitation raises the potential
for conflicts in the form of competing democratic legitimacies, although the
conflicts might be relatively less troublesome in a premier-presidential re­
gime, in which the assembly has ultimate authority over the cabinet and the
president lacks veto power. In a presidential regime there are no means to
reconstitute the government and the president retains his or her veto
against the new majority. Such occurrences can happen with any format,
but are far more likely when midterm elections are employed than when all
assembly elections are scheduled either as counterhoneymoon, concurrent,
or honeymoon elections. For these reasons we are deeply suspicious of the
compatibility of midterm elections with effective government, except for
the specific scenarios we have considered here.
7 To minimize potential conflicts between a president seeking reelection and a premier per­
haps also positioning himself for a run at the presidency, here might be an institutional
design in which no (immediate) reelection would be desirable. If the president and premier
were both presidential candidates, the third party that might be needed to augment the
president's position in the assembly would have an incentive to join with the president's
main opposition in order to provide differentiation for the upcoming campaign. With the
president's ineligibility for immediate reelection, an anticipatory coalition for the next
presidential election could be more easily formed, with the third party perhaps being able to
influence the incumbent president's successor.
 13
Conclusions

This b o ok has been dev oted t o the study of th ose dem ocratic constituti onal
designs in which there are tw o agents of the elect orate. The common ideal­
typical parliamentary regimes have only one such agent, the assembly,
which in Linz's expressi on, is "the only dem ocratically legitimated institu­
ti on" (Linz 1987:5). 1 Many other dem ocratic regimes pr ovide f or tw o
agents of the elect orate, one of which is the assembly, while the other is a
president with s ome p ower over the c omp osition of governments or legisla­
tion, or b oth. W herever the elect orate has tw o agents, it bec omes critical
f or the relative p owers t o be spelled out clearly in the constitution. We have
seen a great range of p owers pr ovided in actual constituti ons featuring
p opularly elected presidents.
In c onsidering different c onstituti onal designs, we have stressed that
am ong the purp oses of dem ocratic instituti ons is to pr ovide what P owell
(1989) refers t o as "citizen c ontr ol" over representatives. Citizens may c on­
trol their representatives thr ough electi ons acc ording to tw o basic m odels.
The first pr ovides that v oters sh ould have the ability to assess clear resp onsi­
bility on the part of an incumbent g overnment and either return it t o p ower
or t oss it out in favor of an alternative g overnment. Instituti ons that pr ovide
v oters this f orm of c ontrol are deemed t o pr ovide elect oral efficiency. A
second basic m odel of citizen c ontr ol assumes that v oters sh ould have a large
menu of partisan ch oices from which t o select, in order that nearly every
v oter has a party cl ose t o his or her ideal p olicy p oint. Instituti ons that
pr ovide v oters this f orm of c ontr ol are deemed representative.
The trade-off, which we intr oduced in Chapter 1, between efficiency and
representativeness is that it is n ot p ossible t o maximize b oth at once. If
1 As we have indicated, Linz's point does not apply equally to all parliamentary regimes, as
some have an upper house that has policymaking powers. Where such an upper house has
no power of confidence over the cabinet, dual democratic legitimacy may be said to exist , as
the government and upper house remain under divided control. Such cases are few, includ­
ing Canada, Germany, and Japan on certain types of legislation. In general, we do not
regard two houses of parliament as being two agents of the electorate. Throughout this
book, we have spoken of "assembly" in a way that refers to both houses whenever there is
bicameralism.
274 Presidents and assemblies
voters are offered only two choices on election day - the ins or the outs -
the choice is efficient, but many voters are bound to feel that they will go
unrepresented or inadequately represented, whoever may win. On the
other hand, if there are many viable choices on the ballot, the system will
be very representative, but assessing responsibility and directing the pre­
ferred composition of the government by means of the vote becomes diffi­
cult. Hence, designers of constitutions must choose one ideal over the
other or else balance the two in some way.
We have argued that parliamentary regimes force this choice very
starkly, simply because there is only one agent. Our assertion is backed up
by the work of Powell (1989) , who finds that parliamentary systems that
afford voters clear governmental accountability are low on representative­
ness. Examples include the United Kingdom and New Zealand. Those with
particularly high scores on representativeness seldom afford clear govern­
mental accountability. Examples are Belgium and Italy. Intriguingly, Pow­
ell also considers an intermediate model of citizen control, which he calls
"government mandates. " Such a constitutional design must combine fea­
tures of the other models that are negatively correlated: party choices (a
component of representativeness) and electoral decisiveness (a feature re­
quired to afford accountability). The implication is that such a regime
would tend to have governments that were coalitions, in order that more
than one party's policy views would be sure to be represented in the legisla­
tive process, but that there would still be a clear government-opposition
divide, so that voters could assess responsibility and identify the alternative
governmental choices. Powell (1989: 121) finds that Greece scores highest
on this model of citizen control. Some other parliamentary systems, such as
Spain, Norway, and Sweden, provide a fair balance between permitting
accountability and identifiability, on the one hand, and being fairly repre­
sentative on the other. However, such a balance may be precarious: Greece
in the late 1980s experienced two consecutive elections in which the two
largest parties failed to achieve a majority but neither had a desirable
coalition partner in the assembly.
Where there is only one agent of the electorate, as in a parliamentary
system, the electoral function that lets citizens exercise control over the
political process cannot maximize the ends of government responsibility
and representativeness at once. If there are two agents, on the other hand,
it may be feasible to maximize both through elections. In two-agent consti­
tutional designs, how identifiability and representativeness are balanced in
the final policy output depends on the ways in which presidential and
assembly powers are balanced, shared, and checked. While the cabinet
must be formed by the assembly when the latter is the only agent of the
electorate, under any format in which there is a second agent, the cabinet
may be derived in whole or in part from the agent whose election can more
readily afford accountability and identifiability through elections: the presi-
 Conclusions 275
dent. Once constitutional designers have settled on a system with two
agents of the electorate, the process of deciding on the exact shape of the
regime has only begun. It is necessary subsequently to decide how cabinet
appointments are to be made, how much legislative power the president
shall have, and what electoral rules to use, including the critical question of
the relative timing of elections of the two agents (the electoral cycle). Such
have been the principal themes of this book.
Returning to The Federalist Papers, we were able to identify one princi­
ple around which regimes featuring two agents could be designed: maximiz­
ing separation of the origin and survival of the two branches of govern­
ment. Following from this principle, we were able to construct a simple
two-dimensional typology (Figures 2.1 and 8.2) that encompasses the ma­
jor ways in which regimes with two agents differ: who has control over
cabinets, and how separate is the survival of the two agents. Presidential
regimes constitute the extreme in both dimensions, although they vary
greatly in how they allocate legislative powers (Figure 8.1). If presidential
regimes are one extreme type in the two dimensions identified, we found
that the type we called premier-presidential approximates the other ex­
treme. Parliamentary regimes are a special case in which the second agent,
the president, either has no authority over governments (or legislation), as
in Ireland, or else there is only the one agent. The exercise also allowed us
to identify a cluster of regimes that exhibit limited or no separation of
powers, yet provide the president great powers over governments and
sometimes over legislation as well. This latter type, which we dubbed
"president-parliamentary," has a dubious record of democratic stability
and is represented by such cases as the German Weimar Republic, the
Chilean "Parliamentary Republic," and the present regime of Peru.
Our remaining task is to assess the two major regime types to which we
have devoted most of our attention in this book: presidentialism and
premier-presidentialism. In doing so we discuss directions for further re­
search.

P O W E R S I N P R E S I D E N T I A L S Y S T E M S : E M P I R IC A L A N D
T H E O R E T I C A L C O N S I DE RA T I O N S

In Chapter 8 we provided estimates of the relative strengths of presidents in

several constitutions that have featured popularly elected preside·nts. In
Chapter 9 we gave a theoretical explanation for why some presidential
regimes have featured far stronger legislative powers than others. Our
work in these areas is certainly preliminary. Much remains to be done on
the relative powers of presidents and assemblies. Let us consider some
further implications of the balance of powers in (pure) presidential sys­
tems, after which we shall consider the implications of premier-presidential
systems.
276 Presidents and assemblies
Empirically, we should hope to see studies of actual power relationships,
preferably on specific policy matters, in individual countries and cross­
nationally. Our empirical knowledge of the mechanisms of congressional
oversight of presidencies and bureaucracies, of the effectiveness of presiden­
tial vetoes, and so on, has been poorly developed by political scientists. We
suggest that researchers on specific countries follow the wave of recent
studies on such matters in the United States. Examples of such studies
include McCubbins and Schwartz (1984) on congressional oversight; Rohde
and Simons (1985), Carter and Schap (1987), and Miller and Hammond
(1989) on the role of vetoes in presidential-congressional relations; and
Kiewiet and McCubbins (1991) on principal-agent relations in the budget­
ary process.
On the theoretical side, studies of power relationships are still in their
infancy even with regard to the most studied of presidential systems, the
United States (see Brams 1989, but, for reservations about the methodol­
ogy, see also Petracca 1989). Such approaches could be usefully applied to
the institutional contexts of various presidential systems, both real and
hypothetical. By expanding the scope of possible constitutional balances of
powers beyond the static example of the U.S., theoretical development
might be enhanced.
A particular area in which more work should be done is the question of
delegation versus abdication of authority. Many studies of presidential de­
mocracies, having reported minimal congressional oversight of the executive
and frequent use of decrees, have assumed that congress was thereby playing
a subservient role and ignoring what are its supposed constitutional preroga­
tives. On the United States, Cater (1964) and Ripley and Franklin (1980)
represent this view, which Shepsle (1988) has called "The Textbook Con­
gress." Works on Colombia (Archer and Chernick 1988) and Bolivia
(Gamarra 1987), and Venezuela (Coppedge 1988), among others, have
made similar observations. We should point out that a lack of explicit con­
gressional action against an executive initiative does not necessarily indicate
congressional abdication. It may even indicate nearly perfect correlation
between congressional and executive desires or acquiescence by the congres­
sional majority to a second-best outcome, or some other more subtle pro­
cess. As Kiewiet and McCubbins (1991) advise, simply reporting a lack of
explicit intervention by the congress is, by itself, nothing more than an
unexplained observation that might conceal an institutional equilibrium.
It is possible that effective oversight and direction of executive acts by
congress are rarer in many presidential democracies than in the United
States. At the same time, our investigation in this book has made us cau­
tious about such claims. W here oversight is less effective, it is often attribut­
able to these systems' institutional designs, which are typically different
from that of the U.S. Because the constitutional allocation of powers to
executive and congress are different, so would their own equilibria be
 Conclusions 277
different. In particular, a congress is constrained in its ability to provide
specific direction to policy and to oversee the policy implementation and
regulatory acts of the executive to the extent that the presidency is constitu­
tionally provided any of the following: superior budgetary authority, a
partial veto, or wide-ranging decree powers. Several Latin American presi­
dential democracies and the Philippines provide at least one of these legisla­
tive faculties to their presidents. One system, Colombia, provided all three
in the constitution that was in effect until 1991. Still, limited congressional
action contrary to executive actions does not indicate a lack of interbranch
equilibrium, as defined by the system's own constitutionally mandated
allocation of responsibilities and capabilities.
Having offered this caveat on the state of our specific knowledge about
power relationships and equilibria among branches in presidential systems
more generally, what do we know that could prove useful to those who
would "craft democracies" (Di Palma 1990), and who would opt for presi­
dentialism?

Balance of assembly-presidential powers: consequences

Our analysis in this volume has suggested that some institutional choices
are more conducive than others to democratic longevity, which is a crude,
but useful, measure of such harder-to-define concepts as legitimacy and
stability. For instance, we do not believe it is merely coincidental that the
two Latin American countries that are, as of 1991, longest-lived and gener­
ally regarded as the "most" democratic, are also two with the most highly
constrained presidential powers over legislation, both in times of "emer­
gency" and in ordinary legislation (see the discussion in Chapter 8). Put
another way, these two systems, Costa Rica and Venezuela, allow their
congresses to be quite dominant. Similarly, in Chapter 9 our reasoning
suggested that the decline and ultimate downfall of Chilean democracy
likely followed the increased concentration of legislative and particularly
fiscal powers in the presidency by more than mere coincidence.

