ECONOMETRIC POLICY EVALUATION: A CRITIQUE

Robert E. Lucas Jr.

1. Introduction

The fact that nominal prices and wages tend to rise more rapidly at the peak of
the business cycle than they do in the trough has been well recognized from the time
when the cycle was first perceived as a distinct phenomenon. The inference that
permanent inflation will therefore induce a permanent economic high is no doubt
equally ancient, yet it is only recently that this notion has undergone the mysterious
transformation from obvious fallacy to cornerstone of the theory of economic policy.
This transformation did not arise from new developments in economic theory.
On the contrary, as soon as Phelps and others made the first serious attempts to
rationalize the apparent trade-off in modern theoretical terms, the zero-degree
homogeneity of demand and supply functions was re-discovered in this new context (as
Friedman predicted it would be) and re-named the "natural rate hypothesis 1". It arose,
instead, from the younger tradition of the econometric forecasting models, and from the
commitment on the part of a large fraction of economists to the use of these models for
quantitative policy evaluation. These models have implied the existence of long-run
unemployment-inflation trade-offs ever since the "wage-price sectors" were first
incorporated and they promise to do so in the future although the "terms" of the trade-
off continue to shift2.
This clear-cut conflict between two rightly respected traditions – theoretical and
econometric - caught those of us who viewed the two as harmoniously complementary
quite by surprise. At first, it seemed that rite conflict might be resolved by somewhat
fancier econometric footwork. On rite theoretical level, one hears talk of a
"disequilibrium dynamics" which will somehow make money illusion respectable while
dp e
going beyond the sterility of =k ( p−p ) . Without underestimating the ingenuity of
dt
either econometricians or theorists, it seems to me appropriate to entertain the
1
See Phelps et at. [31], Phelps' earlier [30] and Friedman [13].
2
The earliest wage-price sector embodying the "trade-off" is (as far as I know) in the 19.55 version of the
Klein-Goldberger model [19]. It has persisted, with minimal conceptual change, into all current
generation forecasting models. The subsequent shift of the "trade-off" relationship to center stage in
policy discussions appears due primarily to Phillips [32] and Samuelson and Solow [33].
possibility that reconciliation along both of these lines will fail, and that one of these
traditions is fundamentally in error.
The thesis of this essay is that it is rite econometric tradition, or more precisely,
the "theory of economic policy" based on this tradition, which is in need of major
revision. More particularly, 1 shall argue that the features which lead to success in
short-term forecasting are unrelated to quantitative policy evaluation, that the major
econometric models are (well) designed to perform the former task only, and that
simulations using these models can, in principle, provide no useful information as to the
actual consequences of alternative economic policies. These contentions will be based
not on deviations between estimated and "true" structure prior to a policy change but on
the deviations between the prior "true" structure and the "true" structure prevailing
afterwards.
Before turning to details, I should like to advance two disclaimers. First, as is
true with any technically difficult and novel area of science, econometric model
building is subject to a great deal of ill-informed and casual criticism. Thus models are
condemned as being "too big" (with equal insight, I suppose one could fault smaller
models for being "too little"), too messy, too simplistic (that is, not messy enough), and,
the ultimate blow, inferior to "naive" models. Surely the increasing sophistication of the
"naive" alternatives to the major forecasting models is the highest of tributes to the
remarkable success of the latter. I hope I can succeed in disassociating the criticism
which follows from any denial of the very important advances in forecasting ability
recorded by the econometric models, and of the promise they offer for advancement of
comparable importance in the future.
One may well define a critique as a paper which does not fully engage the vanity
of its author. In this spirit, let me offer a second disclaimer. There is little in this essay
which is not implicit (and perhaps to more discerning readers, explicit) in Friedman
[11], Muth [29] and, still earlier, in Knight [21]. For that matter, the criticisms I shall
raise against currently popular applications of econometric theory have, for the most
part, been anticipated by the major original contributors to that theory 3. Nevertheless,
the case for sustained inflation, based entirely on econometric simulations, is attended
now with a seriousness it has not commanded for many decades. It may, therefore, be

3
See in particular Marschak's discussion in [25] (helpfully recalled to me by T. D. Wallace) and
Tinbergen’s in [36], especially his discussion of "qualitative policy" in ch. 5, pp. 149-185.
worthwhile to attempt to trace this case back to its foundation, and then to examine
again file scientific basis of this foundation itself.

2. Theory of Economic Policy

Virtually all quantitative macro-economic policy discussions today are

conducted within a theoretical framework which I shall call "the theory of economic
policy", (following Tinbergen [35]). The essentials of this framework are so widely
known and subscribed to that it may be superfluous to devote space to their review. On
the other hand, since the main theme of this paper is the inadequacy of this framework,
it is probably best to have an explicit version before us. One describes the economy in a
time period t by a vector Y t of state variables, a vector X t of exogenous forcing
variables, and a vector ϵ t of independent (through time), and identically distributed
random shocks. The motion of the economy is determined by a difference equation the
distribution of ϵ t , and a description of the temporal behaviour of the forcing variables, x t
.

Y t +1=f ( y t , x t , ϵ t ),

The function f is taken to be fixed but not directly known; the task of empiricists is then
to estimate f . For practical purposes, one usually thinks of estimating the values of a
fixed parameter vector Ɵ , with and f being specified in advance. Mathematically, the
sequence { X T ) of forcing vectors is regarded as being "arbitrary" (that is, it is not
characterized stochastically).

f ( y , x , ϵ)=F ¿ϵ¿,

Since the past x t Values are observed, this causes no difficulty in estimating Ɵ , and in
fact simplifies the theoretical estimation problem slightly. For forecasting, one is
obliged to insert forecasted x t values into F .
With knowledge of the function F and Ɵ , policy evaluation is a straightforward
matter. A policy is viewed as a specification of present and future values of some
components of { x t }. With the other components somehow specified, the stochastic
behaviour of { y t , x t , ϵ t } from the present on is specified, and functionals defined on this
sequence are well-defined random variables, whose moments may be calculated
theoretically or obtained by numerical simulation. Sometimes, for example, one wishes
to examine the mean value of a hypothetical "social objective function", such as under
alternative policies.

