Can Government Purchases Stimulate the Economy?
←
→
Page content transcription
If your browser does not render page correctly, please read the page content below
Journal of Economic Literature 2011, 49:3, 673–685
http:www.aeaweb.org/articles.php?doi=10.1257/jel.49.3.673
Can Government Purchases
Stimulate the Economy?
Valerie A. Ramey*
This essay briefly reviews the state of knowledge about the government spending
multiplier. Drawing on theoretical work, aggregate empirical estimates from the
United States, as well as cross-locality estimates, I assess the likely range of multiplier
values for the experiment most relevant to the stimulus package debate: a temporary,
deficit-financed increase in government purchases. I conclude that the multiplier for
this type of spending is probably between 0.8 and 1.5. ( JEL E23, E62, H50)
1. Introduction rates, however, it became abundantly clear
that more research was needed.
O ne of the few positive effects of the
recent financial crisis has been the
revival of interest in the short-run macroeco-
Given the upsurge in research on this
topic, we now have many more resources to
draw upon when asked “what is the govern-
nomic effects of government spending and ment spending multiplier?” In this essay, I
tax changes. Before 2008, the topic of stim- will begin by briefly reviewing what theory
ulus effects of fiscal policy was a backwater has to say about the potential effects. As I
compared to research on monetary policy. will discuss, “the multiplier” is a nebulous
One reason for the lack of interest was the concept that depends very much on the
belief that the lags in implementing fiscal type of government spending, its persis-
policy were typically too long to be useful tence, and how it is financed. I will then go
for combating recessions. Perhaps another on to review the aggregate empirical evi-
reason was that central banks sponsored dence for the United States, as well as the
many more conferences than government cross-locality evidence on multipliers. I will
treasury departments. When the economy conclude that the U.S. aggregate multiplier
fell off the cliff in 2008 and the Fed reached for a temporary, deficit-financed increase
the dreaded “zero lower bound” on interest in government purchases (that enter sepa-
rately in the utility function and have no
direct effect on private sector production
* University of California, San Diego, and National functions) is probably between 0.8 and 1.5.
Bureau of Economic Research. I wish to thank Davide
Debortoli, Roger Gordon, John Taylor, and Irina Telyu-
Reasonable people can argue, however,
kova for helpful comments. that the data do not reject 0.5 or 2.0.
673674 Journal of Economic Literature, Vol. XLIX (September 2011)
2. Brief Review of the Theory beginning of the period, g T is the transi-
tory component of government spending, g P
In this section, I briefly review the leading is the persistent component of government
theories on the effects of government spend- spending, g is total government spending,
ing. An important point to keep in mind is u(c, n) is the utility function, and f (k, n) is
that all of the theories hinge fundamentally the production function. Primes denote the
on the effect of government spending on next period’s value of variables. There exist
equilibrium hours worked, and how those unique solutions to the utility maximization
hours translate to output. Absent instanta- subproblem, so that optimal labor supply
neous adjustment of the capital stock, total and consumption can be written as:
output can only rise in the short-run if hours
worked rise. Thus, the multiplier is intimately n = h(k, k′ + g) and c = q(k, k′ + g).
linked to the effect of government spending
on equilibrium hours and to the extent of It can be shown that the function h is strictly
diminishing returns to labor. increasing in g since a rise in g represents a
negative wealth effect (and leisure is assumed
2.1 Models in the Neoclassical Tradition
to be a normal good). For the same reason,
In neoclassical models, the key channels the function q is strictly decreasing in g.
through which fiscal policy affects the private Aiyagari, Christiano, and Eichenbaum
economy are wealth effects, intertemporal sub- (1992) decompose the effect of government
stitution effects, and distortions to first-order spending on hours as follows:
conditions (e.g., Robert J. Barro and Robert G.
King 1984, Marianne Baxter and King 1993, ∂ h + _
dn = _
_ ∂ h _
∂ k′ .
dg ∂g ∂ k′ ∂ g
and S. Rao Aiyagari, Lawrence J. Christiano,
and Martin Eichenbaum 1992). To see this, The first term captures the static effect and
consider first a standard neoclassical model the second term captures the dynamic effect.
with no distortionary taxes. The social planner As discussed above, the first term is posi-
maximizes the discounted utility of the repre- tive because of the negative wealth effect.
sentative household subject to the production Because g and k′ enter symmetrically in the h
function and resource constraints. Following function, ∂ h/∂ k′ = ∂ h/∂ g. The size of ∂ k′/∂ g
Aiyagari, Christiano, and Eichenbaum (1992), depends on whether the increase in g is tran-
we can write the standard Bellman equation to sitory or persistent. Aiyagari, Christiano, and
distinguish static from dynamic effects of gov- Eichenbaum (1992) show that
(_
∂ k′ ) > (_
∂ k′ ) ,
ernment spending:
P T
v(k, g , g ) =
P T
max ∂g ∂g
k′≤ f (k, N) − g
{W(k, k′ + g) + E(β v(k′, g P′, g T′ ) | g P)},
so a persistent increase in government spend-
ing raises next period’s desired capital stock
by more. Thus, a persistent increase in gov-
where W(k, k′ + g) ernment spending raises hours more now.
=
max u(c, n). In this model with no distortionary taxa-
c,n∈{0≤n≤N; 0≤c≤ f (k,n)−(g+k′ )} tion, Ricardian equivalence reigns, so it does
not matter whether government purchases
In these equations, c is consumption, n is are financed with current taxes or defi-
hours worked, k is the capital stock at the cit spending. Results change considerably,Ramey: Can Government Purchases Stimulate the Economy? 675
though, when spending is financed with dis- can be as high as 1.2 or as low as –2.5, depend-
tortionary taxes. For example, a rise in cur- ing on the nature of the experiment.
rent distortionary labor income taxes tends
2.2 Models in the Keynesian Tradition
to depress output and hours.
