Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
http:www.aeaweb.org/articles.php?doi=10.1257/jel.49.3.686
An empirical review of the three fiscal stimulus packages of the 2000s shows that they
had little if any direct impact on consumption or government purchases. Households
largely saved the transfers and tax rebates. The federal government only increased
purchases by a small amount. State and local governments saved their stimulus grants
and shifted spending away from purchases to transfers. Counterfactual simulations
show that the stimulus-induced decrease in state and local government purchases was
larger than the increase in federal purchases. Simulations also show that a larger
stimulus package with the same design as the 2009 stimulus would not have increased
government purchases or consumption by a larger amount. These results raise doubts
about the efficacy of such packages adding weight to similar assessments reached more
than thirty years ago. (JEL E21, E23, E32, E62, H50)
686
Taylor: An Empirical Analysis of the Revival of Fiscal Activism in the 2000s 687
2. Methodological Issues in Policy packages is that they will simply repeat the
Evaluation same prediction story over again. You learn
virtually nothing about the efficacy of a
First consider the basic idea behind stimulus package if you use the same models
Keynesian countercyclical fiscal policy as to evaluate its impact ex post that you used
presentedalong with alternative views to predict its impact ex ante. Indeed, this
in college textbooks. A decline in aggregate is one reason for the disagreement about
demand, caused, say, by a decline in invest- the impact of the recent stimulus packages.
ment (I), can be offset by increasing govern- The same models are frequently being used
ment purchases (G) or temporarily increasing in policy evaluation studies, which are then
transfer payments or tax refunds. In terms of referred to in many of the debates about
the Keynesian cross diagram, a shift down policy.3
in the aggregate expenditures line due to To be concrete, consider two models
the fall in investment can be countered by relating the size of the stimulus package S
increasing government purchases, which to output Y. Model A is Y=S+Z and
shifts the line back up. Government pur- model B is Y=Z, where Z is an unobserv-
chasesaugmented by possible multiplier able shock and is a coefficient that I set to
effectsthus fill the gap left by the decline 1.5. Now, suppose that a stimulus is enacted
in investment. Countercyclical changes in with S=2, but Y decreases by 1. Then the
income tax payments and transfers work the shock implied by model A is Z=4 while
same way except that consumption (C) fills the shock implied by model B is Z=1.
the gap. Now consider policy evaluation of the
Estimated macro models used for policy stimulus based on a counterfactual where
evaluationwhether Keynesian or new there is no stimulus so S=0. Economists
Keynesianhave this basic mechanism built using model A would say: Just as we pre-
into them. However, they differ greatly in dicted, the stimulus package worked.
their predictions of the policy impact because Without it, Y would have fallen to 4 rather
of different assumptions about expectations, than 1. The decline in output would have
the marginal propensity to consume, the been four times as deep, a Great Depression
degree of consumption smoothing, the speed 2.0. Economists using model B would sim-
of price adjustment, and crowding out of ply say Just as we predicted the stimulus
other spending as G is raised. For example, package did not work.
Romer and Bernstein (2009) used Keynesian One way to tackle this problem is to look
models without forward looking expectations at the direct effect of the stimulus pack-
to predict the effect of the stimulus pack- ages within the context of the Keynesian
age of 2009the American Recovery and paradigm, but without imposing a rigid
Reinvestment Act (ARRA)before it was
implemented. They predicted large effects 3 For example, the quarterly impact reports by the
of the package with multipliers around 1.5. Congressional Budget Office (2011) focus on existing mod-
In contrast, John F. Cogan et al. (2010) used els while alternatives models are discussed in the testimony
a new Keynesian model to predict the effects by Taylor (2010). An example of how these simulation stud-
ies are referred to in the media is the news article by Jackie
of ARRA before it was implemented. They Calmes and Michael Cooper (2009) who wrote The accu-
predicted a much smaller effect, with multi- mulation of hard data and real-life experience has allowed
pliers averaging 0.5. more dispassionate analysts to reach a consensus that the
stimulus package, messy as it is, is working, offering as evi-
The problem with using these existing dence simulations from the same models which had pre-
macro models for the evaluation of actual dicted large impacts of the stimulus package in advance.