∑ β t u ( y t , xt , ϵt )
t =0

More usually, one is interested in the "operating characteristics" of the system under
alternative policies. Thus, in this standard context, a "long-run Phillips curve" is simply
a plot of average inflation-unemployment pairs under a range of hypothetical policies 4.
Since one cannot treat Ɵ as known in practice, the actual problem policy
evaluation is somewhat more complicated. The fact that Ɵ is estimate from past sample
values affects the above moment calculations for small samples; it also makes policies
which promise to sharpen estimates of Ɵ relatively more attractive. These
considerations complicate without, I think, essentially altering the theory of economic
policy as sketched above.
Two features of this theoretical framework deserve special comment. The first is
file uneasy relationship between this theory of economic policy and traditional
economic theory. The components of the vector-valued function F are behavioural
relationships – demand functions; the role of theory may thus be viewed as suggesting
forms for F , or in Samuelson's terms, distributing zeros throughout the Jacobian of F .
This role for theory is decidedly secondary: microeconomics shows surprising power to
rationalize individual econometric relationships in a variety of ways. More significantly,
this micro-economic role for theory abdicates the task of describing the aggregate
behaviour of the system entirely to the econometrician. Theorists suggest forms for
consumption, investment, price and wage setting functions separately; these
suggestions, if useful, influence individual components of F . The aggregate behaviour
of the system then is whatever it is5. Surely this point of view (though I doubt if many
would now endorse it in so bald a form) accounts for the demise of traditional "business
4
See, for example, de Menil and Enzler [6], Hitsch [16] and Hymans [17].
5
The ill-fated Brookings model project was probably the ultimate expression of this view.
cycle theory" and the widespread acceptance of a Phillips "trade-off" in the absence of
any aggregative theoretical model embodying such a relationship. Secondly, one must
emphasize the intimate link between short-term forecasting and long-term simulations
within this standard framework. The variance of short-term forecasts tends to zero with
the variance of ϵ t ; as the latter becomes small, so also does the variance of estimated
behavior of { y t } conditional on hypothetical policies { x t }. Thus forecasting accuracy in
the short-run implies reliability of long-term policy evaluation.

3. Adaptive Forecasting

There are many signs that practicing econometricians pay little more than lip-
service to the theory outlined in the preceding section. The most striking is the
indifference of econometric forecasters to data series prior to 1947. Within the theory of
economic policy, more observations always sharpen parameter estimates and forecasts,
and observations on "extreme" x t values particularly so; yet even the readily available
annual series from 1929-1946 are rarely used as a check on the post-war fits,
A second sign is the frequent and frequently important refitting of econometric
relationships. The revisions of the wage-price sector now in progress are a good
example6. The continuously improving precision of the estimates of Ɵ within the fixed
structure F , predicted by the theory, does not seem to be occurring in practice. Finally,
and most suggestively, is the practice of using patterns in recent residuals to revise
intercept estimates for forecasting purposes. For example, if a "run" of positive residuals
(predicted less actual) arises in an equation in recent periods, one revises the estimated
intercept downward by their average amount. This practice accounts, for example, for
the superiority of the actual Wharton forecasts as compared to forecasts based on the
published version of the model7.
It should be emphasized that recounting these discrepancies between theory and
practice is not to be taken as criticism of econometric forecasters. Certainly if new
observations are better accounted for by new or modified equations, it would be foolish
to continue to forecast using the old relationships. The point is simply that,
econometrics textbooks notwithstanding, current forecasting practice is not conducted

6
See, for example, Gordon [14l].
7
A good account of this and other aspects of forecasting in theory and practice is provided by Klein [20].
A fuller treatment is available in Evans and Klein [9].
within the framework of the theory of economic policy, and the unquestioned success of
the forecasters should not be construed as evidence for the soundness or reliability of
the structure proposed in that theory.
An alternative structure to that underlying the theory of economic policy has
recently been proposed (in [3] and [5]) by Cooley and Prescott. The structure is of
interest in the present context, since optimal forecasting within it shares many features
with current forecasting practice as just described: Instead of treating the parameter
vector Ɵ as fixed, Cooley and Prescott view it as a random variable following the
random walk where {nt } is a sequence of independent, identically distributed random
variables.

Ɵt +1=Ɵt +nt +1

Maximum likelihood forecasting under this alternative framework ("adaptive

regression") resembles "exponential smoothing" on the observations, with observations
in the distant past receiving a small "weight" - very much as in usual econometric
practice; similarly, recent forecast errors are used to adjust the estimates. Using both
artificial data and economic time series, Cooley and Prescott have shown (in [41]) that
adaptive methods have good short-term foresting properties when compared to even
relatively sophisticated versions of the "fixed Ɵ " regression model. As Klein and others
have remarked, this advantage is shared by actual large-model forecasts (that is, model
forecasts modified by the forecaster's judgment) over mechanical forecasts using the
published versions of the model8.
Cooley and Prescott have proposed adaptive regression as a normative
forecasting method. I am using it here in a positive sense: as an idealized "model" of the
behaviour of large-model forecasters. If the model is, as I believe, roughly accurate, it
serves to reconcile the assertion that long-term policy evaluations based on econometric
models are meaningless with the acknowledgment that the forecast accuracy of these
models is good and likely to become even better. Under the adaptive structure, a small
standard error of short-term forecasts is consistent with infinite variance of the long-
term operating characteristics of the system.

8
See Klein [201].
4. Theoretical Considerations: General

To this point, I have argued simply that the standard, stable-parameter view of
econometric theory and quantitative policy evaluation appears not to match several
important characteristics of econometric practice, while an alternative general structure,
embodying stochastic parameter drift, matches these characteristics very closely. This
argument is, if accepted, sufficient to establish that the "long-run" implications of
current forecasting models are without content, and that the short-term forecasting
ability of these models provides no evidence of the accuracy to be expected from
simulations of hypothetical policy rules.
These points are, I think, important, but their implications for the future are
unclear. After all, the major econometric models are still in their first, highly successful,
decade. No one, surely, expected the initial parameterizations of these models to stand
forever, even under the most optimistic view of the stability of the unknown, underlying
structure. Perhaps the adaptive character of this early stage of macro-economic
forecasting is merely the initial groping for the true structure which, however ignored in
statistical theory, all practitioners knew to be necessary. If so, the arguments of this
paper are transitory debating points, obsolete soon after they are written down.
Personally, I would not be sorry if these were the case, but I do not believe it is. I shall
try to explain why, beginning with generalities, and then, in the following section,
introducing examples.
In section 2, we discussed an economy characterized by

Y t +1=F ¿¿ X t ,Ɵ, ϵ t )

