Blinder - Solow - JME (1976)
Blinder - Solow - JME (1976)
Blinder - Solow - JME (1976)
@ North-Holland
Publishing Company
Alan S. BLINDER
Princeton University, Princeton, NJ 08540, U.S. A.
Robert M. SOLOW*
MIT, Cambridge, MA 02139, U.S.A.
1. Introduction
Ettore Infante and Jerome Stein have produced a fair and painstaking review
of our two articles, Does fiscal policy matter? and Analytical foundations
of fiscal policy (henceforth, Analyticalj. We find their section 2 very balanced
and ver! accurate, and thank them for avoiding what must have been an irresistible temptation to refer to us by the first letters of our respective last names.
However, in sections 3 anti14, they suggest that fiscal policy rests on rather
weak analytical foundations after a11 At the risk of sounding like inveterate
Pollyannas, we disagree. As we stressed in Analytical, there is still much to be
learned about the quantitative dimensions and timing of the economys response
to fiscal-policy actions. But *he theoretical channels and qualitative effects of
fiscal policy seem relatively well established, at least as well established as any
piece of macroeconomic theory.
We have divided our response into several sections. In section 2 we bring our
views on the monetarist-Keynesian controversy up to date (circa spring 19X),
and try to put the contribution of Does fiscal policy matter? into perspective.
Section 3 examines the puzzling results that Infante and Stein obtained with the
fiscalist model, and explains why we find them not so puzzling after all. In
section 4 we restate - more clearly, we hope - the message of Does fiscal policy
matter ?, and show why we think the Infante-Stein critique misses the mark.
Our biggest disagreements are saved for section 5, where we examine the way in
which Infante and Stein relate the theoretical findings from Does fiscal policy
matter? to the results of large-scale econometric models. We find the analysis
there at best misleading and at worst just plain wro::g*
*FinanciA support from the National Scierce Foundation is gratefully acknowledged.
502
A rep/y
503
are made, important issues remain unresolved. How large are the integers
Kand M? What is the height of the peak real multiplier? What is the ultimate
nominal multiplier? Since the economy is no doubt nonlinear, how do these
answers change with variations in initial conditions? The answers to these
questions make a great deal of difference. 3
The portions of our two papers which addressed themselves to the monetaristKeynesian controversy circa 1972 sought to demonstrate that those monetarists
who contended that fiscal policy actions had no long-run effect OIZaggregate
denmd were involved in a logical contradiction. We showed that either the longrun effect is bigger than the short-run effect or the response of the economy is
explosive (in either direction). Zero effect is simply not a logical possibility.
The Infante-Stein critique does not refute this view. They are right, however,
to challenge our notion that the unstable case could be ruled out on logical
grounds. We must admit that a fair reader of our papers might come away with
the view that: Since the economic system we observe seems to be stable, Blinder
and Solow necessarily conclude that the monetarist scenario cannot be cozect
(Infante and Stein, p. 484). That was an error.
Infante and Stein are also correct to question the casual empiricism which
*established the stability condition
1-T
FB > T
where Fs is the IS-LM multiplier effect of bonds (B) on real output (Y), and T
is the marginal propensity to tax and reduce transfers as income rises.4 For one
thing, FB is not simply a transfer multiplier - a point Infante and Stein neglect
in their section 4. An increment in B has effects like a transfer payment through
the accompanying interest payments; but it also has wealth effects. Secondly,
interest on the national debo is taxable; most other transfers are not. Thirdly,
recipients of government interest payments mzy have much lower marginal
propensities to consume than recipients of other transfers. Finally, we no longer
believe that the empirical evidence supports our guesstimate that T is at least
one-ha1 f.
All of this adds up to the admission that the size of Fu is a serious empirical
issue, not to be settled by our armchair speculation or anyone clses. It ma! well
be that (1) is reversed. In that case, a government which stubbornly refused to
finance budget imbalances at least partly by printing or destroyinr! rnonev could
destabilize an otherwise stable economy. We imagine that mosi poverin?ents
would not be so stubborn.
jNone of the evidence on the steepness and speed of adaptation of the long-run Phillips
curve is entirely convincing, and prevailing views may change as the recsnt rathe special
episode recedes.