Electoral cycles and their relation to the balance of powers

In Chapters 9 through 12 we considered the importance of electoral cycles.
We can draw useful conclusions about the interaction of presidential pow­
ers and the electoral cycle in presidential democracies by comparing two
"paradigmatic" cases: Chile and Venezuela. In Chile, the existence of hon­
eymoon elections made for a potentially destabilizing and temporary en­
hancement of the partisan strength of some presidents' parties, even
though the presidents had been elected as part of much broader coalitions.
Indeed, without the honeymoon congressional election of 1965, it is doubt­
ful that President Frei could have obtained the constitutional amendments
278 Presidents and assemblies
of 1970 that further enhanced presidential legislative powers. The large
bloc of Christian Democrats, themselves a majority in one house, com­
bined with a rightist coalition that itself expected to be able to win the
presidency to adopt a plan to increase the presidency's powers.
The Chilean tragedy suggests that the electoral cycle might affect the
salience of cross-branch checks, as does the Venezuelan case, too, but for
different reasons. In Venezuela, congress has come to be dominated by two
principal parties owing to the concurrence of elections to the two branches
(and, of course, owing to the plurality election of the president). At times
this has meant one-party government, rather than the more consensual
form of government envisioned by the regime's founders. On other, but
less frequent, occasions, the lack of an effective presidential veto in the
context of a modified two-party system has shut out the president's party
from legislative power entirely when the president's main opposition party
and smaller parties have combined on certain legislative proposals to legis­
late over the president's head. A more typical pattern of legislative
coalition-building in a country with Venezuela's combination of plurality
presidency, congress elected concurrently by rather large-magnitude PR,
and weak presidential legislative power is a party of the president holding
less than a majority but able to enlist minor parties' support for particular
legislative packages.
Both Venezuela and Chile therefore offer important lessons in the de­
sign of presidential systems. Assuming that the president is elected by
plurality or some variant thereof, such as the double complement rule (see
Chapter 10), so that either there is a two-party system or parties in a
multiparty system are encouraged to forge broad agreements before the
initial presentation of candidates to the electorate, the interaction of the
electoral cycle and veto override procedures affects the balance of congres­
sional-presidential powers.
W ith concurrent elections, the congressional election is likely to be a
contest between two parties, with third parties faring somewhat better in
congressional races than in the contest for the presidency, for reasons
shown in Chapter 11. 2 There is thus an efficient choice, as we have defined
that term. In such a context, especially if a high magnitude is used for the
congressional electoral system, the president's party will be the largest but
often short of a majority. How would the interaction of the resulting party
system and president's legislative powers work out? An extraordinary­
majority override of vetoes would allow the president to block legislation
that entailed, in the president's view, too great a compromise with the
minor parties needed to make the original legislative majority. Because the
veto override would ordinarily require the participation of the principal
2 Again, we wish to emphasize that this supposition would not apply if a majority runoff
system were used to elect the president. Then we would expect multipartism in both
presidential and congressional elections.
 Conclusions 279
opposition party, override ought to be rare. If, on the other hand, the veto
override is a simple majority, minor parties have somewhat greater bargain­
ing power because they can threaten to join forces in a legislative coalition
with the principal opposition to the president. This possibility may actually
preserve accountability between elections more than the extraordinary ma­
jority override does, as responsibility for every piece of legislation can be
made clear and public. The president and his or her party at the next
election will be seen to have been either for the bill or against it, thus
preserving a distinction between government and opposition.
The simple majority override also makes the granting of temporary legis­
lative decree powers less costly for the congress because a simple majority
can prevent an effective veto of any bill rescinding such enabling legisla­
tion. However, it is also true that an extraordinary override on vetoes of
ordinary legislation does not preclude a constitutional provision that de­
crees must be voted on by the congress, with a simple majority against
them being sufficient to prevent their taking effect. We suggest that wher­
ever a constitution is to provide for decree legislation, congressional action
ought to be mandated, with a simple majority sufficient to "veto" the act.
In the multiparty context, the presidential election (if not by majority
runoff) retains efficiency in electoral terms, but the congressional election
may not. In this context, then, the president's retention of a more effective
veto (requiring an extraordinary majority to override) can prevent legisla­
tion over the president's head from possibly becoming the normal state of
affairs. Thus the combination of nonconcurrent elections (without honey­
moon elections) and extraordinary majority override is an important in­
ducement to compromise, for two reasons. First, the electoral cycle makes
unlikely a majority or even near-majority for the president's party. Second,
the extraordinary majority override of a veto virtually prevents the exclu­
sion of the president from legislative coalitions.
In this light, Venezuela's choice in 1961 of a simple majority override is
doubly ironic. In the first place, the presidency in that country was in­
tended to be little more than an arbiter of multiparty coalitions expressed
through a powerful congress. But the simple majority override fosters a
strongly majoritarian bent in the Venezuelan legislative process. Second,
the Venezuelan electoral cycle eventually proved more conducive to a party
system dominated by two major parties rather than to multipartism. In the
end, then, Latin America's most majoritarian - and one of its most
stable - democracies developed largely unintentionally. A look at Venezu­
ela's institutional format might have led to a prediction of these develop­
ments; its founding fathers most likely would not have foreseen the effects
of their constitutional design.
Had Venezuela adopted a cycle with only midterm (or, perhaps, coun­
terhoneymoon) elections, or had presidential elections been conducted
under the majority runoff method, Venezuelan presidential elections could
280 Presidents and assemblies
have become truly inefficient in terms of determining policy. Under such
formats, multipartism would probably have been maintained and the presi­
dent often would have lacked a majority in the congress. Since the congress
is sovereign - able to legislate even over the president's objections -
presidents would have had no legislative authority. Therefore, for a
multiparty system, an extraordinary majority override is desirable as a
means to ensure meaningful presidential participation in the legislative
process. In the two-party context, the choice is less critical, but it still
affects the kinds of coalitions that are likely to form, as well as the ability of
the president to play a meaningful role in policymaking and to be held
accountable at election time for that role.

Political sources ofpresidential strength

In Chapter 9 we noted that several presidential regimes with great legisla­
tive powers also have featured highly decentralized parties with weak lead­
ership control. We suggested that this was a deliberate delegation of pow­
ers by regional politicians whose major purpose in designing such regimes
was to preserve their own autonomy to service their constituents. Features
such as partial vetoes, presidential budgetary authority and decree powers,
as well as nonsynchronized presidential and congressional terms, were insti­
tutional devices to achieve such separation not only of powers but of the
electoral processes themselves. All of these institutional features are de­
signs about which we have expressed skepticism.
Weaker presidencies tended to be associated with stronger parties as well
as concurrent elections; the latter combination, which we have dubbed
"efficient presidentialism," also has a stronger record of democratic success
among presidential systems. We noted that attempts to enhance efficiency
by strengthening presidential powers were of dubious value, as demon­
strated by the decline and ultimate overthrow of Chilean democracy in the
1960s and 1970s.
The link between the gradual strengthening of the presidency in Chile
and the deepening crisis of that regime calls attention not only to the role
that the electoral cycle might play in affecting the balance of powers but
also to the link between politicians' and parties' interests and the resulting
institutional choices. Chile's constitutional reform of 1970 resulted from an
expedient coalition of the then-governing centrist Christian Democrats and
the rightist parties that expected to win the next election. The process by
which Chile's constitution was revised suggests that Geddes's (1990) reason­
ing about the proclivity of parties or leaders who expect to be able to win
presidential elections to therefore favor presidentialism could be modified
to take account as well of parties' choosing how strong that office shall be.
As Chile's once fragmented party system consolidated, it became easier to
forge a coalition that sought to delegate even greater legislative powers in
 Conclusions 281
order to overcome inefficiency in the assembly that was a legacy of the
regime's origins.
The sequence of events in Chile thus calls our attention to the undesirabil­
ity of permitting constitutional revision to be carried out by even a two­
thirds extraordinary majority of a congress that may be temporarily under
the effective sway of one coalition. On matters as crucial to the success of
presidentialism as the relative balance of powers, amendment or revision
should have higher barriers. 3

PREMIER-PRESIDENTI ALISM

We have also devoted considerable attention to potential benefits of

premier-presidentialism, although we have endeavored to make clear that
this regime type is not a mere halfway house between parliamentarism and
presidentialism, as implied by the more common name, semipresidential­
ism, but a regime type unto itself (see Chapter 2).
We have also cautioned that the experiences of the two longest-lived
premier-presidential regimes may not be an accurate guide to how such
regimes would function more generally. In France for many years, presi­
dents regularly had working majorities in the assembly and thus compati­
ble cabinets, partly because of the electoral system used and partly because
of the political polarization engendered by de Gaulle. The experience of
cohabitation between 1986 and 1988, on the other hand, showed that when
the president could not command a majority in the assembly, the premier
would take over as the clear head of government. Yet, given the power of
initiative in the process of appointing the premier, a situation of a presi­
dent and assembly majority confronting one another from across a sharp
political divide may provide a less happy experience than that seen in
France in those years. Indeed, our theoretical model in Chapter 6 does not
predict a prime ministerial appointee at the assembly's ideal point, even
though this was the case in France. Thus we suggested in Chapter 12 ways
to minimize the prospect of cohabitation in premier-presidential regimes.
Indeed, where premier-presidentialism is perhaps most desirable is in pre­
cisely those systems in which there is no clear polarization of the party
system. In such contexts the ability of the president to use powers of
3 In this book we have deliberately not devoted attention to amending procedures because
our primary (though hardly sole) purpose has been to describe and analyze the conse­
quences of actual or hypothetical constitutions once adopted rather than to delve into the
constitutional design process itself. We might recommend, however, that mechanisms for
constitutional revision require the assent of lower elected bodies even in nonfederal sys­
tems, or passage through commissions in which smaller parties are overrepresented. In
some cases, simply raising the threshold for revisions affecting the balance of powers to
75% or so or requiring an intervening congressional election and then a second vote , might
be sufficient to avoid the Chilean problem of fundamental constitutional revision based on
short-term fluctuations in the partisan balance.
282 Presidents and assemblies
initiative to exploit the lack of a clear majority in the assembly (see the
discussion in Chapter 6) is advantageous, in our , view, because it brings
enhanced efficiency to the process of government formation, compared to
multiparty parliamentarism, while still retaining in the hands of the multi­
party assembly ultimate sovereignty over the maintenance of reconstitut­
ing of cabinets.
Finland is a case in which the president has exercised a considerable
degree of power in shaping governments. The generalizability of the Fin­
nish case is, however, hampered by the long tenure of one particular presi­
dent elected by a party-dominated electoral college, which made him in
many ways an agent to whom parties delegated considerable authority.
Kekkonen's presidency thus resembles some "presidents" chosen in parlia­
ments of pure parliamentary systems (as in Israel and Italy) more than
presidents of the other systems we have considered. We have expressed a
preference for a president elected directly, so that the president is capable
of functioning as the agent of the electorate, rather than of the parties.
Historically, premier-presidentialism and the less effective president­
parliamentarism have emerged primarily as alternatives to parliamentar­
ism. Examples include the several European countries that won their inde­
pendence from the empires that were weakened or defeated in the early
twentieth century, such as Austria, Finland, and Germany. Also in France ,
premier-presidentialism emerged as an alternative to a parliamentary re­
gime that was regarded as a failure. The major alternative to presiden­
tialism has tended to be president-parliamentarism. In Latin America, sev­
eral countries, including Ecuador, Peru, and now Colombia, have had
experience with such regimes. The experience has not been a happy one, in
most cases.
As an option for existing presidential systems, especially those with
multiparty systems, premier-presidentialism holds promise as a way to di­
minish the majoritarian tendencies of presidential executives. Because the
prime minister and cabinet are subject to parliamentary confidence, they
are more likely than in a pure presidential system to be of a different party
than the president, when that party does not hold a majority of seats. By
employing midterm elections, such a regime can also limit the extent of the
presidential "pull" on assembly elections, and thus on the cabinet. Such a
design is thus potentially more consensual, but, as indicated in Chapters 4
and 12, this device risks simply transforming presidential-congressional
conflicts into presidential-cabinet conflicts. Thus careful attention should
be paid to both the electoral cycle and to ensuring that the premier is able
to take over as the real head of government and chief policymaker when­
ever the president loses his or her majority in the assembly.
We have devoted a great deal of attention to premier-presidentialism in
this book, in part because it is relatively unfamiliar as a regime type but
also because it shows considerable promise as an effective constitutional
 Conclusions 283
design. In the late 1980s and early 1990s, the design of such regimes cer­
tainly appears to be a growth industry, as premier-presidentialism is emerg­
ing as the regime of choice in much of Central and Eastern Europe and
Africa, and has received favorable consideration in Latin America. In a
few cases, however, what has emerged is not really premier-presidentialism
but, rather, president-parliamentarism. Because of the dubious record of
the latter type, we devoted much care to distinguishing between the two
forms, mainly in Chapters 3, 4, and 6.