The function F and parameter vector Ɵ are derived from decision rules (demand
and supply functions) of agents in the economy, and these decisions are, theoretically,
optimal given the situation in which each agent is placed. There is, as remarked above,
no presumption that ¿) will be easy to discover, but it is the central assumption of the
theory of economic policy that once they are (approximately) known, they will remain
stable under arbitrary changes in the behavior of the forcing sequence { X t }.
For example, suppose a reliable model ¿) is in hand, and one wishes to use it to
assess the consequences of alternative monetary and fiscal policy rules (choices of
X 0 , X 1 X 2,..., where t=Ɵ is "now"). According to the theory of economic policy, one
then simulates the system under alternative policies (theoretically or numerically) and
compares outcomes by some criterion. For such comparisons to have any meaning, it is
essential that the structure ( F , Ɵ ) not vary systematically with the choice of { X t }.
Everything we know about dynamic economic theory indicates that this
presumption is unjustified. First, the individual decision problem: "find an optimal
decision rule when certain parameters (future prices, say) follow 'arbitrary' paths" is
simply not well formulated. Only trivial problems in which agents can safely ignore the
future can be formulated under such a vague description of market constraints. Even to
obtain the decision rules underlying ( F , Ɵ ) then, we have to attribute to individuals
some view of the behavior of the future values of variables of concern to them. This
view, in conjunction with other factors, determines their optimum decision rules. To
assume stability of (F ,Ɵ) under alternative policy rules is thus to assume that agents'
views about the behavior of shocks to the system are invariant under changes in the true
behavior of these shocks. Without this extreme assumption, the kinds of policy
simulations called for by the theory of economic policy are meaningless.
It is likely that the "drift" in Ɵ which the adaptive models describe stochastically
reflects, in part, the adaptation of the decision rules of agents to the changing character
of the series they are trying to forecast9. Since this adaptation will be in most (though
not all) cases slow, one is not surprised that adaptive methods can improve the short-
term forecasting abilities of the econometric models. For longer term forecasting and
policy simulations, however, ignoring the systematic sources of drift will lead to large,
unpredictable errors.

5. Theoretical Considerations: Examples

If these general theoretical observations on the likelihood of systematic

"parametric drift" in the face of variations in the structure of shocks are correct, it
should be possible to confirm them by examination of the specific decision problems
underlying the major components of aggregative models. I shall discuss in turn
consumption, investment, and the wage-price sector, or Phillips curve. In each case, the
"right hand variables" will, for simplicity, be taken as "exogenous" (as components of {

9
This is not to suggest that all parameter drift is due to this source. For example, shifts in production
functions due to technological change ate probably well described by a random walk scheme.
X t }). The thought-experiments matching this assumption, and the adaptations necessary
for simultaneous equations, are too well known to require comment.

5.1 Consumption

The easiest example to discuss with confidence is the aggregate consumption function
since, due to Friedman [11], Muth [28] and Modigliani, Brumberg and Ando [2], [27], it
has both a sound theoretical rationale and an unusually high degree of empirical
success. Adopting Friedman's formulation, permanent consumption is proportional to
permanent income (an estimate of a discounted future income stream), actual
consumption is and actual, current income is

(1) C pt =k y pt

(2) C t=C pt +U t

(3) Y t =Y pt +V t

The variables U t ,V t are independent temporally and of each other and of Y pt .

An empirical "short-run" marginal propensity to consume is the sample moment
corresponding to Cov (Ct ,Y t )/Var (Y t ), or

Var (Y pt )
K
Var ( Y pt )+ Var (V t )

Now as long as these moments are viewed as subjective parameters in the heads
of consumers, this model lacks content. Friedman, however, viewed them as true
moments, known to consumers, the logical step which led to the cross-sectional tests
which provided the most striking confirmation of his permanent income hypothesis10.
This central equating of a true probability distribution and the subjective
distribution on which decisions are based was termed rational expectations by Muth,

10
Of course, the hypothesis continues to be tested as new data sources become available, and anomalies
continue to arise. (For a recent example, see Mayer [26]). Thus one may expect that, as with most
"confirmed" hypotheses, it will someday be subsumed in some more general formulation.
who developed its implications more generally (in [29]). In particular, in [28], Muth
found the stochastic behavior of income over time under which Friedman's
identification of permanent income as an exponentially weighted sum of current and
lagged observations on actual income was consistent with optimal forecasting on the
part of agents11.
To review Muth's results, we begin by recalling that permanent income is that
constant flow Y pt which has the same value, with the subjective discount factor β , as the
forecasted actual income stream:

∞
(4) Y pt = (1 - β) ∑ β E( y t + i/I t )
i

j=0

where each expectation is conditioned on information I t available at t . Now let actual

income y t be a sum of three terms where V t is transitory income, a is a constant, and W t
is a sum of independent increments, each with zero mean and constant variance.

(5) y t = a + w t + v t,

Muth showed that the minimum variance estimator of y t +1 for all i=1 , 2.... is
∞
(l− λ) ∑ λ j y t− j where λ depends in a known way on the relative variances of w t and v t .
t=0

∞
(6) C t= K (1−β ) Y t + Kβ ( 1−λ ) ∑ λ Y t− j+ ut
j

j=0

(This formula differs slightly from Muth's because Muth implicitly assumed that C t
was determined prior to realizing Y t . The difference is not important in the sequel.)
Now let us imagine a consumer of this type, with a current income generated by
an "experimenter" according to the pattern described by Muth (so that the premises of
the theory of economic policy are correct for a single equation consumption function).
An econometrician observing this consumer over many periods will have good success
11
In [12], Friedman proposes an alternative view to Muth's, namely that the weight used in averaging past
incomes ( λ , below) is the same as the discount factor used in averaging future incomes ( β , below). It is
Muth's theory, rather than Friedman's of [12], which is consistent with the cross-section tests based on
relative variances mentioned above.
describing him by (6) whether he arrives at this equation by the Friedman-Muth
reasoning, or simply hits on it by trial-and-error. Next consider policies taking the form
of a sequence of supplements {Y t } to this consumer's income from time T on. Whether
{Y t } is specified deterministically or by some stochastic law, whether it is announced in
advance to the consumer or not, the theory of economic policy prescribes the same
method for evaluating its consequences: add X t to the forecasts of Y t for each t >T ,
insert into (6), and obtain the new forecasts of C t .
If the consumer knows of the policy change in advance, it is clear that this
standard method gives incorrect forecasts. For example, suppose the policy consists of a
constant increase, X t = x , in income over the entire future. From (4), this leads to an
increase in consumption of k x . The forecast based on (6), however, is of an effect in
period t of

