Boondoggles, Rent-Seeking, and Political Checks and Balances: Public Investment under

Unaccountable Governments
Author(s): Philip Keefer and Stephen Knack
Source: The Review of Economics and Statistics, Aug., 2007, Vol. 89, No. 3 (Aug., 2007), pp.
566-572
Published by: The MIT Press
Stable URL: [Link]

REFERENCES
:Linked references are available on JSTOR for this article
[Link]
You may need to log in to JSTOR to access the linked references.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide
range of content in a trusted digital archive. We use information technology and tools to increase productivity and
.facilitate new forms of scholarship. For more information about JSTOR, please contact support@[Link]

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
[Link]

The MIT Press is collaborating with JSTOR to digitize, preserve and extend access to The Review of Economics
and Statistics

This content downloaded from

[Link] on Thu, 31 Jul 2025 [Link] UTC
All use subject to [Link]
 BOONDOGGLES, RENT-SEEKING, AND POLITICAL CHECKS
AND BALANCES: PUBLIC INVESTMENT UNDER
UNACCOUNTABLE GOVERNMENTS
Philip Keefer and Stephen Knack*

Abstract- We show that public investment is dramatically higher in

plored implications of institutions that limit government
countries with low-quality governance and limited political checks and
balances or no competitive elections. This result is robust to a number of (elections and political checks and balances). The evidence
specifications. The most plausible interpretation of these results is that below demonstrates that institutional variables that else-
these governments use public investment as a vehicle to increase their where in the literature are taken as proxies for limited
rent-seeking. This evidence suggests that efforts to increase public invest-
ment in countries with weak governance, or to measure the growth effects government are associated with both less corruption and
of productive public investment using only observed measures of public less (but, we argue, more productive) public investment,
investment, should be undertaken with caution. results that are consistent with predictions by Acemoglu
(2005) about the effects of limited government. Finally,
I. Introduction
previous work has linked corruption to distortions in gov-
ernment spending. One obvious policy conclusion from
and international aid agencies regard
POLICYMAKERS such work is the need to fight corruption more vigorously.
productive public investment as essential for economic
Sachs and Our findings suggest that this may be ineffective in com-
development. colleagues (2004) argue strongly
that high transport costs and other infrastructure weaknesses bating the deeper institutional difficulties that we associate
with those same spending distortions.
are key bottlenecks to African development. The Commis-
sion for Africa report (2005, ch. 7) urges donor nations to
provide $10 billion per year more in aid for African infra- II. Measuring Corruption, Property Rights Security,
structure from 2005-2010, with a further increase to $20 and Political Checks and Balances
billion per year over the following five years. Public invest-
The conclusions here, that public investment rises when
ment is also well-known as a vehicle for rent-seeking,
however. If rent-seeking is a sufficiently powerful motiva- governments have greater incentives to seek rents, depend
tion for public investment expenditures, we might expect an fundamentally on the quality of the two measures of rent-
association between government incentives to seek rents seeking incentives we use. On the one hand, rent-seeking
can be viewed as the extent to which governments tax the
and the quantity of public investment. The results presented
fruits of citizen effort and retain the revenues from these
here show that levels of observed public investment,
taxes for their own purposes. Citizen effort drops when
whether as a fraction of national income or of total invest-
taxes are high. Governments that engage in significant
ment, public and private, are substantially higher in coun-
tries that exhibit low levels of a composite measure of rent-seeking therefore do so because they believe that their
share of the long-run rents from high citizen effort are lower
expropriation and contract repudiation risk, law and order, than the short-run rents that they can extract at the expense
corruption, and bureaucratic quality. Public investment is
of citizen effort (Acemoglu, 2005; Clague et al., 1996). For
also much higher in countries that have noncompetitive
elections and few political checks and balances. example, governments that expect to be expelled from office
The most plausible explanation for the findings reported regardless of whether they perform well have little incentive
to restrain their rent-seeking behavior.1 Such governments
below is that extra public investment associated with weak
also have little reason to refrain from expropriation or from
institutions is unproductive and largely intended to steer
the repudiation of contracts, nor do they have an incentive
rents to government officials or their cronies. These results
to ensure that citizens benefit from the rule of law or from
therefore signal the need for donor agencies to exercise
the services of high-quality bureaucracies.
particular caution in supporting public investment in coun-
Our first indicator of government rent-seeking incentives
tries with a weak institutional environment. They also have
captures these effects. It is a composite of subjective eval-
implications for a number of research directions that are uations of different dimensions of country risk prepared by
prominent in the literature. the PRS Group, a firm specializing in political risks to
The results first suggest some rethinking of research
foreign investment, and published as the International
strategies that use observed public investment data to es-
tablish the effects of public investment on growth. In addi- Country Risk Guide (ICRG). This is a familiar variable; the
particular index here was first used in Keefer and Knack
tion, the results here reveal additional and previously unex-
(1995). The ICRG is provided monthly to subscribers,
Received for publication April 30, 2003. Revision accepted for publi-
mainly multinational investors. To examine this relationship,
cation March 6, 2006.
* Both authors are affiliated with the World Bank. 1A
government laboring under significant negative shocks whose con-
The opinions expressed here are solely those of the authors and do not sequences are attributed by citizens to the government's own actions
represent the views of the World Bank or its directors. would confront such a situation.