PRESIDENTS AND DEMOCRATIZ ATION

As we noted in Chapter 3, the preponderance of recent academic writing on

constitutional forms has stressed the superiority of parliamentarism over
presidentialism, considering only a dichotomous classification of regimes, as
unfortunately has been the case in most previous discussions. The enthusi­
asm among academics for parliamentary regimes has been most especially
extended to newly democratic polities, for which presidentialism is said to
have perils and dangers (Blonde! and Suarez 1981; Linz 1987, 1990;
Mainwaring 1989, 1992a; Di Palma 1990). Yet this academic preference has
hardly been shared by most actual constitution writers or by the authors of
this book. Is it possible that designers of constitutions for new democracies
have legitimately recognized the advantages of the presidential and premier­
presidential forms of government? Or are they simply following habit, par­
ticularly in the case of new democracies in countries that also have a prior
history of presidentialism (Latin America)? Or, are constitution writers
simply following their own self-interest, unable to resist the temptation
stemming from their expectation of being able to hold the power of the
presidency themselves (as Geddes 1990 suggests), even to the detriment of
the broader interest of democratic stability?
As we conclude this book, we should like to suggest that the very possibil­
ity that actual constitution drafters may see their best self-interest served
by presidentialism or premier-presidentialism may in itself be beneficial to
the transition to democracy under certain conditions and as long as legisla­
tive powers are not skewed in the president's favor. We agree with Di
Palma (1990) in his assessment that the fuzzy concept of "legitimacy" of a
new democracy is best judged simply by whether or not the rules of the
game achieve "behavioral compliance," but we disagree with his assess­
ment that parliamentarism is almost universally superior to presidentialism
for new democracies. There may be instances in which a compromise based
on the establishment of a popularly elected president with some circum­
scribed political powers can be crucial to attaining that behavioral compli­
ance and, therefore, to stabilizing a young democracy.
The advantages of parliamentarism for new democracies, according to its
advocates, stem from that regime type's supposed superior ability to regu-
284 Presidents and assemblies
late c onflict. Yet having a p opularly elected president can offer p otential
advantages in c onflict regulati on. These advantages stem from the l ogic of
presidential electi ons in giving v oters identifiability over the ch oice of ex­
ecutive auth ority and even fr om the big-party bias inherent in presidential
electi ons.
On identifiability, presidential elections, unlike parliamentary electi ons in
n onmaj oritarian systems, all ow elites t o present t o v oters alternative g overn­
ments and all ow v oters t o determine what are the p ossible executive options
being placed bef ore them. It might be countered that identifiability is thus a
luxury that only a stable dem ocracy can aff ord. Object ors might further
argue that in a c onflictual envir onment it is m ore imp ortant t o make elec­
ti ons a "permissive" or "feeble"4 reflection of diverse interests with execu­
tive p ower t o be c onstituted afterward. 5 H owever, presumably the desire t o
prevent fragmentation means that extreme P R w ould n ot be adopted. Then
the ir ony of ad opting parliamentarism as a c onflict-regulating pr ocedure, as
adv ocated by Di Palma and others, is that as district magnitude is reduced t o
limit fragmentati on, the number o f c ompetit ors is likewise reduced and s o is
the likelih o od of br oadly inclusive c oaliti ons. As coaliti ons bec ome less
likely, s o d oes the p ossibility of using parliamentarism's principal conflict­
regulating device, the v ote of n o-c onfidence.
In a parliamentary system with single-party g overnment or stable coali­
tions (a diverse gr oup including Britain, Spain, and Norway, f or example),
the identifiability of electi ons is alm ost as great as it w ould be in a presiden­
tial system. But s o are the fixed-term and single-party natures of the execu­
tive, f or which presidentialism is criticized. 6
N ow let us alternatively c onsider those cases in which interests that must
be pr otected in order t o achieve assent t o a c onstitution are s o numer ous
that multiparty c oaliti ons are nearly always necessary. In the multiparty
parliamentary c ontext, the resulting c oalitions typically will not be clear
4 These terms to describe high-magnitude PR with a low threshold (if any) come from
Lijphart (1984) and Sartori (1968). respectively.
5 This characterization, it might be noted, applies not only to parliamentary but also to
assembly-independent systems in which the "president" is chosen by, but not responsible to.
the assembly, as in Bolivia, El Salvador (1982-4), Lebanon, and Switzerland (see Chapter
5). These systems score at least as low on identifiability as any multiparty parliamentary
system. The latter two do so especially, owing to their (unelected) collegial executives.
Unlike multiparty parliamentarism, these systems' executive stability in the face of deep
sociopolitical divisions could be held to make up for what they lack in identifiability.
Moreover, Bolivia's constitutional design may have helped that country become one of the
first Latin American democracies to successfully implement an austerity program . and El
Salvador's can be credited with gaining the behavioral compliance of the extreme right
despite violent polarization between right and left and the latter's abstention from the
regime.
6 Of course, there remains the possibility of early elections even with a single-party parliamen­
tary cabinet, but especially if that single party holds a majority of seats, the early election
will occur only at a time calculated by the ruling party to be to its electoral advantage.
 Conclusions 285
before elections. In such a setting, the relative inability of parties to present
a meaningful picture of what they can hope to achieve - as opposed to
what merely they will ask for - is an argument against parliamentarism. If
there is a presidential election and the electoral formula is the double
complement rule, moderation and conciliation in the choice of executive
will be encouraged without compromising the representativeness of the
assembly, which can be elected by PR, even a very feeble PR. As we
discussed with regard to the Chilean system at mid-century, the improve­
ments we offered in Chapter 5 for collegial presidencies, and the electoral
formats discussed in Chapter 12, the ability of voters to identify a general
direction in which they would like to orient policy and mutual guarantees
against majoritarianism need not be incompatible goals of constitutional
design.
Second, to the extent that new democratic rules are a result of a bargain­
ing process involving established parties (such as those associated with an
outgoing authoritarian regime or restricted democracy) or where democrati­
zation involves incorporating a potentially major party that was previously
excluded, a "presidential compromise" may actually facilitate the striking
of a bargain. W hile Di Palma (1990: 217-18) clearly disparages the compro­
mise struck in Poland in 1989, the establishment of a presidency initially
reserved for the outgoing dictator but subsequently to be an elective posi­
tion could be seen as an ingenious form of conflict regulation. The Polish
bargainers set aside a presidency to provide assurances to the defeated
forces of the old regime while leaving primary legislative responsibility to a
more democratically endowed parliament and a government responsible
thereto.
The Namibian constitutional compromise of 1989 is also an example of a
presidency having been chosen as a bargain between a strong party and
several weak opponents. The dominant SWAPO received the presidency,
but the assembly, over which SWAPO is more likely to lose sway in the
short run than the presidency, has the power of no-confidence. While the
regime is fundamentally president-parliamentary, the willingness of the
president to exercise full presidential powers may be tempered by a provi­
sion that requires a new presidential election if the president chooses to
dissolve the assembly.
Venezuela's constitution of 1961 represents an example of pure presiden­
tialism as a product of accommodation among elites. The presidency had
already been won by a major party, but a multiparty pact to share power
was institutionalized through making the presidency an exceptionally weak
institution in terms of legislation. Thus any of the various regime types with
an elected president may emerge as a bargain that can protect the interests
of nonmajority parties. Having an elected president does not doom the
polity to zero-sum conflicts, competing legitimacies, false mandates, and
exaggerated majoritarian tendencies, as has been charged by various crit-
286 Presidents and assemblies
ics. Any or all of these defects may occur, but are far less likely to do so if
the regime is crafted so as to prevent dominance by the president and
diminish the possibilities of clashes of powers.
Our intention is not to dispute Di Palma's basic point that parlia­
mentarism may contain mechanisms that are conducive to conflict regula­
tion. Rather, we wish to point out that presidentialism or premier­
presidentialism properly crafted can exhibit conflict-dampening advantages
of their own that should not be overlooked. In certain contexts, as when
one party dominates such that it might be able to win control over the
executive by itself under parliamentarism, a presidential or premier­
presidential regime might even offer better opportunities for conflict regula­
tion than would a parliamentary regime. The keys to crafting a system with
a popularly elected executive properly for purposes of conflict regulation
are:
1 . A representative assembly elected by PR to ensure "fair" representation of
diversity and endowed with:
a. superior legislative powers relative to the executive, in the case of presiden­
tialism, or
b. the ability to censure and replace the cabinet, in the case of premier­
presidentialism; and
2. A presidency elected in a way to encourage a broad preelection coalition and
thus moderation and endowed with carefully circumscribed authorities, such as:
a. a veto with more than simple-majority override, but only if the president is the
constitutional head of government, or
b. conditional power to dissolve the assembly and call new elections, but only if
the regime is premier-presidential.

CONCL USI ONS

Much of our discussion in this book has been about how to elect an execu­
tive, or an agent besides parliament with power over the composition of the
executive, without sacrificing the advantages of the more consensual pattern
of democratic governance that is implied by a multiparty system. It is for this
reason that we have focused so much attention on the question of interac­
tions between elections for assembly and the process of forming an execu­
tive. In conclusion, we have argued in this book that trade-offs are involved
when one attempts to maximize the advantages of both an accountable and
electorally identifiable executive as well as an assembly representative of the
diversity of the society and polity. Systems with elected presidents, taken at
face value, would appear not only to strike a balance between the two, but
also to maximize both simultaneously. The reason is the existence of two
agents of the electorate. However, as we have seen, having a separate execu­
tive and assembly requires careful attention to how the powers of the two
overlap and check each other. The institutional choices that comprise vari­
ants on presidentialism and premier-presidentialism, as well as on the elec-
 Conclusions 287
toral rules and electoral cycle employed, may compromise considerably the
advantages of one function or the other.
We have seldom identified a particular institutional feature as universally
preferable to any alternative. W here we have a clear preference, we have
been explicit on it. The basic preference that has guided us in this book has
been that for accountability and identifiability in executive formation
(which normally means an elected presidency) and for broadly representa­
tive assemblies (which normally means PR). Another has been our prefer­
ence for a variant of plurality election of the president, most particularly
the double complement rule, rather than majority runoff. Yet another has
been our warning to stay away from president-parliamentary designs. Our
primary purpose, however, has been not to prescribe but to make clear
what the institutional choices and trade-offs are, given a set of goals that
one might want to maximize. As Lijphart noted in the conclusion of his
study of Democracies (1984), there are many ways to run a successful
democracy. We hope we have successfully highlighted many more such
ways, as well as some less successful ones.
 A P P E N DIX A

Elector al rules for one-se at districts and

co alition-building incentives

We saw in Chapter 10 that in Costa Rica, the rule requiring a candidate to

obtain at least 40% of the vote or else face a runoff has led to a pattern of
competition almost identical to that seen in straight plurality presidential
systems. The Costa Rican rule thus contrasts rather starkly with majority
provisions, in which there is a runoff in the event that no candidate obtains
at least 50% plus one of the votes. In such majority runoff systems, the vote
shares of the first and second finishers in the first round tend to be consider­
ably lower than in plurality rules, including the Costa Rican variant.
Is this outcome fortuitous, or is there something to the 40% rule that
encourages broad preelection coalitions as does pure plurality? This ques­
tion is worth investigating , especially given that there is only one case of a
system using the 40% rule. Perhaps Costa Rica, where a runoff has never
been required, is a special case because of its single well-organized mass
party, the Party of National Liberation (PLN). Perhaps this party has a
"natural" share of the vote that approximates 50% anyway. If so, it might
ordinarily win elections (which it certainly does), and it could only lose to a
united opposition - that is, one that could itself approach the magical 50%
mark. Indeed, in six of the ten elections since 1958, the PLN has obtained
an absolute majority. The lowest winning share of the vote in a Costa Rican
presidential election since the 40% provision was adopted was just over
43% in 1974 (by the official candidate of a split PLN). An opposition
candidate won in 1958 with about 45%. Is this tendency in Costa Rica to
have large blocs inherent to the 40% rule or simply inherent to the Costa
Rican political scene?

CHOOSI NG A THRESHOLD

Let us suppose that all single-member district systems that employ a cate­
gorical ballot - meaning any that require voters to vote for one candidate,
as opposed to ranking candidates by order of preference (Rae 1967) - can
be placed upon a continuum according to the level of votes that the first
finisher must surpass in order to be elected without necessitating a runoff.
Let us call this level of vote the threshold, T, expressed as a percentage.
 Electoral rules and coalition building 289
The useful range of values for T is effectively from nearly zero to 50% of all
votes plus one vote. T = 0 would be operative in the imaginary situation in
which a field of candidates all obtained no votes except for one who got a
single vote or when all candidates obtained a single vote and the winner
was chosen by lot. In either case, when applied to large national elector­
ates, the winner would have a votes share tending toward zero. In more
realistic scenarios (or tiny electorates), T = 0 simply conforms to an elec­
toral rule that is silent about the share of votes needed to win; all that is
thus required is for the winner to surpass the votes of his or her closest
rival. T = 0, then, means straight plurality.
Clearly T = 50% + 1 refers to majority systems. The first-place finisher
is not the winner of the office at stake unless he or she obtains an absolute
majority, one vote more than half of all the votes cast (or valid votes cast)
in the election. Seemingly this is as high as T would logically be set. Indeed,
we know of no electoral law used among mass electorates in which a
candidate for any office must obtain an extraordinary majority (say T =
60% or T = 66.7%). However, theoretically, such a value is possible;
indeed, some parliamentary systems mandate T > > 50% for the election
within parliament of their presidents. W ith such a small electorate (the size
of the assembly) and the practicality of holding several ballots, having
failed candidates withdraw and the opportunities for logrolling and trading
of favors that operate within a parliamentary body with much greater
facility than within mass electorates, such extraordinary majorities are feasi­
ble (if a bit unwieldy). For mass electorates, T = 50% + 1 is certainly the
highest feasible value, simply because that value of T is sure to be met in
the runoff if only two candidates are permitted to contest the second,
decisive round, as is almost always the case. 1