Baxter and King (1993) catalog the possi- The basic idea of the multiplier is illus-
ble range of government spending multipli- trated in the so-called “Keynesian cross
ers using a standard calibration of a dynamic diagram” that is the staple of undergraduate
general equilibrium model. They find that macroeconomics. If interest rates are held
the lowest multipliers result when (1) the constant, then the multiplier for government
increase in government spending is tempo- spending is given by 1/(1 – mpc) and for taxes
rary and (2) governments raise distortion- is given by –mpc/(1 – mpc), where mpc is the
ary taxes concurrently to keep the budget marginal propensity to consume. Allowing
balanced. In this case, the multiplier can be for open economy considerations (i.e., a mar-
as low as negative 2.5. Multipliers for tem- ginal propensity to import) or rises in interest
porary increases in government spending rates lowers the multiplier, whereas allowing
financed with deficits (to be paid with future for accelerator effects in investment can raise
lump-sum taxes) are somewhat higher, but the multiplier. Even in extended models, the
are still substantially below unity. Permanent size of the multiplier is intimately linked to
increases in government spending financed the marginal propensity to consume.
by current or future lump-sum taxes give As Jordi Galí, J. David López-Salido, and
larger multipliers because the greater nega- Javier Vallés (2007) and John F. Cogan et al.
tive wealth effect raises labor supply more (2010) discuss, the typical New Keynesian
and the steady-state capital stock rises, which model (e.g., Frank Smets and Rafael
leads to a rise in investment. In this case, the Wouters 2007) predicts a much smaller
short-run multiplier is just below unity and multiplier. Since the New Keynesian model
the long-run multiplier is around 1.2. builds a sticky-price edifice on a neoclassi-
As Craig Burnside, Eichenbaum, and cal foundation, neoclassical effects tend to
Jonas D. M. Fisher (2004) note, on average in mute the Keynesian multiplier. Cogan et al.
the post–World War II data, large increases (2010) use the Smets–Wouters model to esti-
in government spending are typically fol- mate multipliers that are equal to or less than
lowed by hump-shaped rises in distortionary unity. Galí, López-Salido, and Vallés (2007)
taxes. Although they do not discuss multipli- are able to obtain multipliers as high as 2.0,
ers explicitly, the graphs from the analysis of but only when they make the following two
models with paths of distortionary taxes lead assumptions: (1) at least fifty percent of con-
to higher positive short-run multipliers than sumers are rule-of-thumb consumers, so that
in the lump-sum tax case. The multiplier is the marginal propensity to consume is much
higher because of intertemporal substitution higher than would be the case if consum-
effects: because individuals know that taxes ers behaved optimally; and (2) employment
will be higher in the future, they intertem- is demand-determined, so that workers are
porally substitute more labor to the present always willing to supply as many hours as
when taxes are relatively low. firms demand. These two assumptions essen-
Thus, the neoclassical model predicts that tially convert the New Keynesian model back
the government spending multiplier can be into a traditional Keynesian model.
negative or positive, depending on the extent Within the new Keynesian model, how-
and timing of distortionary taxes. For reason- ever, there is one way in which multipliers
able parameter values, the short-run m ultiplier can be made larger without resorting to676 Journal of Economic Literature, Vol. XLIX (September 2011)
widespread nonoptimizing behavior. This is homogenous agent model with perfect capi-
the case of the “zero lower bound.” Gauti B. tal markets, a temporary rise in transfers now
Eggertsson (2001, forthcoming), Eggertsson should have no effect because of perma-
and Michael Woodford (2003), Christiano, nent income hypothesis considerations and
Eichenbaum, and Sergio Rebelo (2011) and Ricardian equivalence. Oh and Reis (2011)
Woodford (2011) explore fiscal policy in New explore some simple models that relax these
Keynesian models in which the economy is assumptions but are not able to generate
caught in a deflationary spiral at the zero much bigger effects.
lower bound. A deficit-financed increase in All of the models discussed above assume
government spending leads expectations of the economy starts out in a steady-state in
inflation to increase. When nominal inter- which capital is fully utilized and workers are
est rates are held constant, this increase in fully employed. A key question is whether
expected inflation drives the real interest government spending multipliers can be
rate down, spurring the economy. Christiano, greater if the economy starts out with under
Eichenbaum, and Rebelo show that if inter- utilized resources, which is widely believed
est rates are held constant for twelve quar- to be the case in 2009. It seems that this
ters and government spending goes up would be a promising area for more theo-
during this time, the multiplier peaks at 2.3. retical research. Below, I will discuss some
empirical work that has found that the mag-
2.3 Other Considerations
nitude of the multiplier does seem to depend
Most of the models discussed above on the state of the economy.
abstract from three potentially important To summarize this section, the theoreti-
features: (1) productive government spend- cal work on government spending gives a
ing; (2) transfers; and (3) underutilization of wide range of possible values of the multi-
resources. I will briefly discuss how each of plier, depending on the type of model used,
these might change the predictions. the assumptions about how monetary policy
In the last section of their paper, Baxter behaves, the type and persistence of govern-
and King (1993) consider the multiplier ment spending, and how it is financed. It is
effects of an increase in investment in pub- necessary, therefore, to turn to the data to
lic capital. In the case of public capital that see if we can narrow the range.
raises the marginal product of private inputs,
the multiplier can be quite large, somewhere
3. Aggregate Time Series Evidence
between 4.0 and 13.0 in the long run, but
much lower in the short run. Thus, consid- All of the theories of the multiplier dis-
ering productive government spending does cussed above are general equilibrium theo-
not raise the predicted short-run stimulus ries, so aggregate data is the most natural
effects in the neoclassical model. place to study the strength of the multiplier.