1−t
(Δc)t = k x {(1−β )+ β ( 1−λ ) ∑ λ j }
j =0

Since this effect tends to the correct forecast, k x , as t tends to infinity, one might
conjecture that the difficulty vanishes in the "long run". To see that this conjecture is
1
false, consider an exponentially growing supplement X t = x at , 1 ¿ a< . The true effect
β
in t-T is, from (1) and (4),

t
(1− β)a
(Δc)t = k x .
1−aβ

The effect as forecast by (6) is

1−t
λj t
(Δc)t = k x {(1−β ) + β ( 1−λ ) ∑ }a
j=0 a

Neither effect tends to zero, as t tends to infinity; the ratio (forecast over actual) tends to
which may lie on either side of unity.
 aβ ( 1− λ )
(1−aβ) +{1+ }.
( 1− β ) ( a−λ )

More interesting divergences between forecasts and reality emerge when the
policy is stochastic, but with characteristics known in advance. For example, let { X t } be
a sequence of independent random variables, with zero mean and constant variance,
distributed independently of ut , v tand w t . This policy amounts to an increase in the
variance of transitory income, lowering the weight X in a manner given by the Muth
formula. Average consumption, in fact and as forecast by (6), is not affected, but the
variance of consumption is. The correct estimate of this variance effect requires revision
of the weight λ ; evidently the standard, fixed-parameter prediction based on (6) will
again yield the wrong answer, and the error will not tend to vanish for larget .
The list of deterministic and stochastic policy changes, and their combination is
inexhaustible but one need not proceed further to establish file point: for any policy
change which is understood in advance, extrapolation or simulation based on (6) yields
an incorrect forecast, and what is more, a correctible incorrect forecast. What of changes
in policy which are not understood in advance? As Fisher observes, “the notion that one
cannot fool all of the people all of the time [need not] imply that one cannot fool all the
people even some of the time12”.
The observation is, if obvious, true enough; but it provides no support whatever
for the standard forecasting method of extrapolating on the basis of (6). Our knowledge
of consumption behaviour is summarized in (1)-(4). For certain policy changes we can,
with some confidence, guess at the permanent income recalculations consumers will go
through and hope to predict their consumption responses with some accuracy. For other
types of policies, particularly those involving deliberate "fooling" of consumers, it will
not be at all clear how to apply (1)-(4), and hence impossible to forecast. Obviously, in
such cases, there is no reason to imagine that forecasting with (6) will be accurate
either.

5.2 Taxation and Investment Demand

v 2
12
Let σ 2be the variance of v t and σ Δw be the variance of the increments of w t , then the relationship is

√
2
1 σ Δw σ ∆ w 2
1 σ Δw
λ=1+ 1+ .
2 σ 2v σ v 4 σ 2v
 In [15], Hall and Jorgenson provided quantitative estimates of the consequences,
current and lagged, of various tax policies on the demand for producers' durable
equipment. Their work is an example of the current state of the art of conditional
forecasting at its best. The general method is to use econometric estimates of a
Jorgensonian investment function, which captures all of the relevant tax structure in a
single implicit rental price variable, to simulate the effects of alternative tax policies.
An implicit assumption in this work is that any tax change is regarded as a
permanent, once-and-for-all change. Insofar as this assumption is false over the sample
period, the econometric estimates are subject to bias 13. More important for this
discussion, the conditional forecasts will be valid only for tax changes believed to be
permanent by taxpaying corporations.
For many issues in public finance, this obvious qualification would properly be
regarded as a mere technicality. For Keynesian counter-cyclical policy, however, it is
the very heart of the issue. The whole point, after all, of the investment tax credit is that
it be viewed as temporary, so that it can serve as an inducement to firms to reschedule
their investment projects. It should be clear that the forecasting methods used by Hall
and Jorgenson (and, of course, by other econometricians) cannot be expected to yield
even order-of-magnitude estimates of the effects of explicitly temporary tax
adjustments.
To pursue this issue further, it will be useful to begin with an explicit version of
the standard accelerator model of investment behaviour. We imagine a constant returns
industry in which each firm has a constant output-capital ratio λ . Using a common
notation for variables at both the firm and industry level, let K t denote capital at the
beginning of year t. Output during t is λK t . Investment during the year, it, affects next
period's capital according to where δ is a constant physical-rate of depreciation. Output
is sold on a perfect market at a price Pt ; investment goods are purchased at a constant
price of unity. Profits (sales less depreciation) are taxed at the rate Ɵt ; there is an
investment tax credit at tire rate Ψ t .

K t + 1 = i t +(1−δ)K t ,
The firm is interested in maximizing the expected present value of receipts net of
taxes, discounted at the constant cost of capital r . In the absence (assumed here) of

13
[10], p. 113.
adjustment costs, this involves equating the current cost of an additional unit of
investment to the expected discounted net return. Assuming that the current tax bill is
always large enough to cover the credit, the current cost of acquiring an additional unit
of capital is (1−Ψ t), independent of the volume of investment goods purchased. Each
unit of investment yields k units of output, to be sold next period at the (unknown) price
Pt +1. Offsetting this profit is a tax Ɵt +1[ λp t+1 −δ ]. In addition, (1−δ ) units of the
investment good remain for use after period t+ 1; with perfect capital goods markets,
these units are valued at (1−Ψ t+ 1),. Thus letting Et (.) denote an expectation conditional
on information up to period t, the expected discounted return per unit of investment in t
is

1
E ¿ (1 - Ɵt +1 ¿ +δ +(1−δ)(1−Ψ t +1)¿
1+ r t

Since a change in next period's tax rate Ɵt +1 which is not anticipated in t is a "pure
profit tax", Ɵt +1and Pt +1will be uncorrelated. Hence, equating costs and returns, one
equilibrium condition for the industry is
1
(7) E ¿ (1 - Ɵt +1 ¿ δ +(1−δ)(1−Ψ t +1)¿+
1+ r t

1
1−Ψ t + { λ E t (P¿¿ t +1)¿ ¿
1+ r

A second equilibrium condition is obtained from the assumption that the product market
is cleared each period. Let industry demand be given by a linear function, with a
stochastically shifting intercept At and a constant slopeb , so that quantity demanded
next period will be At +1 +bp t+1 . Quantity supplied will be λ times next period's capital.
Then a second equilibrium condition is

λ [i t +(1−δ )K t ]¿ A t +1+ bpt +1

Taking mean values of both sides,

(8) λ [i t +(1−δ)K t ]¿ Et (at +1)+ bEt ( p t+1 )