WHY 40 % ?
We are still left with the question of why constitutional engineers might
settle upon a 40% threshold. Besides its actual use in Costa Rica, the 40%
rule has been proposed in the United States as a replacement for the
existing electoral college. At least in the U.S. case, the motivation for the
proposal appears to be twofold. First, of course, is the desire to avoid the
possibility of the winner of the highest office having been endorsed very
narrowly. Second, there is apparent recognition of what was demonstrated
empirically in Chapter 10; when T = 50% + 1, the number of contestants
grows.
1 Alternatively, T may be set at zero for the second round, as in the French National Assem­
bly elections. (In the French case there is also a threshold on first-round votes that must be
met before a candidate can participate in the runoff in which a plurality is sufficient.
Occasionally rules have provided for T = 50% + l in each of two rounds, then dropped to
plurality for a third round in the event the outcome had yet to be decided.)
290 Appendix A
Let us consider, then, that these two concerns are in the minds of certain
constitutional engineers. We look for a value of T that encourages two
principal challengers for the presidential office. If we set T as high as an
absolute majority, the number of challengers is likely to exceed two, for
two reasons. First, it appears difficult to obtain backing of an absolute
majority; majority winners are empirically relatively rare in single-round
presidential elections, partly because there is almost always some segment
of the political leadership and electorate that prefers to run or support a
certain nonwinner, either for purposes of protest or with the long-run goal
of promoting an alternative for the future. Or, in the case of concurrent
presidential and congressional elections, a minor party may present its own
presidential candidate principally as a means to attract nationwide atten­
tion to its congressional slate, as discussed in Chapter 11.
Second, and stemming from the difficulty of obtaining an endorsement
from an absolute majority, is that a threshold of 50% + 1 increases the
expectations that the first election will not be decisive. If this is the expecta­
tion, then some leaders are bound to prefer the "blackmail" value of achiev­
ing their own voting bloc in the first round in hopes of exacting concessions
from one of the first two finishers between rounds.
If the threshold is set too low, the number of candidates who might be
seen as credible winners may at times exceed two. Ordinarily, the uncer­
tainty of multicandidate competition is expected to induce alliance build­
ing. One can block one's "worst evil" by forming alliances with lesser evils.
It is even possible that the more potential competitors there are, the
stronger the incentive is to form a really broad coalition under plurality
rule. This incentive would be further strengthened in situations in which
there is one formidable incumbent foe that the fragmented opposition
seeks to oust. Under plurality rule, such an opposition must unite if it is to
win, given plurality rule, or else the incumbent might win even with consid­
erably less than a majority. Ferdinand Marcos of the Philippines gambled
and lost that his opposition could not bury its differences to present a
unified front (under plurality rule) against him when he called his "snap"
presidential election for 1986. Panama in 1989 and Nicaragua in 1990 are
two other cases with plurality rule in which fragmented opposition forces
abandoned their differences just long enough to agree on a common-front
presidential candidate. 2
However, our discussion in Chapter 9 on the case of Chile suggests a
possible pitfall of plurality rule when there are three sharply defined blocs.
If these three blocs are approximately of equal size, any one of the three
can reasonably aspire to win. W hile one could say the same about all four
2 One prominent case stands as testimony to the irrationality of failing to coalesce in situa-
tions like those just discussed. In South Korea in 1987. President Rho Tae Woo gambled
and won that his opposition could not present a united front. Rho won the election with
36% of the vote.
 Electoral rules and coalition building 291
candidates in f our-bl oc c ompetition, three-bl oc c ompetiti on may at times
p ose special pr oblems f or plurality rule. What is special ab out three-bl oc
p olitics? If there were f our bl ocs, each having an expected v ote within a
narr ow range ar ound 25% and each having its own p otential candidate, any
two c ould c oalesce and expect a maj ority or near maj ority. These expecta­
ti ons w ould tend t o pr oduce tw o candidates under plurality rule (and f our
under maj ority runoff rule).
H owever, with three bl ocs, each expecting ar ound 33% of the v ote, c oali­
tion building bec omes m ore c omplicated if the ideol ogical distance ( or per­
haps pers onal enmity am ong the candidates) is great. Any tw o t ogether can
expect t o win 60% t o 70% of the v ote, far m ore than needed t o win. F or the
bloc that c ontributed ab out half of such an extra ordinary maj ority - but not
the candidate - t o the c oaliti on, the expected sp oils of office might be s o l ow
as t o deter the formati on of the c oaliti on in the first place. This conditi on, in
which three candidates c ould be a stable outcome despite plurality, w ould be
expected t o occur relatively rarely. When the expectati on of winning and
thus enj oying the wh ole sp oils of office is high f or all three, and being a mere
juni or partner in a c oaliti on is n ot likely t o pr oduce great rewards (especially
given the single-seat character of the presidential electi on), three-way races
may result. Examples of such three-way races are Brazil in 1955, Chile in
1958 and 1970, and the D ominican Republic in 1990.
If we desire t o select a value of T that will prevent situati ons in which any
one of three candidates might win, T must be greater than 33%. But it must
be less than 50% lest candidates' attention be f ocused on a pr obable runoff
instead of the initial electi on. If the b oundary c onditi ons are thus 33% and
50%, either an arithmetic or a ge ometric mean will pr oduce a value of
ar ound 41 % . Thus 40% or a bit higher is a likely value of Tat which only tw o
candidates will be perceived (and perceive themselves) as likely winners, but
the first r ound will be perceived as likely t o be decisive. As a r ound number,
40% has greater appeal than 41 %, and this r ounded value receives the oreti­
cal justificati on fr om the reas oning presented in this appendix.
The 40% rule d oes n ot, h owever, prevent a narrowly end orsed winner. It
merely prevents one end orsed by less than 40% from bec oming president
with out first clearing a second r ound. F or this reas on we have prop osed in
Chapter 10 a "d ouble c omplement rule." This rule w ould have one of the
desirable features of the 40% rule in that it w ould simultane ously safeguard
the elect oral pr ocess against the relatively rare case of fragmentati on in
ordinary plurality electi ons and the c omm on case of fragmentati on in the
first r ound of maj ority runoff elections. The d ouble c omplement rule has
the additi onal advantage of safeguarding the pr ocess against an inherent
p ossibility when the 40% rule is used: the chance of a very narr ow winner,
perhaps by a margin of 45% t o 44%, with s ome third-party candidate
wh ose supp orters might prefer the runner-up over the fr ont-runner.
The d ouble c omplement rule, unlike the others mentioned in this appen-
292 Appendix A
dix, expresses the threshold that must be achieved for victory not only in
terms of the front-runner's share of votes but also in terms of the gap
between that candidate's share of the votes and the vote share of the
runner-up. The simple formula for determining whether a candidate has
met the rule's requirement is given in the main text of Chapter 10.
A further appealing feature of the rule is that it has an almost self­
enforcing elegance. W hile theoretically a candidate could win with 30% or
a lower share of the vote, there is no reason to attach a minimum share
(such as 35% or 40%) in addition to the provision of the double comple­
ment rule. If the largest share won by any competitor were to be 30% and,
further, that candidate were to have met the provision of the double com­
plement rule, the runner-up could have no more than 10% of the vote. In
such an imaginary field of competitors, there would have to be, in addition
to the leader, at least seven candidates, each with no more than 10%. Such
a field is highly unlikely, but were it to occur, the candidate who had won
the 30% would almost appear to have a mandate (and would be a likely
second-round winner anyway). By requiring a minimum winning margin,
rather than a minimum winning share, the double complement rule guards
against the greatest dangers of other methods: narrowly endorsed winners.
 APPEN DIX B

Theoretic al expl an ation for models predicting

number of p arties in presidenti al systems

Our task here is to achieve models for predicting the number of parties in a
political system according to various institutional features. As mentioned in
Chapter 10, the most important of these features has for some time now
been recognized to be district magnitude, M. Additionally, according to the
reasoning regarding Duverger's rule, also discussed near the beginning of
Chapter 11, presidential systems might be expected to be different from
parliamentary systems in their number of parties, for a given magnitude.
That this should be so, at least for those presidential systems employing
concurrent elections and plurality rule for the presidency, was foreshad­
owed by Table 10.2, in which nationwide values of effective number of
parties (N) are given. Our task here is to go beyond the purely empirical
observations of Chapter 10 and seek theoretical explanation.
This discussion builds upon Taagepera and Shugart (1992a, b), in which
are developed general models of the relation between magnitude and num­
ber of parties. By general models, we mean models for the simplest of
general cases, that is, primarily parliamentary systems. When we add to the
PR assembly electoral system the complexity of a concurrently elected
presidency using plurality rule, it is necessary to make further assumptions
in the model. Some of this material has already been discussed in Chapter
11, but the discussion here is more technical.

THEORY O F T H E N U M BE R O F PARTIES

The general model has its origins in theoretical work by Taagepera (1988,
unpublished) which received empirical testing in Shugart (1988). Taage­
pera suggested the following reasoning behind the number of parties in a
district. The number of parties winning seats in any given district - the
actual, not effective, number - has a lower limit of 1 and an upper limit of
M. In the absence of further information, and knowing these absolute
limits, it is justifiable to take as an estimate the geometric average of
these extreme values. The geometric average yields p = (l x M) 5 , or the
square root of M.
294 Appendix B
As explained in Taagepera and Shugart (1992a), this is the expectation
for a simple district. W hen, as in all but a few countries (such as Israel and
the Netherlands), a country's electoral system consists of several districts, p
may be higher. A plausible reason for this is that the districts examined are
not themselves self-contained but, rather, are subnational units within
larger systems. Thus, parties that would be discouraged from competing in
a single nationwide district of, for example, M = 20, might run in several
separate districts of M = 20 within a larger country - even outside the
parties' regions of normal seat-winning strength. Sometimes these parties
win seats in such districts, producing an upward pressure on p.
For simple nationwide districts, p = M·5 provides a good fit. But for
districts within systems made up of several districts, p > M·5 , according to a
factor that corrects for assembly size (Taagepera and Shugart 1992a), pro­
vided that the system is parliamentary. The general form for all these expres­
sions is
p = M\ (B. l )
where k is an exponent that depends on specific institutional features of the
political system. If there is only one district and the system is parliamen­
tary, k = .5, such that
p = M·s. (B.2)
If there are several districts in a parliamentary system, the value of k
must take account of the size of the national assembly of which any given
district is one component:
k = l/[q + (M/S)l//nS], (B.3)
where q is a factor that reduces k to its basic value for a district when there
is no nationwide effect to consider and S is the total number of seats in the
assembly. For parliamentary systems, in order to get back to k = .5, q = l.
(The derivation of this exponent is given in the appendix of Taagepera and
Shugart 1992a). For the usual range of assembly sizes (50 < S < 500), the
value of Equation B.3 changes little with S, for M < 30. For larger M, most
cases are countries in which there is one single nationwide district, hence S
= M. Since there is little effect of S, we can simplify the previous equation
by introducing a median value of S. We then get
k = l![q + .37M2 ], (B.4)
which is the exact value when S = 148, or around the median S observed
worldwide. The above considerations for values of k assume parliamentary
systems. Our next step then is to determine what k is for presidential
systems, or, more precisely, for presidential systems using plurality for the
presidency and electing the congress concurrently.
 Predicting number ofparties 295

TESTING THE N UMBER OF PARTIES I N

PRE S I D E N T I A L S YSTEMS

If the basic model for the general (parliamentary) case has as its starting
point k = .5, what should be the starting point for the variant of presiden­
tial system considered here? For the number of seat-winning parties, per­
haps we should think of the upper limit being the number in the general
case, for which let us use the simpler expression k = .5, while the lower
limit could be k = 0, the value that corresponds to p = 1 for the single-seat
presidency being elected simultaneously.
Such reasoning would suggest an average of these values, such that k =
.25 = ¼. However, this reasoning implicitly incorporates two further as­
sumptions, which may not be valid. The first, valid in only a few cases, is
that the votes for president and congress must be the same. In a few cases
these elections are or have been so "fused." For such cases - for example,
the Dominican Republic in 1966-74 - k = ¼ might apply, provided that the
second implicit assumption is met. That assumption is that the two
branches are equal in importance.
Now what if the elections for president and
1
congress are not fused, as is
usual? Let us hypothesize that, with p = M 14 as the base point for presiden­
tial systems (with plurality presidencies and M > 1 for the concurrently
elected assembly), the value for the congress would tend upward from this
point, toward the general expectation for (parliamentary) assemblies. To
begin a process of predicting a value for votes for the presidency (again, p is
obviously 1),1 we might expect a useful prediction is one that tends down
from p = M 14 • We now have the following expressions, which are testable:
3
Pc = M 'IB, and (B.5)
I
Pp = M IB ' (B.6)
where Pc is the number of seat-winning parties in a district for congress and
pP is a fictitious "presidential seat-winning parties," which we will use to
derive an expression for Nv. It might be useful to think of Pp as the number
of parties that would win seats in the congressional district were the presi­
dential votes used to allocate assembly seats in that district (assuming, of
course, that the votes would be unchanged if this took place).
The model for pc was tested in Figure 11.1. The largest cluster of points is
indeed below what is found for p in the general case (cf. Taagepera and
Shugart 1992a). Figure 11.1 also shows an expression that curves upward
from Equation B.5. This is an adjustment that takes into account that we
are dealing (again, in all but a few cases) with several districts within a
larger nation, rather than a single (nationwide) district. The form of the
adjustment is precisely what is done for the general case in Taagepera and
Shugart. For the upward curve, we use the value of k in Equation B.4, with
296 Appendix B
q set at 1.67, the value it must be if the expression is to reduce to k = .25
when M = S, as the simpler expression assumes.
The expression that fits better is the one adjusted for the effects of nation­
wide politics on individual districts. This is the the upward bowed curve
shown in Figure 11. 1. Thus the existence of several districts matters in the
assembly of presidential systems to the same degree as in parliamentary
systems, even though the values themselves are indeed lower, owing to the
concurrence of a plurality presidential election. Another way of putting this
result is that in presidential systems with the electoral format being consid­
ered here nationwide politics matters twice: once in the form of the presiden­
tial race and again in the form of nationwide politics for the assembly.
Before returning to the number of presidential parties, let us continue
with the congress. U sing a set of equations derived in Taagepera and
Shugart (1992a,b), we can proceed to predictions of the relationship be­
tween effective parties and magnitude in the assemblies of presidential
systems. Each equation is independent of the connection between p and M.
They predict an average relation between p and the share of the largest
party, and between this latter quantity and the effective number of parties. 1
Following the steps in Taagepera and Shugart, we get the following pre­
dicted expression for Ns:
(B.7)
The form of this equation is Ns = M , and is thus lower than the
75k

simplest expression derived in Taagepera and Shugart for parliamentary

systems, which is Ns = AfYB = M 375 • The expressions both for parliamentary
and for presidential systems are empirically slightly too low for their respec­
tive data sets; this is because they do not incorporate the correction for
national politics. The correction, defining k as in Equation B.4, is shown as
the upward curve in Figure B. l, along with Equation B.7.
A final step for the congressional party system is to derive a prediction
for Nv. Again using equations derived by Taagepera and Shugart,2 we get a
prediction that is approximately
NV = l.5M 21 • (B.8)
This is the lower line shown in Figure 11.3. For the data shown, this
expression appears to be too low, although the slope is credible. Perhaps a
better fit would be Nv = l.7M 21 ; however, we shall stick with the theoreti­
cally derived expression. Most of the data points are from Venezuela,
which would be expected to have higher values of Nv. The reason for this is
a provision for additional seats for "vanity parties" (Shugart 1988), allow­
ing any party that accumulates about 0.55% of the votes nationwide to win
1 These equations, each of which is justified theoretically and tested empirically in Taagepera
and Shugart, are s 1 = p-·5, where s1 is the share of seats for the largest party, and N, = s 1 - 15.
2 Here the equations are v1 = (s1 M - 1)/(M - 1) and Nv = v,-1.5 .
 10

,-..,
�
A Costa Rica, 1978-86
� 0 Dominican Republic, 1962, 1982
·e= D Venezuela, 1973-88

=
i:i.

·a=
t,I) 5

4
.i
=
..."'..
.,!.
Q,l
3
0

J
=e=
Q,l
>
2

1l

�
�

2 3 4 5 10 20 30 40 50

District magnitude (M)

Figure B . l . District magnitude and effective number of seat-winning parties, Costa Rica, Dominican Repub­
lic, and Venezuela

Downloaded from Cambridge Books Online from within the IP domain of the University of California on Fri Jan 02 01:40:53 GMT 2015.
 298 Appendix B
40

30

�
20
a
...
CS! 0

.. a
Q,
■ 0
0 ■ 0 0
0 0 0 ■ 0 ■
■ ■
i :!l ■ ■ ■ 6. = 1 1 % M 0
=
hi 10 - 0 0
■
••
■ u

=·= =
CS!
0 ■
� Q, ■ ■ 0
'.Cl � 0 0
&Vii!CS! ■
0 ■
� ;.
0
5 0
t.i,s
a ·i 4 0

� 3

=
Cl) I ■
CS! 0 1978
■ 1988
2
�

2 4 5 10 20 30 40 50

District magnitude (M)

Figure B.2. District magnitude and percentage change in effective number of par­
ties, without vanity parties, 1978 and 1988

a seat. Because the Venezuelan party system can be broken down into
vanity parties, which compete only for nationwide seats, and all other
parties, which compete for district-level seats, it is possible to remove their
votes "experimentally" from the district-by-district totals.
Figure B.2 shows the percent difference in Nv with and without vanity
parties graphed against district magnitude. The figure shows that there is
no relation between M and the extent to which voters who prefer a vanity
party cast a vote for such a party; because the only hope for these parties
and their voters to win a seat is through the nationwide allocation, vanity
parties' votes are not influenced by within-district factors. (Votes for those
minor parties which do win district seats, such as those for Venezuela's
third-largest party, MAS, are indeed influenced by M, as shown in Shugart
1988.)
Figure B.2 shows that the average difference in Nv when vanity parties'
votes are set aside is about 11% . The percent difference between Equation
B.8, above, and what was identified as an empirical best fit (with intercept
1.7) is 11.8%. Our goal was to develop an expression for Nv that would
 Predicting number ofparties 299
predict the likely number of parties in the congressional districts of presiden­
tial systems using concurrent elections and plurality rule for the presidency.
When we remove the votes for parties in Venezuela that would certainly
not compete without the extremely permissive provision for nationwide
seats, we must declare our task quite successful indeed.