The Review of Economics and Statistics, August 2007, 89(3): 566-572

© 2007 by the President and Fellows of Harvard College and the Massachusetts Institute of Technology

This content downloaded from

[Link] on Thu, 31 Jul 2025 [Link] UTC
All use subject to [Link]
 BOONDOGGLES, RENT-SEEKING, AND POLITICAL CHECKS AND BALANCES 567

we use five variables from the ICRG to form an additive except that citizens have no reason to acquiesce to such a
index, "quality of governance."2 These variables are bureau- decision by government. To protect their privileged access
cratic quality, the risk of expropriation and of repudiation of to natural resources or other exogenous sources of rents,
contracts by government, corruption, and the law and order governments restrict citizen influence on political decisions
tradition of the country. The index is measured on a scale of and political careers. Political checks and balances are few
0-50, with higher values indicating better quality of gover- and elections absent.5 Regardless of the historical genesis of
nance. The median values of these are used over the period these institutional arrangements, governments that do not
1974-1998.3 exhibit political checks and balances or that suppress elec-
The ICRG assigns low ratings of its corruption variable toral competition may have a more difficult time making the
("corruption in government") to countries in which top credible commitment to citizens that their assets will be safe
government officials are likely to demand special payments from expropriation after they have exerted effort to accu-
and where illegal payments are generally expected through- mulate them.6 Such governments have no reason to refrain
out lower levels of government. The quality of the bureau- from expropriation, since citizens have no reason to expect
cracy assesses the degree to which the bureaucracy has the that they will, so again we should observe high levels of
strength and expertise to govern without drastic changes in rent-seeking.
policy or interruption of governmental services. Such bu- Our second measure of accountability captures these
reaucracies have established mechanisms for recruitment basic institutional characteristics of countries. The Database
and training, and some autonomy from political pressure. If of Political Institutions (DPI) (Beck et al., 2001) contains an
politicians are unconstrained in the demands that they can objective measure of checks and balances cum elections,
place on bureaucracies, or if they are subject to frequent "checks." The measure is a function of the number of parties
replacement, bureaucratic quality is likely to be lower.4 in the government coalition (for parliamentary systems),
Expropriation risk tracks the risk of confiscation and forced whether the president's party has a majority in the legisla-
nationalization while the risk of contract repudiation as-
ture (presidential systems), and whether elections are gov-
sesses the risk that governments will repudiate or otherwise
erned by closed-list or open-list rules (the former granting
unilaterally change the terms of contracts with foreign more authority to the heads of parties). Since constitutional
businesses. Finally, the extent to which disputes in a country
checks on executive behavior typically mean little if the
are resolved legally and through formal channels is mea-
relevant actors are not elected, the construction of checks
sured in the variable "law and order tradition."
also takes into account the DPFs legislative index of elec-
The quality of governance is a direct measure of the
toral competitiveness (LIEC), scaled one to seven.7 The
incentives of governments to seek rents and to refrain from
average value of checks, 1974-1998, is used in the analysis
establishing institutions that would limit their ability to seek
below.8
rents. The multidimensional index is more informative than
To measure public investment, we follow Levine and
the simple corruption measure for several reasons. First,
Renelt (1992) and Devarajan, Swaroop, and Zou (1996) by
rent-seeking need not be corrupt. On the contrary, efforts to
channel benefits away from citizens to political officials and taking data on public investment from the Government
their constituencies are usually legal. Second, political in-
Finance Statistics (GFS) of the International Monetary
stitutions could limit rent-seeking behavior by politicians Fund. While GFS has some data on public investment by
but bureaucratic corruption could nevertheless be high, state and local governments, its most complete and reliable
because of capacity constraints due to poverty or lack of coverage is of central government expenditures, not includ-
experience in exercising oversight over bureaucrats. Thus,
5
low scores on the index reflect distortions associated with Acemoglu, Johnson, and Robinson (2001) make this argument.
6 These fundamental institutions- elections and checks and balances-
"excessively strong" or "excessively weak" states (Acemo- are the focus here and emphasized by, among many others, North and
glu, 2005). The five dimensions of the quality of gover- Weingast (1989) and Acemoglu, Johnson, and Robinson (2001). However,
incentives to extract rents also vary with the detailed institutions of
nance, taken together, offer a clearer picture of political incen-
democracy (whether governments are parliamentary or presidential, for
tives to pursue rents and to undermine or refrain from building
example), as in Persson and Tabellini (2000).
institutions that would restrain political rent-seeking. 7 Where there are no elections, countries receive a one; the scores rise to
In addition to setting high taxes in anticipation of an early seven when there are multiple candidates and multiple parties, and no
single party or candidate receives more than 75% of the vote. If the
departure from power, governments may also wish to deny legislative index of electoral competitiveness is less than five (where five
citizens access to rents (for example, from natural re- indicates that multiple parties can legally be established, but where only
one party wins any seats in the legislature), checks is always one. For
sources). This might have little influence on citizen effort other coding rules, see Beck et al. (2001).
8 Other
commonly used measures of democracy, such as Polity IV or
2 The results we Freedom House, extend across a wider set of democratic characteristics,
report are not sensitive to the manner in which the index
is constructed. including performance characteristics (such as human rights guarantees).
3 Period medians are more appropriate for ordinal data like the ICRG These additional characteristics cloud the essential role of elections and
indicators, but are in any case correlated at 0.99 with the period means. checks and balances; because of the subjectivity of these variables, it is
4 Noll and Fiorina (1978) is one or the first ol a growing literature not possible to ensure that these multiple characteristics are equally
examining the interaction of political and bureaucratic agents. weighted across countries or over time.