As Hyunseung Oh and Ricardo Reis Numerous studies have been conducted on
(2011) and Cogan and John B. Taylor (2011) a variety of countries. In order to focus this
point out, government purchases barely section, I will concentrate on the U.S. evi-
increased in 2009 and 2010 despite the large dence. Studies of multiple countries, such
stimulus package. As both papers point out, as by Roberto Perotti (2005), Roel Beetsma,
most of the stimulus package was allocated Massimo Giuliodori, and Franc Klaassen
to transfers. Most models, both neoclas- (2008), Daniel Leigh et al. (2010), Ethan
sical and New Keynesian, treat transfers Ilzetzki, Enrique G. Mendoza, and Carlos A.
like a negative lump-sum tax. In the typical Végh (2010), Beetsma and Giuliodori (2011),Ramey: Can Government Purchases Stimulate the Economy? 677
and others, tend to find multipliers in the model, the Klein–Goldberger model, and
range of those discussed here for the United the Brookings model. He found multipliers
States.1 on government spending of about 2.0, both
Since most aggregate studies measure in the short-run and the long-run. He also
what happens on average when government discussed the estimated marginal propensity
spending changes, it is very important to to consume in the models. In the Wharton
keep track of the characteristics of the exper- and Klein–Goldberger models, the short-run
iments covered by the analyses. For example, marginal propensity to consume was esti-
to measure the effect of a deficit-financed mated to be 0.55 and the long-run one was
increase in government spending, one needs estimated to be 0.74.
to focus on periods in which taxes did not Subsequent analyses have tried to come to
change significantly or one needs to control terms with the Robert E. Lucas’ (1976) and
for tax effects. Tax multiplier estimates range Christopher A. Sims’ (1980) critiques of this
from – 0.5 to – 5.0, so it is difficult to choose earlier literature. Most aggregate analyses of
a single number to control for tax effects.2 In the last several decades have relied on vector
addition, because stimulus packages are sup- autoregressions (VARs) or dynamic simula-
posed to be temporary, we ideally would like tions to estimate the effects of government
to measure the effect of temporary changes. spending and tax changes. None of these
Also important is whether the economy had analyses is immune to potential problems of
underutilized resources at the time of the identification, though. For example, Barro
government spending increase. (1981), Robert E. Hall (1986), Valerie A.
Table 1 gives a summary of just a few of the Ramey and Matthew D. Shapiro (1998), Hall
representative studies using aggregate data to (2009), Fisher and Ryan Peters (2010), Ramey
estimate government spending multipliers; it (2011), and Barro and Redlick (2011) all focus
is by no means meant to be exhaustive.3 The on military buildups under the assumption
Michael K. Evans (1969) paper is represen- that this type of government spending is the
tative of the discussion of fiscal multipliers in least likely to respond to economic events.
the heyday of traditional Keynesianism and Nevertheless, there is always the possibility
the big econometric models. Evans com- that the events that lead to these buildups,
pared multipliers for sustained increases in such as the start of World War II and the
government spending across the Wharton start of the Cold War, could have other influ-
ences on the economy apart from the effects
on government spending. For example, dur-
1 However, there is also a literature that finds some evi- ing World War II increased patriotism could
dence that fiscal contractions can be expansionary. See, for have raised labor supply more than would be
example, Francesco Giavazzi and Marco Pagano (1990),
Alberto Alesina and Perotti (1997), and Alesina and Silvia predicted by economic incentives and hence
Ardagna (2010). could raise the multiplier. In contrast, ration-
2 Examples of estimates of the tax multiplier are –0.5
ing and capacity constraints during the world
(Carlo Favero and Giavazzi forthcoming), –1.1 (Barro and
Charles J. Redlick 2011), –3.0 (Christina D. Romer and wars could dampen the multiplier.
David H. Romer 2010), and –5.0 (Andrew Mountford and Numerous studies have followed Olivier
Harald Uhlig 2009).
3 Examples of other studies that use methods similar
Blanchard and Perotti (2002) by using
to some of the studies listed or investigate the robustness Choleski decompositions to identify
of those techniques are Antonio Fatás and Ilian Mihov government spending shocks and by using
(2001), Perotti (2005), Evi Pappa (2005), Dario Caldara assumptions on tax elasticities in a structural
and Christophe Kamps (2008), Caldara (2011), Tommaso
Monacelli, Perotti, and Antonella Trigari (2010), and Jörn VAR (SVAR) to identify tax shocks. SVAR
Tenhofen and Guntram Wolff (2007). methods have the advantage that they are678 Journal of Economic Literature, Vol. XLIX (September 2011)
Table 1
Examples of Aggregate Analyses on U.S. Data
Study Sample Identification Implied spending multiplier
Evans (1969) Quarterly, 1948–62 Based on estimates of equations Slightly above 2.0 in all models
of Wharton, Klein-Goldberger,
and Brookings models
Barro (1981), Hall Annual, various Use military spending as 0.6–1.0
(1986), Hall (2009), samples, some going instrument for government
Barro and Redlick back to 1889 spending
(2011)
Rotemberg and Quarterly, Shocks are residuals from 1.25
Woodford (1992) 1947–89 regression of military spending
on own lags and lags of military
employment
Ramey and Shapiro Quarterly, 1947–late Dynamic simulations or VARs 0.6–1.2, depending on sample
(1998), Edelberg, 1990s or 2000s using Ramey-Shapiro dates, and whether calculated as
Eichenbaum, and which are based on narrative cumulative or peak
Fisher (1999), evidence of anticipated military
Eichenbaum and buildups
Fisher (2005),
Cavallo (2005)
Blanchard and Quarterly, SVARS, Choleski decomposition 0.9 to 1.29, depending on
Perotti (2002) 1960–97 with G ordered first assumptions about trends
Mountford and Quarterly, Sign restrictions on a VAR 0.65 for a deficit-financed
Uhlig (2009) 1955–2000 increase in spending
Romer and Quarterly Average multipliers from Rising to 1.57 by the 8th
Bernstein (2009) FRB/US model and a private quarter
forecasting firm model
Cogan et al. (2010) Quarterly, 1966–2004 Estimated Smets–Wouters 0.64 at peak
model
Ramey (2011) Quarterly, 1939–2008 VAR using shocks to the 0.6 to 1.2, depending on
and subsamples expected present discounted sample
value of government spending
caused by military events, based
on narrative evidence
Fisher and Peters Quarterly, 1960–2007 VAR using shocks to the 1.5 based on cumulative
(2010) excess stock returns of military effects
contractors
Auerbach and Quarterly, 1947–2008 SVAR that controls for Expansion: –0.3 to 0.8
Gorodnichenko professional forecasts, Ramey Recession: 1.0 to 3.6
(forthcoming) news
Key innovation is regime
switching model
Gordon and Krenn Quarterly, 1919–41 Choleski decomposition in VAR 1.8 if no capacity constraints
(2010)Ramey: Can Government Purchases Stimulate the Economy? 679
easy to implement and do not require exten- Fisher and Peters (2010) use excess returns
sive data gathering. Ramey and Shapiro of defense contractor stocks as news to esti-
(1998), Ramey (2011), and Eric M. Leeper, mate multipliers of 1.5 for the period 1960–
Todd B. Walker, and Shu-Chun Susan Yang 2007. Their impulse response functions show
(2011) have criticized traditional VAR meth- no significant rise in taxes for their sample,
ods, though, arguing that most changes in so the increases in government spending
government spending and taxes are antici- they identify appear to be deficit-financed.