Presidential parties
If we return to Equation B .6, we find an expression for the number of
presidential parties, which we expect to be lower than the number of
congressional parties. Equation B.6 obviously predicts only one seat­
winning party, plus a fractional "party" even at very high magnitudes
(when M = 100, pP = 1 .8) . Recall that this is the actual (not effective)
number of seat-winning parties that would be predicted if the votes that
were cast for presidency were used to allocate M seats within the district. It
is apparent that following the procedures used in Taagepera and Shugart
and briefly summarized above would predict unrealistically low values for
N. A solution can be devised if we ask ourselves, what is the expected
number of vote-winning parties (pv) when the number of seat-winning par­
ties is one (as it must be for an M = l presidential election)? The answer, of
course, is two, as predicted by Duverger. Thus, for this simplest of all
districts, which is M = l nationwide, let us use Pv = 2pP . Substituting p ,. for
p in the equations of Taagepera and Shugart,3 we get the following estimate
for NP , the effective number of presidential candidates:
NP = l.68M 09 • (B .9)
As with the expression for Nv , above, a better fit would be higher. Votes
for vanity parties, unlike votes for parties that actually compete for district
seats, are not significantly different for president than for congress; thus
the same adjustment made for congressional N could be made for presiden­
tial N. If this were made, the points for Venezuela would fall in line with
the derived expression, which is expected to work for presidential systems
in which all congressional seats are allocated in districts.
Equations B . 8 and B.9 above suggest a curious aspect of concurrent
elections in countries that use PR for their assemblies, the class of democra­
cies for which they were developed. The prediction at M = l for the
effective number of parties in the congressional election is actually lower
(albeit not by much) than the prediction for the presidential election in that
district, 1 .5 to 1 .7 . Is this feasible? Perhaps it is, if we remember that the
reason that there are multiple presidential candidates even though only two
are "serious" is that there are many districts within the country with M > l .
In other words, once again we must incorporate an effect of national poli-
3 This means using v1 = Pv -.s and, as before, N,. = v 1 -u.
300 Appendix B
tics in a district. If some voters' first choice is for one of these third parties,
they may cast their protest votes for such parties' presidential candidates.
At the same time, however, they may prefer to support a candidate with a
chance of winning their own congressional district, for whatever advan­
tages might accrue to the district from having its local member of congress.
This suggests the possibility of some "reverse splitting" in the smallest
districts (when M = 2 as well as when M = 1). Because the theoretical
predictions are so close, they cannot be sorted out empirically; study would
have to rest upon access to actual ballot papers to see the incidence of such
reverse splits. 4 It is at least plausible, therefore, for the two curves to cross
each other.
4 An alternative possibility is to use an assumption for congressional elections that is analo­
gous to that used for presidential: that when p, < 2, p, = 2p. This would predict a flattening
of the curve for N, at values of M < 5, such that N, = 1.7 = NP when M = 1 . Since the issue
can not be disentangled empirically, owing to the small differences in the expressions'
predictions, we will stay with the simpler equation presented in the text. It is possible,
however, that the question of which expression is more pertinent might depend on the
extent to which voters perceive advantages to casting an effective vote for a district represen­
tative. The more important the district is (in addition to the relative importance of the
congress vis-a-vis the president) , the greater would be the incentive for reverse splitting for
those voters whose first preference, as expressed in their presidential vote, is for a minor
party.
 Bibliography