This content downloaded from

[Link] on Thu, 31 Jul 2025 [Link] UTC
All use subject to [Link]
568 THE REVIEW OF ECONOMICS AND STATISTICS

Table 1. - Descriptive Statistics ^___

Multivariate Sample Settler Mortality Sample

Mean [Link]. N Mean [Link]. N

Public investment/GDP 5.2 3.6 89 5.3 3.4 51

Public investment/total investment 18.4 11.6 86 19.5 12.1 49
Quality of governance index 31.3 11.2 80 26.6 8.3 45
Checks 2.9 1.6 89 2.4 1.2 51
Area (in logs) 16.7 2.1 89 16.9 2.5 51
Population (in logs) 15.7 1.7 89 15.6 1.7 51
Left party largest (prop, of years) 0.32 0.36 89 0.34 0.39 51
Price of investment goods, U.S. = 100 96.5 46.0 89 104.6 53.6 51
Per capita income (logs) 8.4 1.0 87 8.1 0.9 50
Settler mortality (logs) 4.5 1.2 51 4.5 1.2 51
Degrees from equator 25.0 17.1 89 17.1 11.4 51

Multivariate sample includes countries in either equation (3) or (4) in tables 2 and 3. Settler mortality sample includes countries in either equations (1) or (3) of table 4.