pated and showing that this can invalidate However, their government spending shocks
inferences from procedures that do not seem to be quite persistent. In particular, in
account for anticipations. Moreover, Caldara contrast to the work by Ramey (2011) and
(2011) shows that small changes in the others, which shows that government spend-
assumed elasticities of taxes and government ing returns to normal by sixteen quarters,
spending in the structural VAR result in large Fisher and Peters’s estimate suggests a very
differences in the estimated multipliers. persistent increase in government spend-
Table 1 shows, however, that despite sig- ing, barely falling even after twenty quarters.
nificant differences in samples, experiments, (See the lower right panel of their figure 5.).
and identification methods, most aggregate Given that permanent increases in govern-
studies estimate a range of multipliers from ment spending imply larger multipliers than
around 0.6 to 1.8. Moreover, the range within temporary increases in a neoclassical model,
studies is almost as wide as the range across their estimate of 1.5 may be somewhat above
studies, and the standard errors are always the relevant one for considering temporary
large. Thus, despite a healthy debate on stimulus packages.
methodology, most studies are giving similar Several recent aggregate studies consider
answers. the possibility that the multiplier may dif-
These government spending multipliers fer according to the state of the economy.
do not necessarily represent deficit-financed Manuel Coutinho Pereira and Artur Silva
increases in government spending, which is Lopes (2010) and Markus Kirchner, Jacopo
the type most likely to be used in a stimulus Cimadomo, and Sebastian Hauptmeier
package. For example, the lower end of my (2010) use time-varying parameters and
multiplier estimates (Ramey 2011) are from Bayesian estimation techniques and find
samples where the Korean War is dominant, that government spending multipliers are
and hence are samples in which much of the not very different in expansions and contrac-
spending was financed by tax increases. Even tions. In contrast, Alan Auerbach and Yuriy
during World War II, some of the increase Gorodnichenko (forthcoming) use a regime
in government spending was financed with switching model to estimate multipliers that
taxes. Barro and Redlick (2011) control for can differ according to whether the economy
the average marginal tax rate and find gov- is in recession or not. Estimation of such a
ernment spending multipliers of only 0.6. model is far from trivial, and many subtle
Using the framework in Ramey (2011), I issues arise in estimation. When Auerbach
study the effect of holding marginal tax rates and Gorodnichenko do not allow the regime
constant on the path of output and find no to change endogenously, they find identical
significant change from the original multi- impact multipliers in the two regimes, but
plier estimate of approximately unity.4
recompute the impulse response functions holding the
4 In particular, I use the estimates from the VAR Barro–Redlick tax rate constant and calculate the implied
described on pages 29–30 of Ramey (2011). I then multiplier.680 Journal of Economic Literature, Vol. XLIX (September 2011)
different estimated dynamics that imply very government spending. Thus, they observe a
large multipliers in recessions compared large increase in output in response to what
to expansions, 2.2 in recessions and –0.3 appears to be a modest increase in current
in expansions (see table 1 of their paper). government spending. An alternative inter-
The estimated dynamic behavior of some pretation is that the large increase in output
of the variables is odd in this experiment, is the result of firms gearing up for antici-
which I suspect is caused by the assumption pated large future increases in government
that the economy never switches regimes.5 spending. In fact, Barro and Redlick (2011)
Fortunately, Auerbach and Gorodnichenko show that including my news variable elimi-
also discuss results where they allow feed- nates interaction effects with unemployment
back, so that the economy can endogenously in their specification.
switch between regimes. Figure 3 of their Yet another issue is the possibility that the
paper shows the historical multipliers based multiplier is greater at the zero lower bound,
on this experiment. In this case, they obtain as discussed in the theoretical section above.
multipliers between 0 and 0.5 during expan- Some of the authors of these papers have
sions and between 1.0 and 1.5 during reces- argued that since most of the estimates of
sions. The results from this more general multipliers have been over time periods in
model in which the economy is allowed to which interest rates were not at the lower
move between regimes seem more plausible. bound, they do not apply to the current
Robert J. Gordon and Robert Krenn situation.
(2010) also discuss the role of underutilized In fact, we do have historical evidence
capacity on government spending multipli- from periods with very low interest rates.
ers. They create a new quarterly data set From 1939 to the second quarter of 1947,
extending back to 1913 and study the role of the rate on Treasury bills never rose above
government spending in increasing output 0.38 percent although the average annual
in 1940. They find a multiplier of only 0.9 if rate of inflation was six percent over this time
they extend the sample to the fourth quarter period. In Ramey (2011, p. 38), I describe
of 1941, but a multiplier of 1.8 if they stop results showing that when I limit the sample
the sample in the second quarter of 1941. to the period covering 1939 to 1949, I find a
They give arguments and detailed evidence multiplier of 0.7 (but with even larger than
that the U.S. economy started hitting capac- normal standard errors due to the reduced
ity constraints in some sectors after the sec- sample). Thus, I find no evidence of larger
ond quarter of 1941. multipliers during the extended period in
My narrative analysis of this period (Ramey which interest rates were held virtually con-
2009), however, suggests that some of what stant at the zero lower bound.