Since our interest is in the industry investment function, we eliminate Et ( pt +1)between
(7) and (8) to obtain:

r ( 1+ r ) Ψ t−(1−δ)E t ( Ψ t +1 )
λ¿ [ +δ ] + b2 [ ]
1−Et ( Ɵt +1 ) λ 1−E t ( Ɵ t +1 )

Equation (9) gives the industry's "desired" stock of capital, i t + ( 1−δ ) K t , as a

function of the expected future state of demand and the current and expected future tax
structure, as well as of the cost of capital r, taken in this illustration to be constant. The
second and third terms on the right are the product of the slope of the demand curve for
capital, −bλ−2, and the familiar Jorgensonian implicit rental price; the second term
includes "interest" and depreciation costs net of taxes; the third includes the expected
capital gain (or loss) due to changes in the investment tax credit rate.
In most empirical investment studies, firms are assumed to move gradually from
K t to the desired stock given by (9), due to costs of adjustment, delivery lags, and the
like. We assume here, purely for convenience, that file full adjustment occurs in a single
period.
Equation (9) is operationally at the same level as equations (1) and (4) of the
preceding section: it relates current behaviour to unobserved expectations of future
variables. To move to a testable hypothesis, one must specify the time series behaviour
of a t ,Ɵt , and Ψ t (as was done for income in consumption theory), obtain the optimal
forecasting rule, and obtain the analogue to the consumption function (6). Let us
imagine that this has been accomplished, and estimates of the parameters k and b have
been obtained. How would one use these estimates to evaluate the consequences of a
particular investment tax credit policy?
The method used by Hall and Jorgenson is to treat the credit as a permanent or
once-and-for-all change, or implicitly to set Et ( Ψ t +1 ) equal to Ψ t +1. Holding Ɵt constant
at Ɵ , the effect'ofa change in the credit from 0 to Ψ (say) would be the same as a
permanent lowering of the price of investment goods to 1−Ψ or, from (9), an increase
b r +8
in the desired capital stock of . . If the credit is in fact believed by corporations
λ 1−Ɵ
2

to be permanent, this forecast will be correct; otherwise it will not be.

 To consider alternatives, imagine a stochastic tax credit policy which switches
from 0 to a fixed number Ψ in a Markovian fashion, with transitions given by
Pr {Ψ t +1=Ψ ∨Ψ t =0 }=q and Pr {Ψ t +1=Ψ ∨Ψ t =Ψ }=p .14 Then if expectations on next
period's tax credit are formed rationally, conditional on the presence or absence of the
credit in the current period, we have

Et ( Ψ t +1 )=
{ qΨ if Ψ t =0
pΨ if Ψ t=Ψ

The third term on the right of (9) is then

bΨ
2
[−q(1−δ)] if Ψ t =0
λ (1−Ɵ)

bΨ
2
[1+ r− p(1−δ)] if Ψ t =Ψ
λ (1−Ɵ)

The difference between these terms is given by the expression

bΨ
2
[1+ r+ ( q−p ) (1−δ)]
λ (1−Ɵ)

The expression (10) gives the increment to desired capital stock (and, with
immediate adjustment, to current investment) when the tax credit is switched from zero
to Ψ in an economy where the credit operates, and is known to operate, in the stochastic
fashion described above. It does not measure the effect of a switch in policy from a no-
credit regime to the stochastic regime used here. (The difference arises because even
when the credit is set at zero in the stochastic regime, the possibility of capital loss, due
to the introduction of the credit in the future, increases the implicit rental on capital,
relative to the situation in which the credit is expected to remain at zero forever.) By
examining extreme values of p and q one can get a good idea of the quantitative
importance of expectations in measuring the effect of the credit. At one extreme,

14
In particular, the low estimates of “a ” (see [15], Table 2, p. 400), which should equal capital's share in
value added, are probably due to a sizeable transitory component in a variable which is treated
theoretically as though it were subject to permanent changes only.
consider the case where the credit is expected almost never to be offered ( q near 0 ), but
once offered, it is permanent ( p near 1). The effect of a switch from 0 to Ψ , is, in this
case, approximately using (10).

bΨ
2
[r+ δ]
λ (1−Ɵ)

This is the situation assumed, implicitly, by Hall and Jorgenson. At the other extreme,
consider the case of a frequently imposed but always transitory credit (q near 1, p near
0). Applying (10), the effect of a switch in this case is approximately

bΨ
2
[2+ r +δ ]
λ (1−Ɵ)

The ratio of effects is then (2+r −δ)/(r +δ). With r .14 and δ =.15, this ratio is about
715. We are not, then, discussing a quantitatively minor issue.
For a more realistic estimate, consider a credit which remains "off" for an
average period of 5 years and when "switched on" remains for an average of one year.
1
These assumptions correspond to setting p ≅ 0 andq . The ratio of the effect (from
2
(10)), under these assumptions versus those used by Hall and Jorgenson is now

[ 1
]
1+ r + ( 1−δ ) / ( r −δ ). With r =.14 and δ=. 15, this ratio is approximately 4.5. This
5
ratio would probably be somewhat smaller under a more satisfactory lag structure 16, but
even taking this into account, it appears likely that the potential stimulus of the
investment tax credit may well be several times greater titan the Hall-Jorgenson
estimates would indicate17.

15
A tax credit designed for stabilization would, of course, need to respond to projected movements ha the
shift variable a t. In this case, the transition probabilities p and q would vary with indicators (say current
and lagged a t values) of future economic activity. Since my aim here is only to get an idea of the
quantitative importance of a correct treatment of expectations, I will not pursue this design problem
further.
16
The cost of capital of .14 and the depreciation rate of .15 (for manufacturing equipment) are annual
rates from [15]. Since the ratio (2+r −δ)/(r +δ) is not time-unit free, the assumption that all
movement to" ward the new desired stock of capital takes place in one year is crucial at this point: by
defining a period as shorter than one year this ratio will increase, and conversely for a longer period.
17
For the reason given in note 16.
 As was the case in the discussion of consumption behaviour, estimation of a
policy effect along the above lines presupposes a policy generated by a fixed, relatively
simple rule, known by forecasters (ourselves) and by the agents subject to the policy (an
assumption which is not only convenient analytically but consistent with Article 1,
Section 7 of rite U.S. Constitution). To go beyond the kind of order-of-magnitude
calculations used here to an accurate assessment of the effects of the 1962 credit studied
by Hall and Jorgenson, one would have to infer the implicit rule which generated (or
was thought by corporations to generate) that policy, a task made difficult, or perhaps
impossible, by the novelty of the policy at the time it was introduced. Similarly, there is
no reason to hope that we can accurately forecast the effects of future ad hoc tax
policies on investment behaviour. On the other hand, there is every reason to believe
that good quantitative assessments of counter-cyclical fiscal rules, which are built into
the tax structure in a stable and well-understood way, can be obtained.