ing investments by state-owned enterprises.9 We look at or quality of governance. If the sample is divided into two
average public investment over the period as a fraction of equal-sized groups (those above and below checks equal to
GDP and of total investment. 2.5, or those above and below a score of 30 on quality of
A number of other possible determinants of public invest- governance), public investment in the low checks sample is
ment are taken into account in the analysis below. The 6.7% of GDP and in the low quality of governance sample
returns to public investment may depend on the amount of it is 6.1%. Public investment is 3.9% of GDP in the high
initial public infrastructure already in place as well as the checks sample and 3.6% in the high quality of governance
overall level of economic income. We therefore control for
sample.
initial purchasing power parity-adjusted income per capita These differences persist in the analysis below, using
(Summers & Heston, 1991) to capture the effect of both.
ordinary and two-stage least squares estimates of the
The land area of a country (in logs) and the log of initial determinants of average public investment as a fraction
population are included because the economic returns to of GDP and of total investment over the period 1974-
public investment may vary with the size of a country. The 1998. This time period is driven by the availability of the
political demands made on governments by citizens may DPI (which begins in 1975) and ICRG variables (which
vary as well with size.
start in 1982). Like most empirical efforts using these
Spending preferences of governments may be correlated
institutional data, which vary relatively little over time
with their ideological tendencies (left, favoring more redis-
but substantially across countries, we focus our analysis
tribution and government intervention in the economy, or
on cross-section averages.
right, favoring less). The DPI provides information on
Results in table 2 are OLS estimates of determinants of
parties' ideological tendencies, so we control for the per-
centage of years from 1975 to 1998 that the largest party in public investment/GDP. The first two columns display a
the legislature is coded in DPI as left-leaning. Investment is large bivariate association between the quality of gover-
nance and checks and balances. The association changes
likely to be influenced by the price of investment goods, so
we account for the initial price level of investment goods in little when controls are added in the remaining columns. A
a country relative to prices in the United States (Summers & ten-point increase in the quality of governance is associated
Heston, 1991). The U.S. value is 100. Public investment/ with a reduction in public investment of 1.2% of GDP, not
GDP should be lower where the initial price index is higher; controlling for income per capita. The effect of an increase
public investment/total investment should rise if private in checks by one point on the seven-point scale is nearly the
investors are more sensitive to investment prices. Table 1 same. Income per capita is highly correlated both with the
presents the summary statistics on these variables. quality of governance (as Acemoglu, Johnson, & Robinson,
2001, also report) and with checks. Nevertheless, results are
III. Governance, Checks and Balances, and robust to the inclusion of initial income per capita in
Public Investment columns 5 and 6. Initial income per capita is itself not a
A simple exercise suggests that there are dramatic differ- significant determinant of public investment, nor are the
other controls.
ences between countries with high and low values of checks
The dependent variable shifts to public investment/total
9 Barro (1991) investment in table 3, but the results are similar. A ten-point
appears to use the general government public investment,
including decentralized government expenditures from GFS where avail- increase in quality of governance is associated with a
able. For a few dozen countries, Easterly and Rebelo (1993) supplement reduction in the ratio of public to total investment by 3 or 4
GFS data on central government public investment with World Bank
country reports that most notably include data on investments by state- percentage points, depending on the specification. A one-
owned enterprises. point increase in checks reduces public investment/total

This content downloaded from

[Link] on Thu, 31 Jul 2025 [Link] UTC
All use subject to [Link]
 BOONDOGGLES, RENT-SEEKING, AND POLITICAL CHECKS AND BALANCES 569

Table 2. - Determinants of Public Investment/GDP

OLS Regressions
1 2 3 4 5 6

Quality of governance -0.120 -0.121 -0.093

(period median) (0.027) (0.031) (0.048)
Checks -1.145 -1.171 -1.017
(period mean) (0.206) (0.196) (0.229)
Area -0.294 -0.331 -0.286 -0.343
(logsqkm) (0.225) (0.215) (0.239) (0.230)
Population -0.305 -0.211 -0.382 -0.237
(log of initial) (0.352) (0.321) (0.377) (0.354)
Mean years that largest party 0.529 0.336 0.405 0.291
is left-leaning (1.111) (0.943) (1.094) (0.944)
Price of investment 0.004 -0.001 0.003 -0.002
goods (initial) (0.007) (0.007) (0.008) (0.007)
Log per capita income -0.404 -0.362
(initial) (0.636) (0.462)
R2 0.16 0.23 0.30 0.36 0.29 0.36
No. ofobs. 93 114 79 88 77 86
Dependent variable: public investment/GDP, in percent, mean 1974-98 (from IMF's Government Finance Statistics). Quality of governance is an equal-weighted index of the median values over 1982-98 of the
ICRG variables corruption in government, bureaucratic quality, expropriation risk, repudiation of contracts by government, and law and order tradition. The index is measured on a scale of 0-50, with higher values
indicating better quality of governance. Checks is the number of veto players in the political system (median value for 1974-98). Robust standard errors are in parentheses.