40n this see Scarth (1976).
I)
504
505
506
likely to be stable if deficits are financed by prhting money than if they are
@awed by -floating bonds. The second message is that the long-run e$?ct of
government spending on aggregate demand is greater when dejcits are bondjinanced that when they are money-financed. Both messages are logical con-
sequences of the fact that government bonds pay interest, while money does not.
Pnfante and Stein claim to have refuted these messages by showing that the
stability of the rock-bottom model does not imply either the positivity of this
cumulative multiplier [the steady-state multiplier d Y*/dG] or that this multiplier
is larger when deficits are financed through the sale of bonds than when the
deficits are financed by the issuance of new money (p. 485). However, it takes
only a little algebra - actually provided by Infante and Stein in their footnote 6
(p. 487) - to show that d Y*/dG is both positive and bigger under bond-finance
than under money finance whenever the stability condition (1) is satisfied and
TF,
< 1.
(2)
Condition (2), which holds automatically for the model in Does fiscal policy
matteV8
but which becomes an empirical issue when induced investment
(I, > di is allowed, is not only plausible, but almost inevitable. It simply says
that government spending does not Iead to a birdgct surplus. While WCwould argue
that (2) shot&l be assumed as a matter of course, a more fundamental point is
that the mecharlrism which leads to our results operates in reverse if (2) is reversed.
For then raising G leads to a budget surplus and retirements of government
obligations, with consequent contractionary effects on aggregate demand.
Under those circumstances, the whole notion of bond-financed deficit spending
loses meaning.
A direct corollary of the second message is that au open-market ptrrchase of
bortds, while expansionary in the short r-m, is cori/r*actiortaryirl tltc long run.
Infante and Stein find this an unusual result (p. 492), and suggest that: Both
neo-Keynesians and monetarists will be surprised by this direct implication of the
rock-bottom model (p. 492). Furthermore, they assert, because of this result the antithesis of the quantity theory - serious questions and doubts naturally
arise from the use of this simple model (p, 492).
We find it easy to explain the result.g In the short run (meaning here: with
stocks fixed), the open-market purchase expands demand for the usual reasons.
8Pruof.
where
us
(y#$-L.)
>
0.
FG- exceeds Tbecause aL, > 0, and 0 < T < 1 imply that
1 -C1+d,
507
But, with government spending and taxes unchanged, a budget surplus results.
Either money or bonds must be retired to finince this surplus, and these actions
have a contractionary impact on aggregate demand. We know that the cumulative contraction must overwhelm the initial expansion because the condition
for budget balance requires
G+B = T( Y+B).
(3)
With lower B and unchanged G in the new stock equilibrium, Y must be lower
(ifthe model converges).
Those monetarists who have studied the important work of Brunner and
Meltzer carcfuily will also feel at home with this conclusion. Brunner and
Meltzer (1976) work with a model in which long-run output is fixed at full
employment v bile P is free to ad-just. They find that long-run equilibrium price
level is /oI~Y~c(/bv an open market purchase. The government budget constraint
permits no other*conciusioi?. Fix P (as we did) and the result will npp:ar in the
form of reduced real out!lut: fix Y (as Brunner-Mcitzer did) and the result will
appear in the form of l&er P.
5. Rlzwocconomic theory :;i!deconometric models
Infante and Stein (section 4) find glaring discrepancies between the predictions
of our theoretic;~i model and the policy multipliers derived by simulating
empirical macro models. We ;lgree th;it there are major discrepancies, but draw
rather different inferences from them.
Before examining the evidence, there is an important methodological issue.
Suppose that a skii theoretical macro model and a large empirical model reach
dramatic& . different conclusions. Which do we accept ? To us, no general answer
can be provided to this quAon. In each particular case, we must examine the
reasons for the disagreement. Did the theoretical model assume away an
empirically important phenomenon? If so, we would believe the empirical model.
Was the econometric mode specified in a fundamentally illogical way? If so,
WCwould trust the theoretical model.