688 Journal of Economic Literature, Vol. XLIX (September 2011)
Billions of dollars
320 2008
280
240
200
120
80
40
0
01 02 03 04 05 06 07 08 09 10
Though these packages differed in size, and 2009 stimulus packages was that they
duration, and the mechanism for distribution should be temporary, as well as targeted and
of the stimulus payments,6 they were quite timely. This temporary feature distinguishes
similar from the point of view of macroeco- these actions from more permanent changes
nomics because they were all widely viewed as such as the personal income tax rate cuts in
temporary and were justified on the grounds the 1960s and 1980s.
of stimulating or jump-starting consumption.7 Now consider the direct impact which
In fact, a major principle underlying the 2008 these temporary changes in dispos-
able personal income may have had on
6 The 2001 tax rebates could be viewed as an advanced
consumption.8 It would be too narrow an
installment on the more permanent tax cut passed that
year; the 2009 stimulus had more refundable credits and
was implemented in part by a change in withholding. focuses on assessing the Keynesian macroeconomic stabili-
7 Other rationales are sometimes given for stimulus zation rationale for these packages.
packages, including that the payments or government 8 In Taylor (2009), I looked at the 2001 and 2008 pro-
purchases are appropriate in their own right. This paper grams using monthly data from BEA and found that they
690 Journal of Economic Literature, Vol. XLIX (September 2011)
Billions of dollars
12,000
11,600
10,400
10,000
9,600
07Q1 07Q3 08Q1 08Q3 09Q1 09Q3 10Q1 10Q3 11Q1
Figure 2. Quarterly Disposable Personal Income, with and without Stimulus, and Personal Consumption
Table 1
Quarterly PCE Regressions With and Without Stimulus Payments
Disposable personal
incomewithout stimulus 0.857 0.851
(73.0) (60.4)
Notes: The dependent variable is personal consumption expenditures. The t-statistics in parentheses are based on
NeweyWest standard errors. The consumption and income variables are from the Bureau of Economic Analysis
with the stimulus payments calculated as in the text. The oil price variable is West Texas Intermediate from the U.S.
Energy Information Association and net worth variable is Household Net Worth from the Flow of Funds, table
B-100, line 42. Sample period is 2000Q12011Q1.
More precise information about the direct personal income without the stimulus. In the
impact of the stimulus payments on con- third equation, the stimulus payments for
sumption can be obtained from regression 2001, 2008, and 2009 are added as a separate
estimates. Table 1 reports the results from variable. In all three regressions, oil prices
three regressions in which personal con- and household net worth are included as
sumption expenditures is the left-hand side control variables. Higher oil prices would
variable. In regression equation1, displayed be expected to depress consumption while
in column 1, the income variable is simply higher net worth should have a positive
disposable p ersonal income (which includes effect, both with some lag, and this is what
the stimulus). In the second regression the regressions show with the lag equal to
equation, the income variable is disposable two quarters.
692 Journal of Economic Literature, Vol. XLIX (September 2011)
Billions of dollars
(annual rates)
280
240
200
160
80
Grants to state and
local governments
40
Federal government consumption
Federal government investment
0
09Q1 09Q2 09Q3 09Q4 10Q1 10Q2 10Q3 10Q4 11Q1
4.1 Federal Government Purchases 4.2 State and Local Government Purchases
The most striking finding in figure 3 is that State and local governments received
only a small part of ARRA went to purchases substantial grants under ARRA as shown in
of goods and services by the federal govern- the bar chart. The purpose of sending these
ment. Measured as a percentage of GDP the grants to the states was to encourage them
amounts were immaterial: At the maximum to start infrastructure projects and purchase
effect, which occurred in the third quarter other goods and services. But this is not what
of 2010, federal government purchases due happened.
to ARRA reached only 0.21 percent of GDP Consider figure 4, which shows the ARRA
and federal infrastructure only 0.05 percent grants along with the change in state and
of GDP. local government purchases, borrowing, and
These amounts are too small for the stim- expenditures other than government pur-
ulus package to have had a significant effect chases relative to the fourth quarter of 2008 as
on the overall economy. In this case, the published in the BEAs National Income and
debate over the size of the government pur- Product Accounts. ARRA grants increased
chases multiplier is largely moot because steadily from the first quarter of 2009 through
the government purchases multiplier had the third quarter of 2010 before tapering off.