investment by 3.5%. 10Again, adding income per capita in the literature: the log of settler mortality (Acemoglu,
might be expected to weaken these relationships, but they Johnson, & Robinson, 2001), available for former colonies,
remain large and significant with its inclusion in columns 5 is used in the odd-number equations; the degrees of latitude
and 6. As in table 2, income is again insignificant in table 3, from the equator (Hall & Jones, 1999) in the even-numbered
as are most other controls. The price of investment goods is ones. They are used interchangeably as instruments for
significant and positive; private investors may be more quality of governance and checks.11 Panel B in table 4
sensitive to the price of investment goods than public indicates that both instruments are strong predictors of the
investors. quality of governance and checks. Panel A shows that the
These results are robust to alternative specifications. For predicted values of quality of governance and checks are
example, using alternative estimation procedures (median significant and negative, as in the earlier tables. The mag-
regression and robust regression) that downweight the in- nitude of their effect is more than 50% larger than in the
fluence of outlying observations has little effect on these OLS coefficients reported in panel C, using the same sam-
results. Coefficients for checks and quality of governance ples.12
change very little in all of these tests but one, and in that These results would be spurious if the instruments were
case the coefficient of quality of governance substantially correlated with the error terms of the regressions in tables 1
increases in absolute value. and 2. This is not the case. The instruments are each
Although there are good reasons to expect a causal insignificant when added to the OLS specifications of tables
relationship between government accountability and high 1 and 2. We also fail to reject the null hypothesis that
public investment, tables 2 and 3 are insufficient to docu- exclusion of both instruments is appropriate.13
ment such a relationship. Omitted effects could also be
responsible for the relationships identified in these tables. IV. Discussion of Results
For example, internal strife could lead governments to
abandon checks and balances, reduce accountability, and at The inverse association between observed public invest-
the same time deter private investment and lower incomes. ment and the quality of governance or political checks and
Alternatively, ICRG evaluators may, lacking better informa-
tion, infer corruption or expropriation risk from public or
11Each of these instruments is
expected to raise incomes through its
effects on institutions: conditions that reduce settler mortality promote the
private investment or GDP. This might bias upward the creation of institutions that in turn sustain higher growth over long
quality of governance coefficients in tables 2 and 3. Table 4, periods. We therefore expect the correlation of income per capita with
however, provides evidence in support of a causal link predicted accountability variables to be higher than with the raw variables
in tables 2 and 3. This is the case, so table 4 omits income per capita.
among these phenomena. 12Table 4 results could differ from those in tables 2 and 3 because of the
Using as base specifications models 3 and 4 from tables smaller sample of countries for which settler mortality data are available.
2 and 3, the instrumental variable estimates in table 4 use The coefficients in panel C of table 4 indicate that the coefficients on
two instruments for institutions that have a prominent place checks and quality of governance are unaffected by the change in sample,
however.
13
The/?- values from these over-identification tests are, for using checks,
10The standard deviation of the quality of governance is approximately 0.47 (public investment/total investment) and 0.20 (public investment/
11.3 for those countries appearing in any of regressions 3-6 in tables 1 or GDP). The corresponding values for quality of governance are 0.13 and
2; the standard deviation of checks is 1.62. 0.24.

This content downloaded from

[Link] on Thu, 31 Jul 2025 [Link] UTC
All use subject to [Link]
570 THE REVIEW OF ECONOMICS AND STATISTICS

Table 3. - Determinants of Public Investment/Total Investment

OLS Regressions
12 3 4 5 6

Quality of governance -0.392 -0.372 -0.286

(period median) (0.081) (0.095) (0.159)
Checks -3.622 -3.531 -2.404
(period mean) (0.612) (0.637) (0.700)
Area -0.765 -0.668 -0.657 -0.632
(logsqkm) (0.605) (0.814) (0.608) (0.800)
Population (log -0.225 -0.176 -0.580 -0.596
of initial) (0.795) (0.884) (0.871) (1.007)
Mean years that largest 0.243 -1.580 0.024 -1.818
party is left-leaning (2.806) (2.991) (2.759) (3.082)
Price of investment 0.055 0.064 0.047 0.055
goods (initial) (0.019) (0.022) (0.022) (0.023)
Log per capita income -1.146 -2.590
(initial) (2.038) (1.665)
R2 0.16 0.23 0.35 0.36 0.33 0.38
No. of obs. 93 114 78 85 76 83

Dependent variable: public investment/(public + private investment), in percent, mean 1974-98 (from IMF's Government Finance Statistics). Quality of governance is an equal-weighted index of the median values
over 1982-98 of the ICRG variables corruption in government, bureaucratic quality, expropriation risk, repudiation of contracts by government, and law and order tradition. The index is measured on a scale of 0-50,
with higher values indicating better quality of governance. Checks is the number of veto players in the political system (median value for 1974-98). Robust standard errors are in parentheses.