Gordon and Krenn measure as a multiplier Based on these considerations and the esti-
may actually be an anticipation effect. Since mates available, I would argue that, despite
Gordon and Krenn use a standard Choleski significant differences in methodology, the
decomposition in a VAR, they do not con- range of plausible estimates for the multi-
trol for anticipations of future increases in plier in the case of a temporary increase in
government spending that is deficit financed
is probably 0.8 to 1.5. As discussed above, I
5 For example, the impulse response functions suggest truncated the lower estimates because they
that a shock to government spending during a recession were usually accompanied by increases in
leads to a permanently higher level of government spend-
ing and an ever-increasing path of output (relative to distortionary taxes. I truncated the very high-
trend). See figure 2 of their paper. est estimates because of the various concernsRamey: Can Government Purchases Stimulate the Economy? 681
I expressed above. If the increase is under- $1 to Mississippi and finances it by increas-
taken during a severe recession, the esti- ing lump-sum taxes across all states, the true
mates are likely to be at the upper bound of aggregate multiplier is 0, since the taxes and
this range. It should be understood, however, transfers cancel in the aggregate. However,
that there is significant uncertainty involved if we run a panel regression with time fixed
in these estimates. Reasonable people could effects (which net out the economywide rise
argue that the multiplier is 0.5 or 2.0 without in tax liabilities), we will estimate a multiplier
being contradicted by the data. of mpc/(1 – mpc), where mpc is the marginal
propensity to consume. If the marginal pro-
pensity to consume were 0.6, then we would
4. Cross-State Evidence
estimate a multiplier of 1.5 at the state level,
In their recent review of empirical eco- even though the aggregate multiplier for this
nomics, Joshua D. Angrist and Jorn-Steffen experiment is 0.
Pischke (2010) praised the increase in Daniel Shoag (2010) and Emi Nakamura
empirical standards and the many advances and Jón Steinsson (2011) explore in detail
in applied microeconomics, but bemoaned what these experiments mean if we inter-
the fact that macro and industrial organiza- pret states as small open economies in a
tion were slow to adopt some of these new currency union. As the various versions of
approaches. The exciting new literature on their model show, translating the state-level
cross-state effects of government spending is estimates to aggregate estimates depends
both an answer to Angrist and Pischke, but importantly on the type of spending and
also an explanation for why the techniques the assumptions of the theoretical model.
used in applied microeconomics are not Jeffrey Clemens and Stephen Miran (2011)
always suitable for macroeconomics. present a very useful econometric frame-
One reason that the “natural experiment” work for evaluating the economic context
techniques have been slow to diffuse in mac- of the various natural experiments. Their
roeconomics is that it is difficult to use them discussion highlights many of the compli-
to answer macroeconomic questions. As I will cations that arise in most of the cross-state
discuss shortly, there have been numerous empirical work on the subject.
recent papers using panel data or state cross- Table 2 lists some of the papers that have
section data to estimate the effects of govern- estimated state or region multipliers. This
ment spending on state economies. These literature has focused as much on employ-
studies estimate government purchases or ment effects as income effects, which is
transfers multipliers, holding national effects important in this era of jobless recoveries.
constant. Thus, the studies that look at gov- Many (though not all) papers find positive
ernment transfers are answering the question: employment effects. A notable exception
“When the federal government redistributes is the Lauren Cohen, Joshua D. Coval, and
$1 more to Mississippi than to other states Christopher Malloy (2010) paper, which
(with tax liabilities imposed on all states), finds that an increase in earmarks (induced
what happens to income (or employment) by shifts in political power) leads to a decline
in Mississippi relative to other states?” The in corporate employment in the state.
answer to this question is only indirectly In terms of income multipliers, most
related to the aggregate multiplier. To see estimates lie in the range of 0.5 to 2.0.
the difference, suppose the economy behaves As with the aggregate papers, the ranges
according to a simple traditional Keynesian within papers are sometimes as large as
model. In this case, if the government t ransfers the range across papers. Price V. Fishback682 Journal of Economic Literature, Vol. XLIX (September 2011)
Table 2
Examples of Cross-State Analyses
Study Type of data Identification Results
Davis, Loungani, and Military prime Panel VAR, with military Cost of job created ranges from
Mahidhara (1997) contracts, military variables ordered after $34,000 to $400,000 (in $2010),
personnel, panel of oil but before other depending on employment data
states 1956–92 variables source and allowance for spillovers.