Ames, Barry. 1987. Political Survival in Latin America. Berkeley: University of

California Press.
Amy, Douglas. Forthcoming. Real Elections.
Anckar, Dag. 1984. "Presidential Elections in Finland: A Plea for Approval Vot­
ing." Electoral Studies 3 : 1 25-38.
Archer, Ronald P. 1989. "The Transition from Traditional to Broker Clientelism in
Colombia: Political Stability and Social Unrest." Paper presented at the An­
nual Meeting of the Latin American Studies Association, Miami, Florida.
December 4-6.
Archer, Ronald P. , and Marc W. Chernick. 1988. "El Presidente frente a las
Instituciones Nacionales. " In La Democracia en Blanco y Negro: Colombia en
Los Anos Ochenta. Bogota: Fondo Editorial CEREC.
Argentina. 1987. Consejo para la Consolidacion de la Democracia, Reforma
Constitucional: Segundo dictamen del Consejo para la Consolidaci6n de la
Democracia. Buenos Aires.
Arrow, Kenneth. 195 1 . Social Choice and Individual Values. New Haven, Conn. :
Yale University Press.
Arter, David. 198 1 . "Kekkonen's Finland: Enlightened Despotism or Consensual
Democracy?" Western European Politics 4:219-34.
Arter, David. 1985. "Government in Finland: A 'Semi-presidential System?" Parlia­
mentary Affairs 38:472-95 .
Asian-African Legal Consultative Committee (New Delhi), Secretariat. 1968. Con­
stitutions of Asian Countries. Bombay: N. M. Trpathi Private Ltd.
Bagehot, Walter. 1966. The English Constitution. Ithaca, N.Y. : Cornell University
Press.
Barbera, Augusto. 1990. "Un'alternativa neoparlamentare al presidenzialismo."
Democrazia e Diritto 2.
Batlle y Ordonez, Jose. 191 1 . "El P.E . Colegiado. " In El Dia. Montevideo, Uru­
guay.
Berglund, Sten, and Ingemar Worlund . 1990. "The Impact of Electoral Reform: A
Scandinavian Comparative Perspective. " Paper prepared for presentation at
the Western Political Science Association Annual Meeting, Newport Beach,
California, March 22-4.
Birnbaum, Jeffrey H . , and Allan S. Murray. 1987. Showdown at Gucci Gulch:
Lawmakers, Lobbyists, and the Unlikely Triumph of Tax Reform. New York:
Random House.
Blaustein, Albert P. , and Gisbert H. Flanz, eds. 1971-present. Constitutions of the
Countries of the World. Dobbs Ferry, N.Y. : Oceana Publications.
Blonde!, Jean. 1984. "Dual Leadership in the Contemporary World: A Step To-
302 Bibliography
wards Executive and Regime Stability?" In Comparative Government and Poli­
tics: Essays in Honour of S. E. Finer, edited by Dennis Kavanagh and Gillian
Peele. Boulder Colo . : Westview Press.
Blonde!, Jean, and Waldino C. Suarez. 1981. "Las Limitaciones Institucionales de!
Sistema Presidencialista," Criteria (Buenos Aires), February 26, pp. 57-71 .
Boyer, Yves. 1986. "French Foreign Policy: Alignment and Assertiveness." The
Washington Quarterly 9:5-14.
Brager, Stephen. 1990. "The Chilean Elections of 1989: An Account, an Analysis,
and a Forecast for the Future. " Unpublished.
Brams, Stephen J. 1989. "Are the Two Houses of Congress Coequal?" In The
Federalist Papers and the New Institutionalism, edited by Bernard Grofman and
Donald Wittman. New York: Agathon Press.
Bruneau, Thomas C. , and Alex Macleod. 1986. Politics in Contemporary Portugal.
Boulder, Colo. : Lynne Rienner.
Cain, Bruce, John Ferejohn, and Morris Fiorina. 1987. The Personal Vote. Cam­
bridge , Mass. : Harvard University Press.
Cain, Bruce E . , and W. T. Jones. 1989. "Madison's Theory of Representation. " In
The Federalist Papers and the New Institutionalism, edited by Bernard Grofman
and Donald Wittman. New York: Agathon Press.
Cameron, Maxwell. 199 1 . " 'Rational Resignations. ' Coalition Building in Peru
and the Philippines. " Comparative Political Studies. Forthcoming.
Campbell, Angus, Phillip E. Converse, Warren E. Miller, and Donald E. Stokes.
1964. The American Voter. New York: Wiley.
Carey, John M. 1992. "Premier-Presidentialism: The Third Alternative?" In Presi­
dentialism and Democracy, edited by Matthew S. Shugart and Scott Mainwar­
ing. Forthcoming.
Carter, J. R . , and D. Schap. 1987. "Executive Veto, Legislative Override, and
Structure Induced Equilibrium. " Public Choice 52:227-44.
Cater, Douglass. 1964. Power in Washington. New York: Random House .
Chappel, Henry W. , and William R. Keech. 1989. "Electoral Institutions in the
Federalist Papers: A Contemporary Perspective." In The Federalist Papers and
the New Institutionalism, edited by Bernard Grofman and Donald Wittman.
New York: Agathon Press.
Conaghan , Catherine. 1989. "Loose Parties, Floating Politicians, and Institutional
Stress: Presidentialism in Ecuador, 1979-88. " Paper presented to the Research
Symposium, "Presidential or Parliamentary Democracy: Does It Make a Dif­
ference?," Georgetown University, Washington, D.C. , May 14-16.
Conaghan , Catherine M., James M. Malloy, and Luis A. Abugattas. 1990. "Busi­
ness and the Boys: The Origins of Neo-Liberalism in the Central Andes. " Latin
American Research Review 25:3-30.
Coppedge, Michael . 1988. Strong Parties and Lame Ducks: A Study of the Quality
and Stability of Venezuelan Democracy. Ph.D. dissertation, Yale University.
Coppedge, Michael. 1992. "Venezuela: Democratic Despite Presidentialism. " In
Presidential or Parliamentary Democracy: Does it Make a Difference?, edited
by Juan J. Linz and Arturo Valenzuela. Baltimore: Johns Hopkins University
Press.
Cox, Gary. 1987. The Efficient Secret. New York: Cambridge University Press.
Cox, Gary, and Mathew McCubbins. 1992. Parties and Committees in the House of
Representatives. Berkeley: University of California Press.
Craig, Ann L . , and Wayne Cornelius. 1991 . "Houses Divided: Parties and Political
Reform in Mexico. " In Building Democratic Institutions: Parties and Party
 Bibliography 303
Systems in Latin America, edited by Scott Mainwaring and Timothy R. Scully.
Forthcoming.
Cuba. 1940. Constituci6n de la Republica de Cuba. Havana, Cuba: Compaftfa
Editoria de Libros y Folletos O'Reilly.
de Gaulle, Charles. 1969. "The Bayeux Manifesto. " In Politics in Europe, edited by
Arend Lijphart. Englewood Cliffs, N.J. : Prentice Hall.
Diamond, Larry, Juan J. Linz, and Seymour M. Lipset. 1989. Democracy in Devel­
oping Countries. 4 vols. Boulder, Colo. : Lynne Rienner.
DiBacco, Thomas V. 1977. Presidential Power in Latin American Politics. New
York: Praeger.
Di Palma, Giuseppe. 1990. To Craft Democracies. Berkeley: University of Califor­
nia Press.
Dix, Robert H. 1967. Colombia: The Political Dimensions of Change. New Haven,
Conn. : Yale University Press.
Dix, Robert H. 1980. "Consociational Democracy: The Case of Colombia." Com­
parative Politics 12:303-21 .
Drake, Paul. 1978. Socialism and Populism in Chile. Urbana: University of Illinois
Press.
Duverger, Maurice . 195 1 . Les Partis Politiques. Paris: Le Seuil.
Duverger, Maurice . 1954. Political Parties: Their Organization and Activity in the
Modern State, translated by Barbara and Robert North. New York: Wiley.
Duverger, Maurice. 1980. "A New Political System Model: Semi-presidential Gov­
ernment. " European Journal of Political Research 8 : 165-87.
Duverger, Maurice . 1984. "Which Is the Best Electoral System?" In Choosing an
Electoral System, edited by Arend Lijphart and Bernard Grofman. New York:
Praeger.
Duverger, Maurice . 1986. "Duverger's Law Thirty Years Later. " In Electoral Laws
and Their Political Consequences, edited by Bernard Grofman and Arend
Lijphart. New York: Agathon Press.
Epstein, Leon. 1967. Political Parties in Western Democracies. New Brunswick,
N.J. : Transaction Books.
Erikson, Robert. 1988. "The Puzzle of Midterm Loss. " Journal of Politics 50: 1011-
29.
Ertekun, M. Necati Munir. 1984. The Cyprus Dispute and the Birth of the Turkish
Republic of Northern Cyprus. Nicosia: K. Rustem.
Eulau, Heinz, and Kenneth Prewitt. 1973. Labyrinths of Democracy. Indianapolis:
Bobbs-Merrill.
Evans de la Cuadra, Enrique. 1970. Relacion de la Constituci6n Politica de la
Republica de Chile. Santiago: Editorial Jurfdica de Chile.
Federalist Papers, The. 1961. (Alexander Hamilton, James Madison, and John Jay) ,
edited by Clinton Rossiter. New York: New American Library.
Fiorina, Morris P. 1981. Retrospective Voting in American National Elections. New
Haven , Conn.: Yale University Press.
Fiorina, Morris P. 1982. "Group Concentration and the Delegation of Legislative
Authority. " California Institute of Technology, Social Science Working Paper
No. 1 12.
Ford, Henry James. 1898. The Rise and Growth of American Politics: A Sketch of
Constitutional Development. New York: Macmillan.
Gamarra, Eduardo A. 1987. Political Stability, Democratization, and the Bolivian
National Congress. Ph. D . dissertation, University of Pittsburgh.
Gamarra, Eduardo A. 1990. "Bolivia's Perestroika?" Hemisphere 2: 16-19.
304 Bibliography
Gamarra, Eduardo A. 1992. "Hybrid-Presidentialism and Democratization: The
Case of Bolivia." In Presidentialism and Democracy, edited by Scott Mainwar­
ing and Matthew S. Shugart. Forthcoming.
Garcia Laguardia, Jorge Mario, and Edmundo Vasquez Martinez. 1984. Constitu­
ci6n y orden democratico. Guatemala City: Editorial Universitaria de Guate­
mala.
Geddes, Barbara. 1990. "Democratic Institutions as a Bargain Among Self­
Interested Politicians. " Paper prepared for presentation at the Annual Meeting
of the American Political Science Association, San Francisco.
Geddes, Barbara. 1991. "A Game Theoretic Model of Reform in Latin American
Democracies. " American Political Science Review 85:371-92.
Gibson, William M. 1948. The Constitutions of Colombia. Durham, N.C . : Duke
University Press.
Gillespie, Charles G. 1989. "Presidential-Parliamentary Relations and the Prob­
lems of Democratic Stability: The Case of Uruguay." Unpublished.
Gonzalez, Luis E. 1989. "Presidentialism, Party Systems and Democratic Stability:
The Uruguayan Case. " Unpublished.
Grofman, Bernard N. 1987. "Will the Real New Institutionalism Stand Up and
Take a Bow." Unpublished.
Grofman, Bernard N. 1989. "The Federalist Papers and the New Institutionalism:
An Overview." In The Federalist Papers and the New Institutionalism, edited by
Bernard Grofman and Donald Wittman. New York: Agathon Press.
Gunther, Richard. 1989. "Electoral Laws, Party Systems, and Elites: The Case of
Spain." American Political Science Review 83:835-58.
Halperin, S. William. 1963. Germany Tried Democracy: A Political History of the
Reich from 1918 to 1933. Hamden, Conn. : Archon Books.
Hartlyn, Jonathan. 1987. The Politics of Coalition Rule in Colombia. New York:
Cambridge University Press.
Hartlyn, Jonathan. 1989. "Presidentialism and Colombian Politics." Paper pre­
sented to the Research Symposium, "Presidential or Parliamentary Democ­
racy: Does It Make a Difference?," Georgetown University, Washington,
D.C. , May 14-16.
Heise Gonzalez, Julio. 1974. Historia de Chile: El periodo parlamentario, 1861-
1925. Santiago: Editoriales Andres Bello.
Herman, Edward S . , and Frank Brodhead. 1984. Demonstration Elections. Boston:
South End Press.
Herman, Valentine, and John Pope. 1973. "Minority Governments in Western De­
mocracies." British Journal of Political Science 3: 191-212.
Hess, Stephen. 1988. Organizing the Presidency. Washington D.C. : Brookings
Institution.
Hogan, James. 1945. Elections and Representation. Cork, Ireland: Cork University
Press.
Horowitz, Donald. 1990. "Comparing Democratic Systems." Journal of Democracy
1 :73-9.
Horowitz, Donald. 1991. A Democratic South Africa? Constitutional Engineering in
a Divided Society. Berkeley: University of California Press.
Hucko, Elmar, ed. 1987. The Democratic Tradition: Four German Constitutions.
New York: Berg/St. Martin's Press.
Huntington, Samuel. 1991. "Democracy's Third Wave." Journal of Democracy
2: 12-34.
Jacobson, Gary C. 1990. The Electoral Origins of Divided Government: Competi­
tion in U. S. House Elections, 1946-1988. Boulder, Colo . : Westview Press.
 Bibliography 305
Kaminsky, Elijah Ben-Zion. 1989. "The Executive: Stable Governments, Change­
able Institutions." Paper delivered at the APSA Annual Meeting, September.
Karl, Terry L. 1986. "Petroleum and Political Pacts: The Transition to Democracy
in Venezuela. " In Transitions from Authoritarian Rule: Latin America, edited
by Guillermo O'Donnell, Philippe C. Schmitter, and Laurence Whitehead.
Baltimore: Johns Hopkins University Press.
Kelley, R. Lynn. 1977. "Venezuelan Constitutional Forms and Realities." In Vene­
zuela: The Democratic Experience, edited by John D. Martz and David J.
Myers. New York: Praeger.
Kelsen, Hans. 1945. General Theory of Law and State, Part 2. Cambridge: Harvard
University Press.
Kernell, Samuel. 1977. "Presidential Popularity and Negative Voting: An Alterna­
tive Explanation of the Midterm Decline of the Presidential Party. " American
Political Science Review 7 1 :44-66.
Kernell, Samuel. 1986. Going Public: New Strategies of Presidential Leadership.
Washington D . C . : Congressional Quarterly Press.
Ketchman, Ralph, ed. 1986. The Anti-Federalist Papers: and, The Constitutional
Convention Debates. New York: New American Library.
Kiewiet, Roderick, and Mathew McCubbins. 1985. "Appropriations Decisions as
a Bilateral Bargaining Game Between President and Congress. " Legislative
Studies Quarterly 9:181-202.
Kiewiet, Roderick, and Mathew McCubbins. 1991. The Logic of Delegation: Con­
gressional Parties and the Appropriations Process. Chicago: University of Chi­
cago Press.
Kirchheimer, Otto. 1969. "Weimar - And What Then? An Analysis of a Constitu­
tion" (1930), and "Constitutional Reaction in 1932" ( 1932). In Otto Kirch­
heimer, Politics, Law, and Social Change, edited by Frederic S. Burin and
Kurt L. Shell. New York: Columbia University Press.
Kyriakides, Stanley. 1968. Cyprus: Constitutionalism and Crisis Government. Phila­
delphia: University of Pennsylvania Press.
Laakso, Markku, and Rein Taagepera. 1979. "Effective Number of Parties: A
Measure with Application to West Europe. " Comparative Political Studies
12:3-27.
Lepsius, M. Ranier. 1978. "From Fragmented Party Democracy to Government by
Emergency Decree and National Socialist Takeover: Germany. " In The Break­
down of Democratic Regimes, edited by Juan J. Linz and Alfred Stepan. Balti­
more: Johns Hopkins University Press.
Levine , Daniel H. 1973. Conflict and Political Change in Venezuela. Princeton N.J. :
Princeton University Press.
Libai, David, Uriel Lynn , Amnon Rubinstein , and Yoash Tsiddon. 199 1 . Changing
The System of Government in Israel - Proposed Basic Law: The Government.
Jerusalem: The Jerusalem Center for Public Affairs and The Public Committee
for a Constitution for Israel.
Lijphart, Arend. 1977. Democracy in Plural Societies: A Comparative Exploration.
New Haven, Conn. : Yale University Press.
Lijphart, Arend. 1984. Democracies: Patterns of Majoritarian and Consensus
Government in Twenty-one Countries. New Haven, Conn. : Yale University
Press.
Lijphart, Arend. 1989. "Presidentialism and Majoritarian Democracy: Theoretical
Observations. " Paper presented to the Research Symposium, "Presidential or
Parliamentary Democracy: Does It Make a Difference?," Georgetown Univer­
sity, Washington, D .C . , May 1 4-16.
306 Bibliography
Lijphart, Arend. 1990a. "Six Lessons For Latin America. " Government and Opposi­
tion 25:68-84.
Lijphart, Arend. 1990b. "The Political Consequences of Electoral Laws, 1945-85 . "
American Political Science Review 84:481-96.
Lijphart, Arend. 1991. "The Alternative Vote: A Realistic Alternative for South
Africa?" Politikon: The South African Journal ofPolitical Science. Forthcoming.
Lijphart, Arend, and Ronald Rogowski. 1 991 . "Separation of Powers and the
Management of Political Cleavages. " In Political Institutions and their Conse­
quences, edited by Kent Weaver and Bert Rockman. Washington , D.C.: Brook­
ings Institution.
Linz, Juan J. 1978. "Crisis, Breakdown and Reequilibration. " In The Breakdown of
Democratic Regimes, edited by Juan J. Linz and Alfred Stepan. Baltimore:
Johns Hopkins University Press.
Linz, Juan J. 1987. "Democracy, Presidential or Parliamentary: Does It Make a
Difference?" Unpublished.
Linz, Juan J. 1990. "The Perils of Presidentialism. " Journal of Democracy 1 :51-69.
Linz, Juan J. and Alfred Stepan. 1978. The Breakdown of Democratic Regimes.