balances has two possible explanations. One seems more unable to demonstrate their credibility. They therefore in-
plausible, but we can discard neither based on the evidence crease productive public investment in order to offset the
presented above. First, in countries where governance or fact that their lack of credibility has driven off private
checks on political behavior are weak (where rulers are investment. This explanation seems less plausible because
strong), governments face fewer political costs from rent- governments unable to make credible commitments (for
seeking. Because public investment is a favored vehicle for example, governments with short horizons) also tend to
extracting rents, observed public investment rises. This have fewer incentives to promote growth. The findings here
result is consistent with the argument in Acemoglu (2005) therefore at least advise caution when increasing public
that strong rulers are freer to extract revenues for their own investment in countries with a weak institutional environ-
benefit. The evidence is insufficient to test the further ment.
prediction that strong states that are unable to collect reve- The findings also have implications for a number of
nues will undertake lower productive public investment, research directions prominent in the literature. A significant
since productive public investment is unobserved. body of research uses observed public investment data to
It could be, however, that public investment is not dis- assess the impact of public investment or public infrastruc-
proportionately used to seek rents. This does not seem to be ture on economic growth. Pritchett (1996) and others have
the case. From pork barrel spending in the United States to argued that observed public investment is a poor proxy for
white elephant steel factories in Nigeria, public investment productive investment. Our results suggest that it may not
has served purposes other than maximization of growth, only be a noisy indicator, but a systematically biased one as
ranging from securing political support to increasing per- well: there may be an inverse correlation between observed
sonal fortunes. In Turkmenistan, where roads were crum- and productive public investment. Attempts to use observed
bling and water was unavailable for hours on end, the public investment to assess the effect of public investment
authorities built an international airport with the capacity to on growth are unlikely to yield accurate estimates without
receive 4.5 million visitors a year, though only a few controlling for the possibility of this inverse correlation.
hundred thousand used the airport; authorities further in- Another line of research investigates the effects of cor-
sisted on building the control tower on the wrong side of the ruption on government decision making and specifically on
terminal, blocking the controllers' view of the runway.14 the allocation of public spending. Tanzi and Davoodi (1997)
Statistical evidence suggests as well that public investment conclude that public investment is strongly associated with
improves the quality of infrastructure only when the quality greater corruption. They argue explicitly that this "extra"
of governance is high (Keefer & Knack, 2002). 15 public investment is aimed at rent-seeking. Mauro (1998)
The second explanation for the inverse association of finds similar, though weaker, evidence of the same phenom-
public investment and governance is that governments with enon. Our findings, particularly the strong relationship be-
no interest in rent-seeking nevertheless find themselves tween public investment spending and political checks and
balances and competitive elections, indicate that the con-
14"Palaces and in Central Asia," Washington Post, November nection between government decisions and corruption is the
Poverty
11, 1994, p. A35. Robinson and Torvik (2005) provide more examples.
15There, the product of deeper phenomena that may influence both
quality of infrastructure is approximated by an index of
such indicators as electricity losses, kilometers of paved roads, and access corruption and public investment simultaneously. Support-
to adequate sanitation. ing this, the results reported above are equally strong using

This content downloaded from

[Link] on Thu, 31 Jul 2025 [Link] UTC
All use subject to [Link]
 BOONDOGGLES, RENT-SEEKING, AND POLITICAL CHECKS AND BALANCES 571

Table 4. - Instrumental Variables Estimates of the Determinants of Public Investment/GDP

Public Investment/GDP (%) Public Inv./Total Inv. (%)

12 3 4 5 6 7 8

Panel A: Two-Stage Least Squares

Quality of -0.166 -0.179 -0.603 -0.466

governance (0.099) (0.043) (0.285) (0.116)
Checks -1.515 -1.356 -6.621 -4.484
(0.814) (0.338) (3.306) (1.157)
Area -0.472 -0.402 -0.313 -0.373 -1.364 -0.936 -0.738 -0.885
(logsqkm) (0.274) (0.233) (0.251) (0.216) (0.780) (0.617) (0.976) (0.713)
Population (log of -0.023 -0.138 -0.136 -0.136 0.881 0.043 0.062 0.222
initial) (0.398) (0.300) (0.369) (0.300) (1.136) (0.794) (1.449) (0.994)
Left-party largest 2.758 0.741 2.475 0.434 5.519 0.632 4.643 -0.826
(mean) (1.121) (0.932) (1.111) (0.877) (3.330) (2.537) (4.745) (3.065)
Price of investment 0.001 0.002 -0.002 -0.001 0.047 0.052 0.044 0.058
goods (initial) (0.009) (0.008) (0.008) (0.007) (0.025) (0.022) (0.036) (0.027)
Panel B: First-Stage Regressions

Log settler -3.676 -0.438 -3.661 -0.416

mortality (0.959) (0.134) (0.976) (0.136)
Degrees lat. 0.450 0.056 0.455 0.054
from equator (0.051) (0.008) (0.053) (0.008)
Area -0.127 -0.976 -0.047 -0.149 -0.124 -0.982 -0.043 -0.152
(logsqkm) (0.717) (0.586) (0.092) (0.083) (0.727) (0.589) (0.093) (0.085)
Population (log of 0.859 1.370 0.057 0.225 0.865 1.349 0.068 0.234
initial) (0.980) (0.740) (0.134) (0.112) (0.993) (0.744) (0.136) (0.114)
Left-party largest 1.968 2.235 0.398 0.374 2.098 1.913 0.552 0.477
(mean) (2.882) (2.435) (0.388) (0.352) (3.029) (2.512) (0.407) (0.376)
Price of investment 0.027 -0.010 -0.001 -0.003 0.027 -0.010 -0.001 -0.003
goods (initial) (0.024) (0.022) (0.003) (0.003) (0.024) (0.022) (0.003) (0.003)