Decreases in military spending have
larger effects than increases
Hooker and Knetter Military procurement Assume military Elasticity of nonfarm payroll
(1997) contracts, panel of procurement contracts employment to real military
states 1963–94 uncorrelated with state contracts per capita is 1.8; decreases
economy in military spending have larger
effects than increases
Fishback and Various types of New Interaction of swing Income multiplier of –0.57 to 1.67,
Kachanovskaya (2010) Deal spending, panel voting and aggregate depending on type of spending;
of states, 1930–40 government spending negligible impact on employment
Cohen, Coval, and Federal earmarks, Whether state senators Decrease in corporate employment,
Malloy (2010) Panel of states, and representatives investment and R&D, suggesting
1967–2008 control powerful crowding out of private activity
committees
Chodorow-Reich et al. Medicaid spending Variations due to pre- $100,000 in spending results in 3.5
(2010) from ARRA in cross- recession medicaid job years
section of states, Dec. spending
2008–June 2010
Wilson (2011) Total ARRA spending State ARRA spending Approximately $25,000 per job
in cross-section of instrumented by Wall created, but job is short-lived
states, Feb. 2009–Feb. Street Journal forecasts
2010, with Oct. 2010
follow-up
Shoag (2010) State government Changes in state Income multiplier around 2.0; each
spending, panel of spending caused by $35,000 generates one additional job
states, 1987–2008 excess returns to state
pension fund returns
Clemens and Miran State government Interaction of state 0.3 to 3.0, depending on
(2011) outlays, panel of states, balanced budget rules specification. Standard errors are
1988–2004 with business cycle large
Nakamura and Military prime contracts, State-specific sensitivity 1.5 income multiplier
Steinsson (2011) panel of states to aggregate changes in
military spending
Suárez Serrato and Federal spending on Changes in federal 1.88 income multiplier; $30,000 per
Wingender (2011) localities, panel of spending on states job created
counties, 1970–2009 caused by updates of
population estimates
based on the CensusRamey: Can Government Purchases Stimulate the Economy? 683
and Valentina Kachanovskaya (2010) study mechanism by which government spend-
New Deal outlays, which are of particular ing raises GDP. Some of the papers find that
interest because of parallels between the government spending leads consumption to
Great Depression and the Great Recession. decline, consistent with the negative wealth
According to their estimates, the types of effect of the neoclassical model. Others find
outlays that had the highest multiplier were that consumption increases, consistent with
public works and relief, with a multiplier of rule-of-thumb consumers. Household stud-
1.7. In contrast, payments to farmers to take ies, such as the work by Jonathan A. Parker
their land out of production had an income et al. (2011), can help shed light on this issue.
multiplier of –0.5. Christopher J. Nekarda and Ramey (2011)
Despite using very different identification present evidence that industry markups do
methods, many of these cross-state studies not change in response to government spend-
find multipliers on purchases or transfers of ing, as required by the New Keynesian model.
about 1.5 to 1.8 for income and an implied Thus, more research is required before we
cost of around $35,000 per job created. understand the mechanism.
Several studies also find that the multiplier It is important to note that none of these
is significantly higher during times of higher estimates sheds light on the welfare conse-
slack (e.g., Shoag 2010, Juan Carlos Suárez quences of temporary increases in govern-
Serrato and Philippe Wingender 2011, and ment spending to stimulate the economy.
Nakamura and Steinsson 2011). These find- Such an analysis would require a better
ings suggest that some types of stimulus understanding of the mechanisms, as well
spending that redistribute resources from as assumptions about whether government
low unemployment states to high unemploy- purchases enter the utility function.
ment states could result in sizable aggregate
multipliers. More research is needed, how- References
ever, to understand how these local multi-
Aiyagari, S. Rao, Lawrence J. Christiano, and Martin
pliers translate to aggregate multipliers. Eichenbaum. 1992. “The Output, Employment, and
Interest Rate Effects of Government Consumption.”
Journal of Monetary Economics, 30(1): 73–86.
5. Conclusions Alesina, Alberto, and Silvia Ardagna. 2010. “Large
Changes in Fiscal Policy: Taxes versus Spending.” In
We now have many more estimates of fis- Tax Policy and the Economy, Volume 24, ed. Jeffrey
cal multipliers than we did in Fall 2008 and R. Brown, 35–68. Chicago and London: University of
Chicago Press.
early 2009, when policymakers were trying to Alesina, Alberto, and Roberto Perotti. 1997. “Fiscal
decide whether to use fiscal policy to try to Adjustments in OECD Countries: Composition and
stimulate the economy. Many of the studies Macroeconomic Effects.” International Monetary
Fund Staff Papers, 44(2): 210–48.
are so recent, however, that the profession Angrist, Joshua D., and Jorn-Steffen Pischke. 2010.
needs more time to interpret results and to “The Credibility Revolution in Empirical Econom-
check their robustness before coming to any ics: How Better Research Design Is Taking the Con
Out of Econometrics.” Journal of Economic Perspec-
firm conclusions. At this point, it seems that tives, 24(2): 3–30.
the bulk of estimates imply that the aggregate Auerbach, Alan, and Yuriy Gorodnichenko. Forthcom-
multiplier for a temporary rise in government ing. “Measuring the Output Responses to Fiscal Pol-
icy.” American Economic Journal: Economic Policy.
purchases not accompanied by an increase Barro, Robert J. 1981. “Output Effects of Govern-
in current distortionary taxes is probably ment Purchases.” Journal of Political Economy,
between 0.8 and 1.5. 89(6): 1086–1121.
Barro, Robert J., and Robert G. King. 1984. “Time-
Despite the increase in the number of Separable Preferences and Intertemporal–Sub-
estimates, there is still no consensus on the stitution Models of Business Cycles.” Quarterly684 Journal of Economic Literature, Vol. XLIX (September 2011)
Journal of Economics, 99(4): 817–39. orporate Downsizing?” National Bureau of Eco-
C
Barro, Robert J., and Charles J. Redlick. 2011. “Mac- nomic Research Working Paper 15839.
roeconomic Effects from Government Purchases Davis, Steven J., Prakash Loungani, and Ramamohan
and Taxes.” Quarterly Journal of Economics, 126(1): Mahidhara. 1997. “Regional Labor Fluctuations:
51–102. Oil Shocks, Military Spending, and Other Driv-
Baxter, Marianne, and Robert G. King. 1993. “Fiscal ing Forces.” Board of Governors of the Federal
Policy in General Equilibrium.” American Eco- Reserve System International Finance Discussion
nomic Review, 83(3): 315–34. Paper 578.
Beetsma, Roel, and Massimo Giuliodori. 2011. “The Edelberg, Wendy, Martin Eichenbaum, and Jonas D.
Effects of Government Purchases Shocks: Review M. Fisher. 1999. “Understanding the Effects of a
and Estimates for the EU.” Economic Journal, Shock to Government Purchases.” Review of Eco-
121(550): F4–32. nomic Dynamics, 2(1): 166–206.