5.3 Phillips Curves

A third example is suggested by the recent contr0versyover the Phelps-

Friedman hypothesis that permanent changes in the inflation rate will not alter the
average rate of unemployment. Most of the major econometric models have been used
in simulation experiments to test this proposition; the results are uniformly negative.
Since expectations are involved in an essential way in labour and product market supply
behaviour, one would presume, on the basis of the considerations raised ill section 4,
that these tests are beside the point18. This presumption is correct, as the following
example illustrates.
It will be helpful to utilize a simple, parametric model which captures the main
features of the expectational view of aggregate supply - rational agents, cleared markets,
incomplete information19. We imagine suppliers of goods to be distributed over N
distinct markets , i=1 ..... N . To avoid index number problems, suppose that the same
(except for location) good is traded in each market, and let y ¿ be the log of quantity
supplied in marketi in periodt . Assutne, further, that the supply y ¿ is composed of two

18
It should be noted that this conclusion reinforces the qualitative conclusion reached by Hall and
Jorgenson [15], p. 413.
19
Sargent [34] and I [23] have developed this conclusion earlier in similar contexts.
factors where y ¿p denotes normal or permanent supply, and y c¿ cyclical or transitory
supply (both, again, in logs).

y ¿ = y ¿p+ y c¿

We take y ¿pto be unresponsive to all but permanent relative price changes or, since the
latter have been defined away by assuming a single good, simply unresponsive to price
changes. Transitory supply y c¿ varies with perceived changes in the relative price of
goods in i:

c e
y ¿ = β ( p¿ − p¿),

where p¿ is the log of the actual price in i at t , and pe¿ is the log of tim general
(geometric average) price level in the economy as a whole, as perceived in market i.
Prices will vary from market to market for each t, due to file usual sources of fluctuation
hi relative demands. They will also fluctuate over time, due to movements in aggregate
demand. We shall not explore the sources of these price movements (although this is
easy enough to do) but simply postulate that the actual price in i at t consists of two
components:

p¿ = pt + z ¿

Sellers observe the actual price p¿; the two components cannot be separately observed.
The component pt varies with time, but is common to all markets. Based on information
obtained prior to t (call it I t−1) traders in all markets take pt to be a normally distributed
random variable, with mean pt (reflecting this past information) and variance σ 2. The
component z ¿reflects relative price variation across markets and time: z ¿is normally
distributed, independent of pt and Z j ¿unlessi= j, s=t ¿ ,with mean0 and varianceτ 2 .
s

The actual general price level at t is the average over markets of individual prices,
 N N
1 1
N ∑ P ¿= Pt + ∑Z
N i=t ¿
i=t

We take the number of markets N to be large, so that the second term can be
neglected, and pt is the general price level. To form the supply decision, suppliers
estimate pt ; assume that this estimate pe¿ is the mean of the true conditional distribution
of pt . The latter is calculated using the observation that p¿is the sum of two independent
normal variates, one with mean 0 and variance τ 2 .; one with mean pt and variance σ 2.

P¿ =¿ E { p¿ ∨, I t−1 } ¿(1−Ɵ) p¿+ Ɵ p t

e

2
τ
It follows that where Ɵ = 2 2
Ɵ +τ

Based on this unbiased but generally inaccurate estimate of the current general
level of prices, suppliers in i follow

y ¿ =¿ β [ p ¿−( ( 1−Ɵ ) p ¿ +Ɵ p t 0 ) ]=Ɵβ ¿]

c

Now averaging over markets, and invoking the law of large numbers again, we
have the cyclical component of aggregate supply:

e
y ¿ =Ɵβ ¿ )

Re-introducing the permanent components,

y t =Ɵβ ¿) + y pt

Though simple, (11) captures the main features of the expectational or "natural
rate" view of aggregate supply. The supply of goods is viewed as following a trend path
y pt which is not depeudent on nominal price movements. Deviations from this path are
induced whenever the nominal price deviates from the level which was expected to
prevail on the basis of past information. These deviations occur because agents are
obliged to infer current general price movements on the basis of incomplete
information.
It is worth speculating as to the sort of empirical performance one would expect
from (11). Ill doing so, we ignore the trend component y pt , concentraring on the
determinants of pt , β and Ɵ . The parameter β reflects intertemporal substitution
possibilities in supply: technological factors such as storability of production, and tastes
for substituting labor supplied today for supply tomorrow. One would expect β to be
reasonably stable over time and across economies at a similar level of development. The
2
τ
parameter Ɵ is the ratio 2 2 .τ 2 reflects the variability of relative prices within
σ +τ
the economy; there is no reason to expect it to vary systematically with demand policy,
2
σ is the variance of the general price level about its expected level; it will obviously
increase with increases in the volatility of demand20. Similarly, pt , the expected price
level conditional on past information, will vary with actual, average inflation rates.
Turning to a specific example, suppose that actual prices follow the random
walk where ϵ t is normal with mean π and variance σ 2.

(12) pt = pt −1 +ϵ t

Then pt = pt −1 + π and (11) becomes

(13) y t =Ɵβ( p ¿ ¿t− pt −1)Ɵβ + y pt ¿

Over a sample period during which π and σ 2 remain roughly constant, and if y pt can be
effectively controlled for, (13) will appear to the econometrician to describe a stable
trade-off between inflation and real output. The addition of lagged inflation rates will
not improve the fit, or alter this conclusion in any way. Yet it is evident from (13) that a
sustained increase in the inflation rate (an increase in π ) will not affect real output.

20
This model is taken, with a few changes, from my earlier [24].
 This is not to say that a distributed lag version of (11) might not perform better
empirically. Thus let the actual rate of inflation follow a first-order autoregressive
scheme or where 0< ρ<1and ϵ t is distributed as before.