Baltimore : Johns Hopkins University Press.
Linz, Juan J. , and Arturo Valenzuela, eds. 1992. Presidential or Parliamentary
Democracy: Does It Make a Difference? Baltimore: Johns Hopkins University
Press.
Lipset, Seymour M. 1990. "The Centrality of Political Culture. " Journal of Democ­
racy 1 : 80-3.
Loosemore , John, and Victor J. Hanby. 197 1 . "The Theoretical Limits of Maximum
Distortion." British Journal of Political Science 1 :467-77.
Lowi, Theodore J. 1979. The End of Liberalism: The Second Republic of the United
States. New York: Norton.
Mackie , Thomas T. , and Richard Rose. 199 1 . The International Almanac of Elec­
toral History, 3rd ed. Washington, D.C. : Congressional Quarterly.
Mainwaring, Scott . 1989. "Presidentialism in Latin America: A Review Essay. "
Latin American Reseach Review 25 : 157-79.
Mainwaring, Scott. 1 99 1 . "Politicians, Parties, and Electoral Systems: Brazil in
Comparative Perspective . " Comparative Politics, [October] :21-43.
Mainwaring, Scott. 1992a. "Presidentialism, Multiparty Systems, and Democracy:
The Difficult Equation. " Comparative Political Studies. Forthcoming.
Mainwaring, Scott . 1992b. "Institutional Dilemmas of Multiparty Presidential De­
mocracy: The Case of Brazil . " In Presidentialism and Democracy, edited by
Scott Mainwaring and Matthew S. Shugart. Forthcoming.
Malloy, James M . , and Eduardo A. Gamarra. 1987. "The Transition to Democracy
in Boliva. In Authoritarians and Democrats: Regime Transition in Latin Amer­
ica, edited by James M. Malloy and Mitchell A. Seligson. Pittsburgh: Univer­
sity of Pittsburgh Press.
March, James G . , and Johan P. Olsen. 1984. "New Institutionalism: Organizational
Factors in Political Life . " American Political Science Review 78:734-49.
Marotta, Daniel. 1985. El Uruguay y el Segundo Colegiado. Montevideo: IDEAS.
Martz, John D. 1966. Acci6n Democratica: Evolution of a Modern Political Party.
Princeton, N.J. : Princeton University Press.
McCoy, Terry L. 197 1 . "Congress, the President, and Political Instability in Peru. "
I n Latin American Legislatures: Their Role and Influence, edited by Weston H .
Agor. New York: Praeger.
McCubbins, Mathew, and Thomas Schwartz. 1984. "Congressional Oversight Over-
 Bibliography 307
looked: Police Patrols Versus Fire Alarms. " American Journal of Political Sci­
ence 28: 165-79.
McCubbins, Mathew, and Frances Rosenbluth. 1992. "Electoral Structure and the
Organization of Policymaking in Japan. " Unpublished.
McDonald, Ronald H . , and J. Mark Ruhl. 1989. Party Politics and Elections in
Latin America. Boulder, Colo. : Westview Press.
McKelvey, Richard D . 1976. "lntransitivities in Multidimensional Voting Models:
Some Implications for Agenda Control." Journal of Economic Theory 12: 131 ,
472-82.
Miller, Gary J. , and Thomas H. Hammond. 1989. "Stability and Efficiency in a
Separation-of-Powers Constitutional System." In The Federalist Papers and the
New Institutionalism, edited by Bernard Grofman and Donald Wittman. New
York: Agathon Press.
Moe, Terry M. 1987. "An Assessment of the Positive Theory of 'Congressional
Dominance . ' " Legislative Studies Quarterly 12:475-520.
Moe, Terry M. 1989. "The Politics of Bureaucratic Structure ." In Can the Govern­
ment Govern?, edited by John E. Chubb and Paul Peterson. Washington,
D.C. : Brookings Institution.
Montesquieu, Charles de Secondat, baron de. 1989. The Spirit of the Laws. New
York: Cambridge University Press.
Needler, Martin. 1959. "The Theory of the Weimar Presidency. " The Review of
Politics 21:692-8.
Needler, Martin. 1965. "Cabinet Responsibility in a Presidential System: The Case
of Peru . " Parliamentary Affairs 18: 156-61.
Neumann, Franz. 1942. Behemoth: The Structure and Practice of National Social­
ism. New York: Oxford University Press.
Neustadt , Richard E. 1960. Presidential Power, the Politics of Leadership. New
York: Wiley.
Nogueira Alcala, Humberto. 1986a. El regimen semipresidencial: Una nueva forma
de gobierno democratico? Santiago: Editorial Andante.
Nogueira Alcala, Humberto. 1986b. Teoria y practica democratica. Santiago: Edito­
rial Andante.
Nousiainen, Jaakko. 1988. "Bureaucratic Tradition, Semi-presidential Rule and
Parliamentary Government: The Case of Finland." European Journal of Politi­
cal Research 16:221-49.
Opello, Walter C. 1985. "Portugal's Political Development. Boulder, Colo . : West­
view Press.
Opello, Walter C. 1986. "Portugal's Parliament: "An Organizational Analysis of
Legislative Performance . " Legislative Studies Quarterly 1 1 :291-319.
Ordeshook , Peter. 1986. Game Theory and Political Theory. New York: Cambridge
University Press.
Pak, Chi-Young. 1980. Political Opposition in Korea, 1945-1960. Seoul, Republic
of Korea: Seoul University Press.
Pennock, Roland. 1979. Democratic Political Theory. Princeton, N.J. : Princeton
University Press.
Perera, N. M. 1979. Critical Analysis of the New Constitution of the Sri Lanka
Government. Colombo: V. S. Raja.
Petracca, Marc P. 1989. "The Distribution of Power in the Federal Government:
Perspectives from The Federalist Papers - A Critique. " In The Federalist Papers
and the New Institutionalism, edited by Bernard Grofman and Donald Wittman.
New York: Agathon Press.
308 Bibliography
Peirce, Neal R . , and Lawrence D. Longley. 1981. The People's President: The
Electoral College in American History and the Direct Vote Alternative. New
Haven, Conn. : Yale University Press.
Pierce, Roy. 1990. "The Executive Divided Against Itself: France 1986-1988."
Paper presented at the Annual Meeting of the American Political Science
Association, San Francisco.
Pinoargote, Alfredo. 1982. La republica de papel. Guayaquil, Ecuador: Artes
Graficas Senefelder.
Pitkin, Hannah. 1967. The Concept of Representation. Berkeley: University of Cali­
fornia Press.
Polsby, Nelson. 1983. Consequences of Party Reform. New York: Oxford University
Press.
Polyviou, Polyvios G. 1980. Cyprus: Conflict and Negotiation, 1960-1980. London:
Duckworth.
Powell, G. Bingham, Jr. 1982. Contemporary Democracies: Participation, Stability,
and Violence. Cambridge, Mass. : Harvard University Press.
Powell, G. Bingham, Jr. 1989. "Constitutional Design and Citizen Electoral Con­
trol." Journal of Theoretical Politics l: 107-30.
Prothro, James W. , and Patricio E. Chapparo. 1974. "Public Opinion and the
Movement of the Chilean Government to the Left, 1952-72." Journal of Poli­
tics 36:2-34.
Rae, Douglas W. 1967. The Political Consequences of Electoral Laws. New Haven,
Conn. : Yale University Press.
Ranney, Austin. 1962. The Doctrine of Responsible Party Government. Urbana:
University of Illinois Press.
Rasmussen, Eric. 1989. Games and Information: An Introduction to Game Theory.
New York: Blackwell.
Riker, William H. 1967. The Theory of Political Coalitions. New Haven, Conn. :
Yale University Press.
Riker, William H. 1982a. Liberalism Against Populism: A Confrontation Between
the Theory of Democracy and the Theory of Social Choice. San Francisco: W.
H. Freeman.
Riker, William H. 1982b. "The Two-Party System and Duverger's Law: An Essay
on the History of Political Science." American Political Science Review 76:753-
66.
Ripley, Randall, and Grace Franklin. 1980. Congress, the Bureaucracy, and Public
Policy. Homewood, Ill . : Dorsey Press.
Robinson, Donald L . , ed. 1985. Reforming American Government. Boulder, Colo.:
Westview Press.
Rogowski, Ronald. 1990. "Comment on 'Separation of Powers and the Manage­
ment of Political Cleavages' by Arend Lijphart" (draft). February.
Rohde, David, and Dennis Simons. 1985. "Presidential Vetoes and Congressional
Responsibility. " American Journal of Political Science 29:397-427.
Rose, Richard. 1988. "Presidents and Prime Ministers." Society 25:61-7.
Rustow, Dankwart. 1970. "Transitions to Democracy." Comparative Politics 2:337-
63.
Sartori, Giovanni. 1968. "Political Development and Political Engineering." In vol.
17 of Public Policy, edited by J. D. Montgomery and A. 0. Hirschman. Cam­
bridge: Cambridge University Press.
Sartori, Giovanni. 1976. Parties and Political Systems: A Framework for Analysis.
New York: Cambridge University Press.
Sartori, Giovanni. 1992. "Neither Presidentialism nor Parliamentarism." In Presi-
 Bibliography 309
dential or Parliamentary Democracy: Does It Make a Difference?, edited by
Juan J. Linz and Arturo Valenzuela. Baltimore: Johns Hopkins University
Press.
Schap, David. 1986. "Executive Veto and Informational Strategy: A Structure­
induced Equilibrium Analysis." American Journal of Political Science 30:755-
70.
Schelling, Thomas C. 1960. The Strategy of Conflict. Cambridge, Mass. : Harvard
University Press.
Schlesinger, Joseph A . , and Mildred Schlesinger. 1991. "The Reaffirmation of a
Multiparty System in France." American Political Science Review 84: 1077-
1 101 .
Scott, Robert E . 1973. "National Integration Problems and Military Regimes in
Latin America. " Latin American Modernization Problems, edited by Robert
E. Scott. Chicago: University of Illinois Press.
Shepsle, Kenneth. 1988. "The Changing Textbook Congress: Equilibrium Congres­
sional Institutions and Behavior. " Center for American Political Studies Occa­
sional Paper 88-31 , Harvard University.
Shugart, Matthew. 1985 . "The Two Effects of District Magnitude: Venezuela as a
Crucial Experiment." European Journal of Political Research 13:353-64.
Shugart, Matthew S. 1988. Duverger's Rule, District Magnitude, and Presiden­
tialism. Ph. D . dissertation, University of California, Irvine.
Shugart, Matthew S. 1989. "Electoral Cycles and the Party System in Presidential
Democracies." Paper presented to the Research Symposium, "Presidential or
Parliamentary Democracy: Does It Make a Difference?," Georgetown Univer­
sity, Washington, D . C . , May 14-16.
Shugart, Matthew S. 1992a. "Leaders, Rank and File, and Constituents: Electoral
Reform in Colombia and Venezuela." Electoral Studies. Forthcoming.
Shugart, Matthew S. 1992b. "Guerrillas and Elections: An Institutionalist Perspec­
tive on the Costs of Conflict and Competition . " International Studies Quarterly
36: 12 1-52.
Shugart, Matthew S . , and Raul A. Fernandez. 199 1 . "Constitutional Reform and
Economic Restructuring in Colombia." Paper presented at the Annual Meet­
ing of the North American Association of Colombianists, Ibague , Colombia,
August.
Shugart, Matthew S . , and Rein Taagepera. 1992. "Majority Versus Plurality Election
of Presidents: A Proposal for a 'Double Complement Rule. ' " Unpublished .
Sigler, Jay A . , Albert Blaustein, and Gisbert H. Flanz. 1981 . "Chile." In vol. 5 of
Constitutions of the Countries of the World, edited by Albert Blaustein and
Gisbert Flanz. Dobbs Ferry, N.Y. : Oceana Publications.
Stauffer, Robert B . 1970. "Congress in the Philippine Political System." Legisla­
tures in Developmental Perspective, edited by Allan Kornberg and Lloyd D .
Musolf. Durham, N.C . : Duke University Press.
Stauffer, Robert B . 1975. The Philippine Congress: Causes of Structural Change.
Beverley Hills, Calif.: Sage.
Stedman, Murray S . , Jr. , ed. 1968. Modernizing American Government: The De­
mands of Social Change. Englewood Cliffs, N.J. : Prentice-Hall.
Stockman, David A. 1986. The Triumph of Politics: How the Reagan Revolution
Failed. New York: Harper & Row.
Stokes, William S. 1949. "The Cuban Parliamentary System in Action, 1940-1947. "
Journal of Politics 1 1 :335-64.
Strom , Kaare. 1986. "Deferred Gratification and Minority Governments in Scandi­
navia." Legislative Studies Quarterly 1 1 :583-605.
310 Bibliography
Strom, Kaare. 1990. Minority Government and Majority Rule. New York: Cam­
bridge University Press.
Suarez, Waldino Cleto. 1982. "El poder ejecutivo en America Latina: su capacidad
operativa bajo regimenes presidencialistas de gobierno. " Revista de £studios
Polfticos, 29: 109-44.
Suleiman, Ezra N. 198 1 . "Presidential Government in France. " In Presidents and
Prime Ministers, edited by Richard Rose and Ezra Suleiman. Washington
D.C. : American Enterprise Institute.
Suleiman, Ezra N. 1986. "Toward the Disciplining of Parties and Legislators. " In
Parliaments and Parliamentarians in Democratic Politics, edited by Ezra
Suleiman. New York: Holmes & Meier.
Suleiman, Ezra N. 1989. "French Presidentialism and Political Stability. " Paper
presented to the Research Symposium , "Presidential or Parliamentary Democ­
racy: Does It Make a Difference?," Georgetown University, Washington,
D.C . , May 14-16.
Sundquist, James L. 1981. The Decline and Resurgence of Congress. Washington
D.C. : Brookings Institution.
Sundquist, James L. 1986. Constitutional Reform and Effective Government. Wash­
ington D . C . : Brookings Institution.
Taagepera, Rein. 1989. "A Theory of the Number of Parties." Unpublished.
Taagepera, Rein, and Bernard Grofman. 1985. "Rethinking Duverger's Law: Pre­
dicting the Effective Number of Parties in Plurality and PR Systems - Parties
Minus Issues Equals One . " European Journal of Political Research 13:341-52.
Taagepera, Rein, and Matthew S. Shugart. 1989. Seats and Votes: The Effects and
Determinants of Electoral Systems. New Haven, Conn . : Yale University Press.
Taagepera, Rein, and Matthew S. Shugart. 1992a. "Theory of the Number of
Parliamentary Parties: A Quantitative Model of Duverger's Mechanical Ef­
fect. " Unpublished.
Taagepera, Rein, and Matthew S. Shugart. 1992b. "Theory of the Number of Elec­
toral Parties and Deviation from Proportional Representation. " Unpublished.
Taylor, Philip. 1960. Government and Politics in Uruguay. New Orleans: Tulane
University Press.
Therborn , Goran. 1979. "The Travail of Latin American Democracy." New Left
Review 1 13 :71-109.
Urzua Valenzuela, German. 1986. Hist6ria Polftica Electoral de Chile, 1931-1973.
Santiago: Universidad de Chile.
Valenzuela, Arturo. 1978. The Breakdown of Democratic Regimes: Chile. Vol. 4 of
The Breakdown of Democratic Regimes, edited by Juan J. Linz and Alfred
Stepan. Baltimore: Johns Hopkins University Press.
Valenzuela, Arturo. 1989a. "Party Politics and the Failure of Presidentialism in
Chile: A Proposal for a Parliamentary Form of Government." Paper presented
to the Research Symposium, "Presidential or Parliamentary Democracy: Does
It Make a Difference?," Georgetown University, Washington, D. C. , May 14-16.
Valenzuela, Arturo. 1989b. "Chile: Origins, Consolidation, and Breakdown of a
Democratic Regime. " In Democracy in Developing Countries, Vol 4: Latin
America, edited by Larry Diamond, Juan J. Linz, and Seymour Martin Upset.
Boulder, Colo. : Lynne Rienner.
Valenzuela, Arturo, and Alexander Wilde. 1979. "Presidential Politics and the
Decline of the Chilean Congress. " In Legislatures in Development: Dynamics
of Change in New and Old States, edited by Joel Smith and Lloyd Musolf.
Durham: Duke University Press.
 Bibliography 31 1
Vazquez Carrizosa, Alfredo. 1979. El Poder Presidencial en Colombia. Bogota:
Enrique Dobry, Editor.
Veloso Abueva, Jose, and Raul P. de Guzman, eds. 1969. Foundations and Dynam­
ics of Filipino Government and Politics. Manila: Bookmark.
Verney, Douglas. 1959. The Analysis of Political Systems. London: Compton Print­
ing Works.
Warnapala, W. A. Wiswa. 1980. "Sri Lanka's New Constitution. " In Asian Survey
20:914-30.
Warnapala, W. A. Wiswa, and L. Dias Hewagama. 1983. Recent Politics in Sri
Lanka: The Presidential Election and the Referendum of 1982. New Delhi:
Navrang.
Weber, Max. 1968. "Parliament and Government in a Reconstructed Germany. " In
vol. 3 of Economy and Society, edited by Guenther Roth and Claus Wittich.
New York: Bedminster Press.
Whitehead, Laurence. 1986. "Bolivia's Failed Democratization, 1977-80. " In Tran­
sitions from Authoritarian Rule: Latin America, edited by Guillermo O'Don­
nell, Philippe C. Schmitter, and Laurence Whitehead. Baltimore: Johns Hop­
kins University Press.
Wilson, A. Jeyaratnam. 1979. Politics in Sri Lanka: 1947-1979. London: Mac­
millan.
Wood, David M. 1968. "Majority vs. Opposition in the French National Assembly,
1956-65: A Guttman Scale Analysis. " American Political Science Review
62:88-109.
Zafrullah, H. M. 1981 . Sri Lanka's Hybrid Presidential and Parliamentary System.
Kuala Lumpur: University of Malaya Press.
 Index