virtually nothing to multiply at the federal But state and local government purchases
level. hardly changed at all during this period. The
694 Journal of Economic Literature, Vol. XLIX (September 2011)
Billions of dollars
(annual rates)
150
ARRA Grants
100
Other expenditures
50
Government purchases
0
50
Borrowing (net)
100
150
2009Q1 2009Q3 2010Q1 2010Q3 2011Q1
coefficient on the ARRA grant variable in A=0. The counterfactual path is compared
the other expenditures equation is positive; with the actual path of ARRA grants in fig-
since the sum of these coefficients must be ure5 and the results of the simulation are
approximately zero, they are nearly equal shown in figures 69.
but of opposite signs, meaning ARRA had In each of figures 6 through 9, the histori-
no effect on the sum of purchases and other cal data are shown along with the counterfac-
expenditures. tual simulation. Also shown is the dynamic
simulation of the three-equation system; this
4.4 Counterfactual Simulations with the
simulation sets the variable A equal to the
Estimated Model
actual ARRA grants, but sets the residuals to
To investigate the counterfactual hypoth- zero rather than to the estimated residuals.
esis of no ARRA program, and thereby illus- In all cases, the dynamic simulations closely
trate the impact of ARRA, one can simulate track the historical data indicating that the
the three-equation system for the case where model fits the data well.
696 Journal of Economic Literature, Vol. XLIX (September 2011)
Billions of dollars
140
120
ARRA Grants
100
80
60
40
20
Counterfactual
0
07Q1 07Q3 08Q1 08Q3 09Q1 09Q3 10Q1 10Q3 11Q1
Figure 5. Actual and Counterfactual ARRA Grants to State and Local Governments
Figure 6 shows that in the absence of the governments were liquidity or borrowing
2009 stimulus grants, net borrowing by state constrained following the financial crisis and
and local governments would have been they simply could not have borrowed more if
greater than it was with the grants. This is ARRA had not existed. At the least increased
consistent with the view that state and local borrowing spreads would have reduced the
governments tried to smooth their expen- incentives to borrow.
ditures in the face of temporary changes in However, an examination of the changes
income, much as households without bor- in state and local government financial assets
rowing constraints did. and liabilities using the Flow of Funds data
from the Federal Reserve shows that the
4.5 The Plausibility of the Counterfactual counterfactual increase in net borrowing
would have been quite likely even if there
One might question the plausibility of were such borrowing constraints. As a matter
these simulations, arguing that many state of accounting, an increase in net borrowing
Taylor: An Empirical Analysis of the Revival of Fiscal Activism in the 2000s 697
Billions of dollars
240
200
160
120
80
Counterfactual simulation
40 Dynamic simulation
Historical data
0
07Q1 07Q3 08Q1 08Q3 09Q1 09Q3 10Q1 10Q3 11Q1
Figure 6. Borrowing (net) by State and Local Governments: Historical and Counterfactual without ARRA
occurs when the net acquisition of financial decrease in net borrowing during the first
assets is smaller than the net increase in lia- two years of ARRA). Thus state and local
bilities. So net borrowing can decrease when governments were adding significantly to
there is a decrease in the acquisition of finan- their financial assets as ARRA grants came
cial assets. in. Indeed, it appears that they were sav-
According to annual Flow of Funds ing the grant money rather than using it to
data, net borrowing fell by about $118 bil- increase expenditures.