When we apply this rather eclectic methodology to the issues raised by
Infante and Stein, we find that we must side with our theoretical specification
(for re;tsons explained below). This is not terribly surprising. Why does anyone
ther with simple and unrealistic theoretical model? of the macroeconomy?
imagine because simple analytical models have tight logical structures which
can be used to check on the rather unwieldy empirical models, and to suggest
new effects that may be worth capturing.
Our analysis of long-run wealth effects is a case in point. Most of the econometric models surveyed by Infante and Stein omit wealth effects entirely, and
trc;it government interest payments 1s exogenous. Very few pay attention to the
50s
government budget constraint [see Christ (1971)]. How, then, can they be used
to confirm or refute our theoretical notions?
Infante and Stein compare our theoretical notions to the findings of the
econometric models on five questions. They conclude that the models do not
support the implications drawn by Blinder and Solow from the rock-bottom
model (p. 499). Let us examine each of these questions in turn.
Do Y and P react in the same direction ? The verdict of the models seems to be
%IO- probably not in the short run and definitely not in the long run.
First, we would take issue with Infante and Steins identification of secondquarter multipliers from econometric models with our impact multipliers. Our
impact multipliers refer to a hypothetical scenario in which all the lags in the
IS-LM sector work themselves out, while P remains fixed. While P is nearly
unresponsive to G within two quarters, empirical consumption and investment
functions certainlv suggest that the IS-LM lags are far longer than two quarters.
Second, there is a transitory phenomenon that is always ignored by theoretical
modli:ls. Short-run surges in aggregate demand lead to productivity improvements which hold down, and may even lower, prices. This is the reason why
dP;dG is negative for a few quarters in most econometric models. Third, most
econ,ometric moJe1 builders would admit that price-determination is one of their
weakest links. WJien Fromm and Klein wrote the article which Infante and Stein
use. killmodels were demand-oriented, and paid almost no attention to the supply
tide of the economy. In Wharton 111,for example, a stimulus to aggregate demand
wcrs prices for at least 4 years.
Do mhpIicr paths reach appropriate hits? We would expect (see footnote 1)
both real and nominal values to show damped oscillations towards limiting
values which would be relatively small or zero for the former and substantially
positive for the latter.
Infante and Stein use 40.quarter multipliers (except for the MPS model) as
empirical representations of our theoretical long-run multipliers. It is well
known. however, that these econometric models are designed almost exclusively
for short-run analysis, and often exhibit unbelievable behavior if the computer is
allowed to whir away long enough. In our view, 40,quarter multipliers tell more
about the long-run behavior of difference-equation systems than they do about
the U.S. economy. This is why we restricted our examination of results from the
models to 12 quarters in Analytical. We are, therefore, unimpressed by the fact
that 40-quarter simul Itions are at variance with our theoretical notions.
Despite this, let us look at the evidence. Infante and Stein have the most
trouble accounting for the real multipliers, dY/dG, which look bizarre by anyones standards (see their table 3). What is going on here? Inspection of Fromm
and Kleins table 5 shows that:
For the Brookings and DHL models, the simulations seem to be converging
toward zero or a small positive value.
509
(2) The MPS model after 16 quarters gives the appearance of convergence.
But those familiar with its properties know that it damps very slowly. If
allowed to run 40 quarters, it would probably still be cycling.
(3) DRI and Wharton show the behavior that MPS would have shown: cycles
which are either anti-damped or damp very slowly.
(4) The BEA model simulation looks as if the computer blew a fuse.
If we ignore the BEA model, these results do not look so bad after all.
In fact, they look better than they should, for most of the simulations reported
by Fromm and Klein were based on strstaind irtcrements ift nomitzalgovertfment
spetding. Real multipliers were computed by dividing the observed change in
real GNP, quarter by quarter, by the observed change in real G. The latter falls
endogenously over time as the government deflator rises. So the real multipliers
refer to u /urge initial incwasc in G f~~/low~d by a wccession of smal! decreases.
It is no wonder, then, that such multipliers may turn negative and take a long
time to damp down.
Does the stabiiitvI condition hold? Here Infante and Stein commit the same
error we did in our (futile) exercise in casual empiricism - trying to use an
empirical tax multiplier to represent FB. Furthermore, we repeat that two-quarter
econometric multipliers are not the same as our theoretical impact multipliers.