accountability: 9, 34-5, 44-5, 49n, 90, 91, Bolivia: 137-8, 144, 149n, 161, 172, 284n
96, 102, 125 , 266, 274-5, 286 executive selection in: 81-5, 86-7, 92n,
Africa: 206, 217n, 283 188, 210, 212
Alessandri, Arturo: 85, 201, 204, 246, 247 party system: 224n, 241
Alessandri, Jorge: 87 Bork, Robert: 110
Alfonsfn, Raul: 72 Bosch, Juan: 89
Algeria: 59 Brazil: 3, 93, 136, 138, 140-1, 142, 149n,
Allende, Salvador: 32, 35n, 36, 87, 104, 151, 157, 174, 175, 183-4, 186, 188,
201-2, 203, 204, 217, 246, 247, 248, 190-1 , 199, 204, 205, 291
252, 265 electoral cycle: 221, 256
Alvarez, Innocente: 117 electoral rules: 176, 239n, 257n
amendments party system: 222, 237, 241n
constitutional: 72, 141, 191, 281 reform of 1961: 192
legislative: 139-40 breakdown of democracy: 38-43, 157-8
appointments, cabinet: 62, 106-11 , 117-26, Britain: see Great Britain
130, 152-3 Bruning, Heinrich: 70
arbiter, president as: 48-9, 53, 63, 73n Bulgaria: 9, 91-2, 163
Argentina: 20n, 72, 158 Burma: 172
electoral college: 20n, 210, 212, 216n Bush, George: 110, 118n
electoral cycle: 243, 265
electoral rules: 176 cabinet responsibility: see censure; parlia­
party system: 221 mentary confidence
Aristide, Jean Bertrand: 163n Canada: 45, 79-80, 228, 277n
assembly-independent regimes: 26, 78-85, censure of cabinet ministers: 68-9, 111-26,
94, 188 132, 153
Australia: 73 and appointments: 117-1 8
Austria: 7, 50n, 53n, 71-2, 80, 282 Chamorro, Violeta: 238
authoritarianism: 36-8 checks and balances: 19, 46-8, 52, 95, 106,
125-6, 141
Bagehot, Walter: 8, 171 Chile: 32, 35, 36, 103-4, 126, 129, 139, 140,
Balaguer, Joaquin: 89 141, 144, 146, 154, 157, 175, 177, 193,
Balmaceda, Jose Manuel: 114 204, 205, 210, 217n, 221 , 277, 290-1
Banzer, Hugo: 82 as "archetype": 174, 178-83, 203
Barre, Raymond: 271 constitutional design of: 193, 278
Batista, Fulgencio: 116 electoral cycles: 243-8, 251 , 253, 259,
Batlle, Cesar: 97 278
Batlle, Lorenzo: 97 electoral rules: 176, 224
Batlle Berres, Luis: 97, 115 executive selection in: 85-7
Batlle y Ordonez, Jose: 97 Parliamentary Republic: 1 13- 14, 138,
Belaunde Terry, Fernando: 36, 113 149, 157, 161, 275
Belgium: 12, 46, 274 party system: 178-83, 221, 224, 237,
Betancourt, Romulo: 89 241n, 256
bicameralism: ln, 273n reform of 1970: 200 , 280-1
314 Index
Chirac, Jacques: 57, 59-60, 108, 123-4 Eanes, Ramalho: 57n, 64-5
closed lists: see proportional representation, Ebert, Friedrich: 69
closed-list Ecuador: 89, 1 14, 137, 144, 149, 152, 158,
cohabitation: 56-8, 124, 267, 270, 271 217n, 258n, 282
colegiado (Uruguay): 97-9, 101, 102, 1 15 electoral cycle: 243, 265
collegial executive: see executives, collegial electoral rules: 178
Colombia: 74n, 92, 93n, 121, 137, 139, 151 , party system: 178, 223-4, 262
158, 174, 175, 177, 184-5, 1 86, 190-1, efficiency: 7-8, 8-9, 169, 215-16, 227, 228,
205 , 272 238, 242, 266, 273, 280
electoral cycle: 243, 264 El Salvador: 80-1, 123n, 144, 158, 217n,
electoral rules: 222, 239n, 257n 284n
emergency powers: 141-2, 144, 146 electoral cycle: 243, 250-1
executive selection: 188 election, popular: 4-6, 19, 20- 1 , 188
National Front: 36, 43n electoral college: 5, 15, 20n, 62, 77, 170-1 ,
party system: 222 212-13, 216n, 217, 225, 267-8
reform of 1991 : 141-2, 161, 199, 258n, electoral cycle: 32, 58, 92, 178, 188, 207-8,
282 212, 216n, 217, 219-21 , 223-4, 226-58,
concurrent elections: see electoral cycle 259-72
constitutional design: 14-15, 56, 198-9, concurrent: 86-7, 242, 260, 261
257-8, 283-6 counterhoneymoon: 243, 257, 264-5
co-presidency: 103-4 honeymoon: 242, 246-50, 251-3, 257,
Costa Rica: 122, 143, 158 , 177, 196, 197, 260, 265-72
198, 205, 277 nonconcurrent: 86-7, 171-2, 188, 263
electoral cycle: 261 electoral rules: 206-25
electoral rules: 176, 225 emergency powers: 131, 143-4
party system: 233, 237-8, 256 executives
presidential elections: 217, 288-90 collegial or plural: 20-1, 93-105, 219,
Cox, Gary: 168 285
Cuba: 1 16, 121, 138, 153, 158 , 159 properties of: 4
Cyprus: 99-101 , 102 selection by assembly: 78-87, 188
Czechoslovakia: 3
Federalist Papers: 14-15, 18-19, 21, 95, 105,
D'Aubuisson, Roberto: 81 106
de Gaulle, Charles: 6, 52, 57n, 59, 281 Figueres, Jose: 197
deadlock: 35, 47, 104, 126-7, 147 Finland: 7, 53, 61-3, 120, 124, 162, 217n,
decree powers: 69, 70, 131, 133, 140-6, 151, 259, 266-8, 269n, 282
279, 280 electoral college: 210, 212-13, 216n,
delegation of authority: 83-5, 188, 267, 267-8
280-1, 282 France: 6, 23, 39, 52, 57, 58-61, 66, 120,
democratization: 6, 39, 283-6 123-4, 142, 152, 154, 259, 281
Denmark: 9, 12 electoral cycle: 243, 250- 1 , 253, 263
Diaz, Porfirio: 88 emergency powers: 144
DiPalma, Giuseppe: 42, 283-6 Fourth Republic: 12, 172
dismissal of cabinet ministers: 109, 1 19, majority runoff in: 213-14, 289n
120-6, 153 party system: 214, 269-72
dissolution of assembly: 25, 53, 58, 65, 69, Third Republic: 6, 69
73, 106, 126-9, 154, 159 Frei, Eduardo: 35n, 103, 200, 201-2, 203,
district magnitude: 1 1 , 12, 21, 101 , 293 244, 246, 247, 252, 265
divided government: 55 Freitas do Amaral, Diego: 216
Dominican Republic: 89, 158, 172, 177, 196, Fujimori, Alberto: 33n, 1 13, 215
197, 198, 209, 233, 295 , 291
double complement rule: 217-18, 219, 225, Gallegos, Romulo: 89
242n, 260, 264, 278, 291 Garcia, Alan: 113
dual democratic legitimacies: 32-6, 5 1-2, Germany: 3, 9, 47, 273n
56-7, 165 Weimar Republic: 7, 57n, 68-71, 125,
Duarte, Jose Napoleon: 8 1 , 251 , 265 142, 157, 160, 161n, 162, 275, 282
Duverger, Maurice: 23, 206, 207 Gierek, Edward: 37
Duverger's law or rule: 206-8 , 293 Giscard d'Estang, Valery: 52
 Index 315
Gorbachev, Mikhail: 37 neo-parliamentarism: 163-4
Goulart, Joiio: 36, 93, 192 Nepal: 3, 172
Grau San Martin, Ramon: 1 16-1 7 Netherlands: 9, 80, 164, 294
Great Britain: 9, 37, 168-70, 173, 228, 284n Neustadt, Richard: 48
Greece: 213, 267, 274 new institutionalism: 1, 14
Guatemala: 89, 1 16, 135, 152 New Zealand: 173, 228, 274
Guyana: 3 Nicaragua: 89, 136, 149, 197, 198, 209, 238,
290
Haiti: 73, 163, 214n, 261n Nigeria: 3, 107, 1 10, 158, 159, 218, 225, 257n
Hamilton, Alexander: 21, 95-6 no-confidence vote: see censure
Haya de la Torre, Raul: 1 12 constructive: 12, 43, 47n, 153
Herrera, Luis Alberto de: 97 Norway: 45, 80, 173, 274, 284
Hindenburg, Paul von: 57n, 69, 70-1 Nujoma, Sam: 74, 90
Hitler, Adolf: 42n, 57n, 68, 7 1 number of parties, effective: 179
Honduras: 135
Hungary: 3, 9 origin and survival: see separation of powers
Ortega, Daniel: 89
Ibanez, Carlos: 179, 1 82, 244 outsider presidents: 33, 5 1 , 63, 73n, 215
Iceland: 7, 50n, 71-2, 120, 124-5, 140-1 ,
142, 162 Paasikivi, J. K.: 62
identifiability: 9, 45-6, 102, 267, 284, 285 Pacheco: 1 1 5
India: 39 Pakistan: 3
Indonesia: 163 Panama: 290
inefficiency: 8, 169, 229n Papen, Franz von: 70
Ireland: 7, 45, 73, 163, 219, 275 Paraguay: 37, 92, 126, 137, 154
Israel: 9, 12, 46, 79, 164, 213, 267, 282, 294 parliamentary confidence: 4, 7, 52 (see also
Italy: 46, 164, 172, 274 , 282 censure)
parliamentary system: 2-3, 10-12, 26-7,
Japan: 169, 273n 78-80, 1 19, 162-3, 168-70, 1 87, 265,
Jayewardene, J. R. : 66-7 267-8, 273, 274, 275
and democratic breakdown: 39-43
Kekkonen, Urho: 62-3, 266-8, 282 directly elected prime minister: 163-5
Kenya: 163 and number of parties: 226-9, 232-3
Koivisto, Mauno: 213, 267 Paz Estenssoro, Victor: 82
Korea: 107, 1 10, 140, 142, 162, 259, 290n Paz Zamora, Jaime: 82
Krushchev, Nikita: 37 Perez, Carloz Andres: 146
Peru: 33n, 74, 92, 93, 1 12- 13, 121, 125,
Lebanon: 80, 163, 284n 128-9, 144, 146, 154, 214-15, 217, 224,
legitimacy of democracy: 277, 283 258n, 262, 275 , 282
Lijphart, Arend: 20-1 , 30-2, 94-6 Philippines: 93, 107, 1 10, 122, 157, 159, 175,
Linz, Juan: 12, 29, 30, 34-5, 8 1 , 180, 273 204, 209, 222-3, 257, 277, 290
Pinochet, Augusto: 37
Madison, James: 21-7, 95, 205 plural societies: 77, 80, 93-4, 105, 218- 19
magnitude: see district magnitude plurality: 209, 213, 215-17, 219-23, 225, 288
Mainwaring, Scott: 175, 180 concurrent pluralities: 218, 225
majoritarianism: 30-2, 47, 48 , 50-1 , 94 Poland: 138, 215, 285
majority runoff elections: 210, 213-16, 225, Pompidou, Georges: 57n, 59
250-1 , 258, 264, 288, 289n, 291 Portugal: 57n, 63-5, 120, 125, 153, 154, 157,
mandates: 9, 32, 45, 57-8 161, 2 16, 266
Marcos, Ferdinand: 290 Powell, G. Bingham: 273-4
Mexico: 37, 88, 92, 257n premier-presidentialism: 53-7, 108, 1 19-20,
minor parties: 253-7 123-5, 135, 138, 154, 158, 159-60, 275,
Mitterrand, Frarn;ois: 52, 59-60, 108 , 123- 282-3
4, 242, 251 , 252, 263, 265, 268-9, 271 advantages of: 49-53
Montesquieu: 18 defined: 23-4
president-parliamentarism: 53-7, 1 1 8 , 125,
Namibia: 73-4, 90, 121, 138, 143, 154, 161n, 138, 160-1, 162, 164, 275, 282, 283
261-2, 285 defined: 24
316 Index
presidential powers Sri Lanka: 65-7, 92, 157, 160, 161n, 162,
budgetary initiative: 139-40, 152 219
delegated: 131-3, 144-6 Stroessner, Alfredo: 37
introduction of legislation: 139-40, 147, Suksi, Markku: 213
151 Supreme Court: 1 10, 122-3
legislative: 131-47, 148-52, 154-8 Sweden: 9, 45, 80, 274
nonlegislative: 152-8 Switzerland: 26, 78, 80, 84-5, 102, 161, 284n
regulatory: 143
see also, decree powers; dismissal; disso­ Taagepera, Rein: 293
lution; emergency powers; referendum; Taif accord: 80
vetoes temporal rigidity: 28-30, 34, 50
presidentialism: 4-6, 12-15, 158, 159, 165 Thatcher, Margaret: 37
advantages of: 43-9 third parties: see minor parties
archetype: 174, 178-83, 204, 241 , 242 ticket-splitting: see split-ticket voting
and authoritarianism: 36-8 Tower, John: 1 10
defined: 19 Trujillo, Rafael: 197
and democratic breakdown: 38-43 Turkey: 41, 163
efficient: 194 Tyminski, Stanislaw: 215, 216
"optimal": 198, 205
Preuss, Hugo: 69 United Kingdom: see Great Britain
principal-agent relations: 14, 187 (see also United States: 4-6, 20n, 35, 51n, 77, 88,
delegation of authority) 132-3, 158, 159, 171-2, 173, 178 , 185-
proportional representation: 4, 9, lln, 30, 6, 194, 225, 289n
45n, 102 appointment powers: 107-10, 1 17, 1 12-3
closed-list: 171 , 194, 239, 257n electoral college: 2 10, 212, 217, 218
deviation from: 30 electoral cycle: 243, 265
electoral rules: 257
party system: 222-3
Quadros, Janio: 93, 192 Uruguay: 97-8, 101-2, 127-8, 139-40, 154,
159, 172, 175, 178
Redslob, Robert: 69
reelection, limitations on: 35, 87-91 , 105, Vargas, Getulio: 191
191 , 198 Vargas Llosa, Mario: 33n, 215
referendum: 59, 66, 69, 152 Venezuela: 20, 89-90, 92, 122, 136, 144,
representativeness: 9, 3 1 , 208, 242, 273, 286 146, 158, 177, 194-7, 1948, 205, 209,
Rho Tae Woo: 290n 277
Riker, William: 179, 206 constitutional design of: 197-8, 278-80,
Romania: 261 n 285-6
Roosevelt, Franklin: 8 8 electoral cycle: 248-50, 256, 261 , 278
electoral rules: 176, 221 , 238n, 239n,
240n
Sanguinetti, Julio: 127-8 party system: 221 , 233, 241 , 296-9
Sartori, Giovanni: 22 Verney, Douglas: 20
Schelling, Thomas: 61, 136 vetoes: 19, 63, 65, 134-8, 144-6
Schleicher, Kurt von: 71 partial or item: 134-5, 136-7, 149-51 ,
semipresidentialism: see premier- 277, 280
presidentialism provisions for override: 135-8, 149
separation of powers: 15, 1 8 vice-presidency: 91-3
origin and survival: 19, 22, 158-65 Virginia Plan: 77
Siles Suazo, Hernan: 82
Soares, Mario: 65, 216 Walesa, Lech: 215, 216
South Africa: 218-19 Weber, Max: 69
Spain: 47n, 256n, 274, 284n
split-ticket voting: 239, 300 Zimbabwe: 3