Beetsma, Roel, Massimo Giuliodori, and Franc Klaas- Eggertsson, Gauti B. 2001. “Real Government Spend-
sen. 2008. “The Effects of Public Spending Shocks ing in a Liquidity Trap.” http://www.newyorkfed.
on Trade Balances and Budget Deficits in the Euro- org/research/economists/eggertsson/JMP2.pdf.
pean Union.” Journal of the European Economic Eggertsson, Gauti B. Forthcoming. “What Fis-
Association, 6(2–3): 414–23. cal Policy Is Effective at Zero Interest Rates?” In
Blanchard, Olivier, and Roberto Perotti. 2002. “An NBER Macroeconomic Annual 2010, Volume 25, ed.
Empirical Characterization of the Dynamic Effects Daron Acemoglu and Michael Woodford. Chicago
of Changes in Government Spending and Taxes on and London: University of Chicago Press.
Output.” Quarterly Journal of Economics, 117(4): Eggertsson, Gauti B., and Michael Woodford. 2003.
1329–68. “The Zero Bound on Interest Rates and Optimal
Burnside, Craig, Martin Eichenbaum, and Jonas D. Monetary Policy.” Brookings Papers on Economic
M. Fisher. 2004. “Fiscal Shocks and Their Con- Activity, 1: 139–211.
sequences.” Journal of Economic Theory, 115(1): Eichenbaum, Martin, and Jonas D. M. Fisher. 2005.
89–117. “Fiscal Policy in the Aftermath of 9/11.” Journal of
Caldara, Dario. 2011. “The Analytics of SVARs: A Money, Credit, and Banking, 37(1): 1–22.
Unified Framework to Measure Fiscal Multipliers.” Evans, Michael K. 1969. “Reconstruction and Estima-
http://people.su.se/~dcald/JMP_Caldara.pdf. tion of the Balanced Budget Multiplier.” Review of
Caldara, Dario, and Christophe Kamps. 2008. “What Economics and Statistics, 51(1): 14–25.
Are the Effects of Fiscal Shocks? A VAR-Based Fatás, Antonio, and Ilian Mihov. 2001. “The Effects
Comparative Analysis.” European Central Bank of Fiscal Policy on Consumption and Employment:
Working Paper 877. Theory and Evidence.” Center for Economic and
Cavallo, Michele. 2005. “Government Employment Policy Research Discussion Paper 2760.
and the Dynamic Effects of Fiscal Policy Shocks.” Favero, Carlo, and Francesco Giavazzi. Forthcoming.
Federal Reserve Bank of San Francisco Working “Measuring Tax Multipliers: The Narrative Method
Paper 2005-16. in Fiscal VARs.” American Economic Journal: Eco-
Chodorow-Reich, Gabriel, Laura Feiveson, Zachary nomic Policy.
Liscow, and William Gui Woolston. 2010. “Does State Fishback, Price V., and Valentina Kachanovskaya. 2010.
Fiscal Relief during Recessions Increase Employ- “In Search of the Multiplier for Federal Spending in
ment? Evidence from the American Recovery and the States During the New Deal.” National Bureau of
Reinvestment Act.” http://econgrads.berkeley.edu/ Economic Research Working Paper 16561.
gabecr/files/2011/05/Does-State-Fiscal-Relief-Dur- Fisher, Jonas D. M., and Ryan Peters. 2010. “Using
ing-Recessions-Increase-Employment.pdf. Stock Returns to Identify Government Spending
Christiano, Lawrence J., Martin Eichenbaum, and Shocks.” Economic Journal, 120(544): 414–36.
Sergio Rebelo. 2011. “When Is the Government Galí, Jordi, J. David López-Salido, and Javier Vallés.
Spending Multiplier Large?” Journal of Political 2007. “Understanding the Effects of Government
Economy, 119(1): 78–121. Spending on Consumption.” Journal of the Euro-
Clemens, Jeffrey, and Stephen Miran. 2011. “The pean Economic Association, 5(1): 227–70.
Role of Fiscal Institutions in Analysis of Fiscal Pol- Giavazzi, Francesco, and Marco Pagano. 1990.
icy.” http://www.people.fas.harvard.edu/~clemens/ “Can Severe Fiscal Contractions Be Expansion-
StateBudgetCuts.pdf. ary? Tales of Two Small European Countries.” In
Cogan, John F., Tobias Cwik, John B. Taylor, and Volker NBER M acroeconomics Annual 1990, ed. Olivier
Wieland. 2010. “New Keynesian versus Old Keynes- J. Blanchard and Stanley Fischer, 75–111. Cam-
ian Government Spending Multipliers.” Journal of bridge, Mass. and London: MIT Press.
Economic Dynamics and Control, 34(3): 281–95. Gordon, Robert J., and Robert Krenn. 2010. “The
Cogan, John F., and John B. Taylor. 2011. “What the End of the Great Depression 1939–41: Policy Con-
Government Purchases Multiplier Actually Multi- tributions and Fiscal Multipliers.” National Bureau
plied in the 2009 Stimulus Package.” Unpublished. of Economic Research Working Paper 16380.