# 33 ^58 ~21 49 33 ^58 ^28 47

Panel C: OLS

Quality of -0.110 -0.121 -0.360 -0.373

governance (0.055) (0.032) (0.165) (0.095)
Checks -1.279 -1.171 -4.230 -3.531
(0.380) (0.196) (1.267) (0.637)

No. of obs. 45 79 51 88 44 78 49 85

Panel A reports the two-stage least squares estimates with public investment/GDP (in percent, mean 1974-98) as the dependent variable in columns 1-4 and public investment/total investment (in percent, mean
1974-98) as the dependent variable in columns 5-8, instrumenting for the institutional variable with log settler mortality and/or degrees from equator. Panel B reports the corresponding first-stage regressions. Panel
C reports the OLS coefficient from regressing the indicated public investment variable on the indicated institutional variable, for the same (2SLS) sample, with the other control variables listed in panel A (full results
not reported to save space).

a quality of governance index that excludes the corruption In our tables 2 and 3 regressions using checks, an objec-
variable; at the same time, corruption used alone is not as tive measure of two key democratic institutions (elections
statistically significant. and political checks and balances), checks remains a signif-
Others have found no connection between institutions- icant determinant of public investment even when we con-
- same noninstitutional measures.16 Even if
specifically, democracy and many government spending trol for these
decisions. Mulligan, Gil, and Sala-i-Martin (2004) contrast citizens cannot systematically influence who benefits from
two arguments. On the one hand, the presence of democratic redistributive government spending, and provided that pub-
institutions should influence public spending by giving the lic investment is, indeed, a preferred vehicle for rent-
poor greater opportunity to redistribute spending to their seeking, the robustness of these results implies that demo-
own benefit. On the other, all leaders, democratic or not, cratic institutions influence the degree to which citizens can
seek to maximize the gains from dealing with interest control rent-seeking behavior by politicians.
groups, so that underlying economic interests rather than The findings here also contribute to the literature on the
political institutions should determine cross-country differ- effects of credible commitment on policymaking. Typically,
ences in observed public spending outcomes. Mulligan et al.
find that noninstitutional variables are significant determi- 16Their five noninstitutional variables are a communist dummy, a
nants of social and education spending, while their measure British legal origin dummy, the log of population, the fraction of the

of democracy is not. Our results suggest that political population over 65, and the log of initial income. Adding British legal
origin and the fraction of the population over 65 does not change our
institutions affect spending at least to the extent that they results; mean years that the largest party is left-wing captures the same
facilitate citizen control of political rent-seeking. effect as the communist dummy.

This content downloaded from

[Link] on Thu, 31 Jul 2025 [Link] UTC
All use subject to [Link]
572 THE REVIEW OF ECONOMICS AND STATISTICS