Δp t =ρΔpt −1+ϵ t

(14) pt =(1¿¿ – ρ) pt−1 −ρpt −2 +¿ ϵ t ¿ ¿

Then combining (11) and (14):

(15) y t =βΔp t −βΔρpt −1+Ɵβ π + y pt

In econometric terms, the "long-run" slope, or trade-off, would be the sum of the
inflation coefficients, or Ɵβ(1¿ ¿ – ρ), ¿which will not, if (14) is stable, be zero. In
short, one can imagine situations in which empirical Phillips curves exhibit long lags
and situations in which there are no lagged effects. In either case, the "long-run" output-
inflation relationship as calculated or simulated in the conventional way has no bearing
on the actual consequences of pursuing a policy of inflation.
As in the consumption and investment examples, the ability to use (13) or (15) to
forecast the consequences of a change in policy rests crucially on the assumption that
the parameters describing the new policy (in this case π , σ 2 and ρ ) are known by agents.
Over periods for which this assumption is not approximately valid (obviously there
have been, and will continue to be, many such periods) empirical Phillips curves will
appear subject to "parameter drift," describable over the sample period, but
unpredictable for all but the very near future.

6. Policy Considerations

In preceding sections, I have argued in general and by example that there are
compelling empirical and theoretical reasons for believing that a structure of the form

y t +1=F ( y t , x t , Ɵ, ϵ t )
 ( F known, Ɵ fixed, x t "arbitrary") will not be of use for forecasting and policy
evaluation in actual economies. For short-term forecasting, these arguments have long
been anticipated in practice, and models with good (and improvable) tracking properties
have been obtained by permitting and measuring "drift" in the parameter vector Ɵ .
Under adaptive models wllich rationalize these tracking procedures, however, long-run
policy simulations are acknowledged to have infinite variance, which leaves open the
question of quantitative policy evaluation.
One response to this situation seldom defended explicitly today though in
implicit form probably dominant at the most "practical" level of economic advice-
giving, is simply to dismiss questions of the long-term behavior of the economy under
alternative policies and focus instead on obtaining what is viewed as desirable behavior
in the next few quarters. The hope is that the changes in Ɵ induced by policy changes
will occur slowly, and that conditional forecasting based on tracking models will
therefore be roughly accurate for a few periods. This hope is both false and misleading.
First, some policy changes induce immediate jumps in Ɵ : for example, an explicitly
temporary personal income tax surcharge will (c.f. section 5.1) induce an immediate rise
in propensity to consume out of disposable income and consequent errors in short-tern1
conditional forecasts21. Second, even if the induced changes in Ɵ are slow to occur, they
should be counted in the short-term "objective function", yet rarely are. Thus
econometric Phillips curves roughly forecast the initial phase of the current inflation,
but not the "adverse" shift in the curve to which that inflation led.
What kind of structure might be at once consistent with the theoretical
considerations raised in section 4 and with operational, accurate policy evaluation? One
hesitates to indulge the common illusion that "general" structures are more useful than
specific, empirically verified ones; nevertheless, a provisional structure, cautiously
used, will facilitate the remainder of the discussion. As observed in section 4, one
cannot meaningfully discuss optimal decisions of agents under arbitrary sequences { x t }
of future shocks. As an alternative characterization, then, let policies and other
disturbances be viewed as stochastically disturbed functions of the state of the system,
or (parametrically) where G is known, λ is a fixed parameter vector, and ηt a vector of
disturbances.

21
This supply function for goods should be thought of as drawn up given a cleared labor market in i . See
Lucas and Rapping [22] for an analysis of the factors underlying this function.
(16) x t ¿ G( y t , λ ,η t )

Then the remainder of the economy follows where, as indicated, the behavioral
parameters Ɵ vary systematically with the parameters λ governing policy and other
"shocks".

(17) y t +1 ¿ F ( y t , xt ,Ɵ( λ), ϵ t )

The econometric problem in this context is that of estimating the function Ɵ ( λ ) .

In a model of this sort, a policy is viewed as a change in the parameters λ , or in
the function generating the values of policy variables at particular times. A change ill
policy (in λ ) affects the behavior of the system in two ways: first by altering the time
series behavior of x t ; second by leading to modification of the behavioral parameters
Ɵ(λ) governing the rest of the system. Evidently, the way this latter modification can be
expected to occur depends crucially on the way the policy change is carried out. If the
policy change occurs by a sequence of decisions following no discussed or pre-
announced pattern, it will become known to agents only gradually, and then perhaps
largely as higher variance of "noise". In this case, the movement to a new Ɵ ( λ ) ,if it
occurs in a stable way at all, will be unsystematic, and econometrically unpredictable.
If, on the other hand, policy changes occur as fully discussed and understood changes in
rules, there is some hope that the resulting structural changes can be forecast on the
basis of estimation from past data of Ɵ ( λ ) .
It is perhaps necessary to emphasize that this point of view towards conditional
forecasting, due originally to Knight and, in modem form, to Muth, does not attribute to
agents unnatural powers of instantly divining file true structure of policies affecting
them. More modestly, it asserts that agents' responses become predictable to outside
Observers only when there can be some confidence that agents and observers share a
common view of the nature of the shocks which must be forecast by both. The
preference for "rules versus authority" in economic policy making suggested by this
point of view, is not, as I hope is clear, based on any demonstrable optimality properties
of rules-in-general (whatever that might mean). There seems to be no theoretical
argument ruling out the possibility that (for example) delegating economic decision-
making authority to some individual or group might not lead to superior (by some
criterion) economic performance than is attainable under some, or all, hypothetical rules
ill the sense of (16). The point is rather that this possibility cannot ill principle be
substantiated empirically. The only scientific quantitative policy evaluations available to
us are comparisons of the consequences of alternative policy rules.

7. Concluding Remarks

This essay has been devoted to an exposition and elaboration of a single

syllogism: given that the structure of all econometric model consists of optimal decision
rules of economic agents, and that optimal decision rules vary systematically with
changes in the structure of series relevant to the decision maker, it follows that any
change in policy will systematically alter the structure of econometric models.
For the question of the short-term forecasting, or tracking ability of econometric
models, we have seen that this conclusion is of only occasional significance. For issues
involving policy evaluation, in contrast, it is fundamental; for it implies that
comparisons of the effects of alternative policy rules using current macro econometric
models are invalid regardless of the performance of these models over the sample
period or in ex ante short-term forecasting.
The argument is, in part, destructive: the ability to forecast the consequences of
"arbitrary", unannounced sequences of policy decisions, currently claimed (at least
implicitly) by the theory of economic policy, appears to be beyond the capability not
only of the current-generation models, but of conceivable future models as well. On the
other hand, as the consumption example shows, conditional forecasting under the
alternative structure (16) and (17) is, while scientifically more demanding, entirely
operational.
In short, it appears that policy makers, if they wish to forecast the response of
citizens, must take the latter into their confidence. This conclusion, if ill-suited to
current econometric practice, seems to accord well with a preference for democratic
decision making.