Cohen, Lauren, Joshua D. Coval, and Christopher Hall, Robert E. 1980. “Labor Supply and Aggregate
Malloy. 2010. “Do Powerful Politicians Cause Fluctuations.” Carnegie–Rochester ConferenceRamey: Can Government Purchases Stimulate the Economy? 685 Series on Public Policy, 12: 7–33. Working Paper 16684. Hall, Robert E. 1986. “The Role of Consumption in Pereira, Manuel Coutinho, and Artur Silva Lopes. Economic Fluctuations.” In The American Business 2010. “Time-Varying Fiscal Policy in the U.S.” Bank Cycle: Continuity and Change, ed. Roger J. Gor- of Portugal Working Paper 2010-21. don, 237–55. Chicago and London: University of Perotti, Roberto. 2005. “Estimating the Effects of Chicago Press. Fiscal Policy in OECD Countries.” Center for Eco- Hall, Robert E. 2009. “By How Much Does GDP Rise nomic and Policy Research Discussion Paper 4842. If the Government Buys More Output?” Brookings Perotti, Roberto. 2008. “In Search of the Transmis- Papers on Economic Activity, 2: 183–231. sion Mechanism of Fiscal Policy.” In NBER Mac- Hooker, Mark A., and Michael M. Knetter. 1997. “The roeconomics Annual 2007, Volume 22, ed. Daron Effects of Military Spending on Economic Activity: Acemoglu, Kenneth Rogoff, and Michael Wood- Evidence from State Procurement Spending.” Jour- ford, 169–226. Chicago and London: University of nal of Money, Credit, and Banking, 29(3): 400–421. Chicago Press. Ilzetzki, Ethan, Enrique G. Mendoza, and Carlos A. Ramey, Valerie A. 2009. “Defense News Shocks, Végh. 2010. “How Big (Small?) are Fiscal Multi- 1939–2008: Estimates Based on News Sources.” pliers?” National Bureau of Economic Research http://econ.ucsd.edu/~vramey/research/Defense_ Working Paper 16479. News_Narrative.pdf. Kirchner, Markus, Jacopo Cimadomo, and Sebastian Ramey, Valerie A. 2011. “Identifying Government Hauptmeier. 2010. “Transmission of Government Spending Shocks: It’s All in the Timing.” Quarterly Spending Shocks in the Euro Area: Time Varia- Journal of Economics, 126(1): 1–50. tion and Driving Forces.” European Central Bank Ramey, Valerie A., and Matthew D. Shapiro. 1998. Working Paper 1219. “Costly Capital Reallocation and the Effects of Leigh, Daniel, Pete Devries, Charles Freedman, Government Spending.” Carnegie–Rochester Con- Jaime Guajardo, Douglas Laxton, and Andrea Pes- ference Series on Public Policy, 48: 145–94. catori. 2010. “Will It Hurt? Macroeconomic Effects Romer, Christina D., and Jared Bernstein. 2009. of Fiscal Consolidation.” In World Economic Out- “The Job Impact of the American Recovery look: Recovery, Risk, and Rebalancing, 93–124. and Reinvestment Plan.” http://otrans.3cdn. Washington, D.C.: International Monetary Fund. net/45593e8ecbd339d074_l3m6bt1te.pdf. Leeper, Eric M., Todd B. Walker, and Shu-Chun Romer, Christina D., and David H. Romer. 2010. Susan Yang. 2011. “Foresight and Information “The Macroeconomic Effects of Tax Changes: Esti- Flows.” National Bureau of Economic Research mates Based on a New Measure of Fiscal Shocks.” Working Paper 16951. American Economic Review, 100(3): 763–801. Lucas, Robert E. 1976. “Econometric Policy Evalu- Rotemberg, Julio J., and Michael Woodford. 1992. ation: A Critique.” Carnegie–Rochester Conference “Oligopolistic Pricing and the Effects of Aggregate Series on Public Policy, 1: 19–46. Demand on Economic Activity.” Journal of Political Monacelli, Tommaso, Roberto Perotti, and Antonella Economy, 100(6): 1153–1207. Trigari. 2010. “Unemployment Fiscal Multipliers.” Shoag, Daniel. 2010. “The Impact of Government Journal of Monetary Economics, 57(5): 531–53. Spending Shocks: Evidence on the Multiplier from Mountford, Andrew, and Harald Uhlig. 2009. “What State Pension Plan Returns.” http://www.people.fas. Are the Effects of Fiscal Policy Shocks?” Journal of harvard.edu/~shoag/papers_files/shoag_jmp.pdf. Applied Econometrics, 24(6): 960–92. Sims, Christopher A. 1980. “Macroeconomics and Nakamura, Emi, and Jón Steinsson. 2011. “Fiscal Reality.” Econometrica, 48(1): 1–48. Stimulus in a Monetary Union: Evidence from U.S. Smets, Frank, and Rafael Wouters. 2007. “Shocks and Regions.” http://columbia.edu/~en2198/papers/fiscal. Frictions in US Business Cycles: A Bayesian DSGE pdf. Approach.” American Economic Review, 97(3): Nekarda, Christopher J., and Valerie A. Ramey. 2011. 586–606. “Industry Evidence on the Effects of Government Suárez Serrato, Juan Carlos, and Philippe Wingen- Spending.” American Economic Journal: Macroeco- der. 2011. “Estimating Local Fiscal Multipliers.” nomics, 3(1): 36–59. http://www.ocf.berkeley.edu/~jcsuarez/Files/ Oh, Hyunseung, and Ricardo Reis. 2011. “Targeted Suarez_Serrato-Wingender-ELFM.pdf. Transfers and the Fiscal Response to the Great Tenhofen, Jörn, and Guntram B. Wolff. 2007. Recession.” Columbia University Department of “Does Anticipation of Government Spending Matter? Economics Discussion Paper 1011-10. Evidence from an Expectation Augmented VAR.” Pappa, Evi. 2005. “New-Keynesian or RBC Trans- Deutsche Bundesbank Discussion Paper 14. mission? The Effects of Fiscal Shocks in Labour Wilson, Daniel J. 2011. “Fiscal Spending Multipliers: Markets.” Centre for Economic Policy Research Evidence from the 2009 American Recovery and Discussion Paper 5313. Reinvestment Act.” Federal Reserve Bank of San Parker, Jonathan A., Nicholas S. Souleles, David S. Francisco Working Paper 2010-17. Johnson, and Robert McClelland. 2011. “Consumer Woodford, Michael. 2011. “Simple Analytics of the Spending and the Economic Stimulus Payments Government Expenditure Multiplier.” American of 2008.” National Bureau of Economic Research Economic Journal: Macroeconomics, 3(1): 1–35.
This article has been cited by: 1. John B. Taylor. 2011. An Empirical Analysis of the Revival of Fiscal Activism in the 2000s. Journal of Economic Literature 49:3, 686-702. [Abstract] [View PDF article] [PDF with links] 2. Jonathan A. Parker. 2011. On Measuring the Effects of Fiscal Policy in Recessions. Journal of Economic Literature 49:3, 703-718. [Abstract] [View PDF article] [PDF with links]
You can also read