this literature has found that governments unable to make Barro, Robert, "Economic Growth in a Cross Section of Countries,"
credible commitments create a greater risk of expropriation Quarterly Journal of Economics 106:2 (1991), 407-^44.
Beck, Thorsten, George Clarke, Alberto Groff, Philip Keefer, and Patrick
on citizens or limit government ability to conduct monetary Walsh, "New Tools in Comparative Political Economy: The Data-
base of Political Institutions," World Bank Economic Review 15:1
policy and that political checks and balances mitigate the
(2001), 165-176.
credibility problem.17 Acemoglu (2005) offers theoretical Clague, Christopher, Philip Keefer, Stephen Knack, and Mancur Olson,
support for the argument that limited government should "Property and Contract Rights in Autocracies and Democracies,"
also affect government spending. The results here support Journal of Economic Growth 1:2 (1996), 243-276.
Commission for Africa, Our Common Interest: Report of the Commission
this line of analysis, demonstrating that political checks and
for Africa, [Link]
balances also influence the composition of government [Link] (2005).
spending, and may do so because of the influence they have Devarajan, Shantayanan, Vinaya Swaroop, and Heng-fu Zou, "The Com-
on government incentives to seek rents. position of Public Expenditure and Economic Growth," Journal of
Monetary Economics 37:2 (1996), 313-344.
Easterly, William, and Sergio Rebelo, Fiscal Policy and Economic
V. Conclusion Growth: An Empirical Investigation," Journal of Monetary Eco-
nomics 32:3 (1993), 417^58.
We document that observed public investment as a frac- Hall, Robert E., and Charles I. Jones, "Why Do Some Countries Produce
So Much More Output per Worker than Others?" Quarterly Jour-
tion of national income or total investment is much higher in nal of Economics 114:1 (1999), 83-116.
countries that exhibit weak governance or few political Keefer, Philip, and Stephen Knack, "Institutions and Economic Perfor-
checks and balances. Further research is needed to deter- mance: Cross-Country Tests Using Alternative Institutional Mea-
sures," Economics and Politics 7:3 (1995), 207-227.
mine the extent to which public investment in these coun-
"Boondoggles and Expropriation: Rent-Seeking and Policy Dis-
tries is less productive and intended to serve the private tortion when Property Rights are Insecure," World Bank policy
interests of government officials, or is rather an effort by research working paper no. 2910 (2002).
Keefer, Philip, and David Stasavage, "The Limits of Delegation: Veto
those same public officials to compensate for their inability
Players, Central Bank Independence, and the Credibility of Mon-
to offer a secure environment to private investors. If the etary Policy," American Political Science Review 97:3 (2003),
former explanation is correct, scaling up aid for infrastruc- 407^23.
Levine, Ross, and David Renelt, "A Sensitivity Analysis of Cross-Country
ture or other public investment in countries with weak Growth Equations," American Economic Review 82:4 (1992), 942-
governance will have little impact on economic growth. 963.
Advocates of dramatic increases in aid for infrastructure Mauro, Paulo, "Corruption and the Composition of Government Expen-
diture," Journal of Public Economics 69:2 (1998), 263-279.
often acknowledge that scaled-up funding should be accom-
Mulligan, Casey, Ricard Gil, and Xavier Sala-i-Martin, "Do Democracies
panied by anticorruption programs, to ensure the resources Have Different Public Policies than Non-democracies?" Journal of
are well spent (Commission for Africa, 2005). However, Economic Perspectives 18:1 (2004), 51-74.
Noll, Roger, and Morns Fiorina, Voters, Legislators and Bureaucrats: A
policies to fight corruption similarly may be ineffective in Rational Choice Perspective on the Growth of Bureaucracy," Jour-
counteracting the more deeply seated incentive structure nal of Public Economics 9:3 (1978), 239-254.
that leads to distortions in public investment. North, Douglass, and Barry Weingast, "Constitutions and Commitments:
The Evolution of Institutions Governing Public Choice in 17th
Century England," Journal of Economic History 49:4 (1989),
17
Acemoglu, Johnson, and Robinson (2002) and North and Weingast 803-832.
(1989) emphasize the importance of political checks and balances for the Persson, Torsten, and Guido Tabellini, Political Economics: Explaining
security of property rights. Keefer and Stasavage (2003) demonstrate that Public Policy (Cambridge, MA: MIT Press, 2000).
political checks and balances improve the credibility of government Pritchett, Lant, "Mind Your Ps and Qs: The Cost of Public Investment Is
monetary policy, thereby reducing inflation. Not the Value of Public Capital," World Bank policy research
working paper no. 1660 (1996).
REFERENCES Robinson, James, and Ragnar Torvik, "White Elephants," Journal of
Public Economics 89:2-3 (2005), 197-210.
Acemoglu, Daron, "Politics and Economics in Weak and Strong States," Sachs, Jeffrey, with John W. McArthur et al., Ending Afnca s Poverty
Journal of Monetary Economics 52:7 (2005), 1199-1226. Trap," Brookings Papers on Economic Activity 2004:1 (2004),
Acemoglu, Daron, Simon Johnson, and James Robinson, "The Colonial 117-240.
Origins of Comparative Development: An Empirical Investiga- Summers, Robert, and Alan Heston, "The Penn World Table (Mark V): An
tion," American Economic Review 91:5 (2001), 1369-1401. Extended Set of International Comparisons, 1950-88," Quarterly
"Reversal of Fortune: Geography and Institutions in the Making of Journal of Economics 106:2 (1991), 327-336.
the Modern World Income Distribution," Quarterly Journal of Tanzi, Vito, and Hamid Davoodi, "Corruption, Public Investment and
Economics 117:4 (2002), 1231-1294. Growth," IMF working paper WP/97/139 (1997).

This content downloaded from

[Link] on Thu, 31 Jul 2025 [Link] UTC
All use subject to [Link]