REFERENCES

ADIE, Douglas K. The Importance of Expectations for the Phillips Curve Relation,
Research Paper No. 133, Department of Economics, Ohio University (undated).
ANDO, Albert; MODIGLIANI, Franco. The Life Cycle Hypothesis of Saving;
Aggregate Implications and Tests, American Economic Review, v. 53, 1963, pp. 55-
84.

COOLEY, Thomas F.; PRESCOTT, Edward C. An Adaptive Regression Model,

International Economic Review, June, 1973, 364-71.
COOLEY, Thomas F.; PRESCOTT, Edward C. Tests of the Adaptive Regression
Model, Review of Economics and Statistics, April 1973, 248-56.

COOLEY, Thomas F.; PRESCOTT, Edward C. Estimation in the Presence of

Sequential Parameter Variation, Econometrica, forthcoming.

MENIL, George; ENZLER, Jared J. Prices and Wagesi n the FRB-MIT-Penn

Econometric Model, in ECKSTEIN, Otto, ed., The Econometrics of Price
Determination Conference. Washington: Board of Governors of the Federal Reserve
System and Social Science Research Council, 1972, pp. 277-308.

DOLDE, Walter, Capital Markets and the Relevant Horizon for Consumption
Planning, Yale doctoral dissertation, 1973.

ROBERT, Eisner. Fiscal and Monetary Policy Reconsidered, American Economic

Review, v. 59 (1969), pp. 897-905.

EVANS, Michael K.; KLEIN, Lawrence R. The Wharton Econometric Forecasting

Model. 2nd, Enlarged Edition. Philadelphia: University of Pennsylvania. Economics
Research Unit, 1968.

FISHER, Franklin M., Discussion in ECKSTEIN, Otto, ed., The Econometrics of

Price Determination Conference. Washington: Board of Governors of the Federal
Reserve System and Social Science Research Council, pp. 113-115.

FRIEDMAN, Milton, A Theory of the Consumption Function. Princeton: Princeton

University Press, 1957.

FRIEDMAN, Milton, Windfalls, the 'Horizon, and Related Concepts in the Permanent
Income Hypothesis, in Carl F. Christ, et. al., eds., Measurement in Economics.
Stanford: Stanford University Press, 1963, pp. 3-28.

FRIEDMAN, Milton, The Role of Monetary Policy, American Economi Review r v.

58, 1968, pp. 1-17.

GORDON, Robert J., Wage-Price Controls and the Shifting Phillips Curve, Brookings
Papers on Economic Activity, 1972, no. 2, pp. 385-421.

HALL, Robert E. JORGENSON, Dale W. Tax Policy and Investment Behavior,

American Economic Review, v. 57, 1967, pp. 391-414.

HIRSCH, Albert A., "Price Simulations with the OBE Econometric Model," in
ECKSTEIN, Otto, ed., The Econometrics of Price Determination Conference.
Washington: Board of Governors of the Federal Reserve System and Social Science
Research Council, pp. 237-276.

HYMANS, Saul H., "Prices and Price Behavior in ECKSTEIN, Otto, ed., The
Econometrics of Price Determination Conference. Washington: Board of Governors
of the Federal Reserve System and Social Science Research Council, pp. 309-322.

KLEIN, Benjamin, "The Effect of Price Level Unpredictability on the Composition of

Income Change," unpublished working paper, April, 1973.

KLEIN, Lawrence R.; GOLDBERGER, Arthur S. An Econometric Model of The

United States: 1929-1952.Amsterdam: North Holland, 1955.

KLEIN, Lawrence R. An Essay on the Theory of Economic Predietion. Helsinki:Yrjo

Jaimsson Lectures, 1968.

KNIGHT, Frank H., Risk, Uncertainty and Profit. Boston: Houghton-Mifflin, 1921.

LUCAS JR., Robert E; RAPPING, Leonard A. Real Wages, Employment, and Inflation.
Journal of Political Economy, v. 77, 1969, pp. 721-754.

LUCAS JR., Robert E; Econometric Testing of The Natural Rate Hypothesis. in

ECKSTEIN, Otto, ed., The Econometrics of Price Determination Conference
(Washington: Board of Governors of the Federal Reserve System and Social Science
Research Council), pp. 50-59.

LUCAS JR., Robert E. Some International Evidence on Output-Inflation Trade-Offs.

American Economic Review, v. 63 (1973).

MARSCHAK, Jacob, "Economic Measurements for Policy and Prediction," in HOOD,

William C.; KOOPMANS, Tjalling G., eds., Studies in Econometric Method. Cowle
Commission Monograph 14, New York: Wiley, 1953, pp. 1-26.

MAYER, Thomas. Tests of the Permanent Income Theory with Continuous Budgets.
Journal of Money, Credit, and Banking., v. 4, 1972, pp. 757-778.

MODIGLIANI, Franco; BRUMBERG, Richard, Utility Analysis and the Consumption

Function: An Interpretation of Cross-Section Data, in KURIHARA; K. K. ed., Post-
Keynesian Economics. New Brunswick: Rutgers University Press, 1954.

MUTH, John F. Optimal Properties of Exponentially Weighted Forecasts. Journal of

the American Statistical Association, v. 55, 1960, pp. 299-306.

MUTH, John F. Rational Expectations and tide Theory of Price Movements,

Econometrica, v. 29, 1961, pp. 315-335.

PHELPS, Edmund S. Money Wage Dynamics and Labor Market Equilibrium. Journal
of Political Economy, v. 76, 1968, pp. 687-71 1.

PHELPS, Edmund S., et al. The New Microeconomics in Employment and Inflation
Theory. New York: Norton, 1970.

PHILLIPS, Alban. W. The Relation Between Unemployment and the Rate of Change of
Money Wage Rates in the United Kingdom, 1861-1957. Economica. 2 v. 25 (1958), pp.
283-299.

SAMUELSON, Paul A.; SOLOW, Robert M. Analytical Aspects of Anti-Inflation

Policy. American Economic Review, v. 50 (1960), pp. 177-194.

SARGENT, Thomas J. A Note on the 'Accelerationist' Controversy. Journal of Money,

Credit, and Banking, v v. 3 (1971), pp. 721-725.. 3, 1971, pp. 721-725.

TINBERGEN, Jan. On the Theory of Economic Policy. Amsterdam: North Holland,

1952.

TINBERGEN, Jan. Econoinic Policy: Principles and Design. Amsterdam: North

Holland, 1956.