DWS art oplw-market purchasr raLw Y and P in the short mu ?Infante and
Stein, using some results tabulated by Christ (1975), find that Y generally rises
but P often falls. The explanation for this perversity probably lies in the shortrun productivity gains cited earlier.
DWJ au opcmmarket ptrrchu.sc iowr Y arld P i/l th lotlg rrrlz ? Infante and
Stein find that most of the models imply that Y and P rise instead. Our reasons
for ignoring the 4O-quarter mllltipliers- of econometric models were explained
previously. lo But, even if these were waived, the fact remains that none of the
models contains the causal chain from budget imbalances to both future wealth
and future interest payments, which is what makes an open-market purchase
contractionary in our model.
6, Concludingcommeet
Infante and Stein give the distinct impression that there is no convergence of
views m the operation of fiscal policy. We think tl:,ere is convergence, on both
the empirical and theoretical levels. The c%ef concliusions of our two papers are
also implied by the Brunner and Meltzer model. When Blinder and Solow agree
with Brunner and Meltzer, is there not convergence of views? It is Aso our
lThe monetary-policy
simulations
quarters in the DRI and MQEM model.,, 16 quarters in the Wharton, 14 quarters in the
Liu-Hwn(monthly) model, and 64 quarters in the Hickman-Coen (annual) model.
510
impression that a repeat in 1976 of the Fromm-Klein comparison of fiscalpolicy multipliers would show the models much closer together. The current
Wharton model looks much more like the MPS than it did then; DRI is now far
less fiscalist; the Brookings model is a thing of the past.
The absurd results which Infante and Stein obtain with the fiscalist model
show that it is no better, and no worse, than iMV = P Y. Even a model as simple
as our rock-bottom model of aggregate demand does not have similarly absurd
implications as long as deficit spending leads to deficits rather than to surpluses.
When the discrepancies between our theoretical predictions and results from
econometric models are held up to closer scrutiny, either the disparities look less
significant or the econometric models fall apart. This does not mean that
empirical results have supported our theoretical suggestions, only that the latter
have never really been tested.
That some kind of consensus now exists does not mean that we have at last
stumbled upon the truth. There are already new theoretical developments,
surprising empirical findings and novel policy issues which raise the possibility
that the conceptual underpinnings of fiscal policy may be quite different in the
year 2000. But the undeniable fact that we 30 not know everything does not
mean that we do not know anything. Looking back over the Infante-Stein
critique remindc us of a sentence in Analytical: We do not believe that the roof
fell in, much less that the foundation crumbled (p. 11).
References
Blinder, A.S. and R.M. Solow, 1973, Does fiscal policy matter?, Journal of Public Economics 2,
319-337.
Blinder, A.S. and R.M. Solow, 1974, Analytical foundations of fiscal policy, in: A.S. Blinder
et al., The economics of public finance (Brookings, Washington, DC) 3-l 15.
Brunner, K. and A.H. Meltzer, 1976, An aggregative theory for a closed economy, in: J.L.
Stein, e&, Monetarism (North-Holland, Amsterdam) 69-103.
Christ, C.F., 1971, Econometric models of the financial sector, Journal of Money, Credit and
Banking 3, no. 2419-449.
Christ, C.F., 1975, Judging the performance of econometric models of the U.S. economy,
International Economic Review 16,54-73.
Fromm, G. and L.R. Klein, 1973, A comparison of eleven econometric models of the United
Sta?es, American Economic Review 63,385-393.
Infante, E.L. and J.L. Stein, 1976, Does fiscal policy matter?, Journal of Monetary Economics
2,473~500.
Scarth, WM., 1976,, A note on the crowding out of private expenditures by bond-financed
increases in government spending, Journal od Public Economics, 5,385-387.
Stein, J.L., 1976, Monetarism (North-Holland, Amsterdam).
Tobin, J. and W.H. Buiter, 1976, Long-run effects of fiscal and monetary policy on aggregate
demand, in: J.L. Stein, ed., Monetarism (North-Holland, Amsterdam) 273-309.