lion from 2008 to 2010, or during the first These data suggest, therefore, that the
two years of ARRA. This is consistent with counterfactual is quite plausible: For net
quarterly data from the BEA shown in fig- borrowing to have increased in the counter-
ure 4. During this same period, there was factual compared with history, it would have
a net increase in liabilities of $53 billion been enough for the states simply to have
and net acquisition of financial assets of not increased their acquisition of finan-
$171 billion (which gives the $118 billion cial assets by as much as they did. So even
698 Journal of Economic Literature, Vol. XLIX (September 2011)
Billions of dollars
2,350
2,300
2,250
2,200
2,150
2,100
Counterfactual simulation
Dynamic simulation
Historical data
2,050
2,000
07Q1 07Q3 08Q1 08Q3 09Q1 09Q3 10Q1 10Q3 11Q1
without increasing their liabilities, the state same in the absence of the ARRA grants
governments could have increased their as they were with the ARRA grants. In this
net borrowing by reducing their acquisi- sense, ARRA had no impact on total state
tion of financial assets. Given that state and expenditures. But figures 8 and 9 also illus-
local governments increased their financial trate the striking divergence of the compo-
assets by such a large amount during 2009 nents of total expenditurespurchases and
and 2010, the counterfactual net borrowing other expendituresthat ARRA caused.
path in figure 6 seems quite plausible. Why did ARRA cause states to shift
funds away from government purchases
4.6 Expenditure Switching: Hypothesis and toward these other expenditures, which
Test consist largely of transfer programs such
as Medicaid and Temporary Assistance to
Figure 7 shows that total state and local Needy Families (TANF)? Medicaid is the
expenditures would have been about the public health insurance for poor women
Taylor: An Empirical Analysis of the Revival of Fiscal Activism in the 2000s 699
Billions of dollars
1,880
1,840
1,800
1,760
1,720
Counterfactual simulation
Dynamic simulation
Historical data
1,680
1,640
07Q1 07Q3 08Q1 08Q3 09Q1 09Q3 10Q1 10Q3 11Q1
and children, the disabled, and elderly in To test this hypothesis, Cogan and Taylor
nursing homes, while TANF is cash welfare (2010) split ARRA grants (A) into Medicaid
for the poor. One hypothesis is that ARRA (M) and non-Medicaid (N), using the BEA
stipulated that states receiving additional satellite account for ARRA, and ran the
Medicaid grants could not reduce ben- regressions in equations (1), (2), and (3),
efits or restrict eligibility rules relative to with A replaced by M and N. The Medicaid
what they were on July 1, 2008. In some grant variable in the government purchases
states, this meant reversing benefit reduc- equation was negative and significant; the
tions or eligibility restrictions that were estimated coefficient was 0.318 with a
implemented in the previous seven months t-statistic of 2.3. In contrast, the coefficient
before ARRA was passed in February 2009. of the non-Medicaid coefficient was insig-
This hold-harmless provision could have nificantly different from zero. Hence the
forced states to shift funds away from pur- statistical results confirm the hold harmless
chases to transfers. hypothesis.
700 Journal of Economic Literature, Vol. XLIX (September 2011)
Billions of dollars
500
480
460
Counterfactual simulation
Dynamic simulation
Historical data
440
420
400
380
360
07Q1 07Q3 08Q1 08Q3 09Q1 09Q3 10Q1 10Q3 11Q1
To the extent that government purchases less than they would have been without
had a greater impact on GDP than tem- ARRA.
porary transferswhich the permanent As early as the summer of 2009, it was
income theory predictsthen ARRA could becoming apparent that the recovery of the
have had a negative effect on the economic U.S. economy from the recession of 200709
recovery by reducing purchases and increas- had little to do with government purchases
ing transfers by the same amount. Moreover, related to the stimulus. For example, Cogan,
according to the simulations in figure 8, the Taylor, and Volker Wieland (2009) reported
cumulative negative effect on state and local that nondefense government purchases con-
government purchases was $85 billion. This tributed less than 1 percentage points to
was nearly three times as large as the $30 bil- the 5.4 percentage point real GDP growth
lion cumulative positive effect of ARRA on improvement from the first to the second
federal government purchases. The results quarter of 2009. A comprehensive interna-
indicate that government purchases were tional comparison by Hyunseung Oh and
Taylor: An Empirical Analysis of the Revival of Fiscal Activism in the 2000s 701
Ricardo Reis (2011) shows that government simple example of model A versus model B
purchases did not increase by very much in presented above, these results are evidence
many other countries during the recovery against the views represented by model A,
from the recent recession. The review here and thus against using such models to show
of empirical work on how ARRA worked that things would have been worse.
provides an explanation in the case of the Others argue that the stimulus was too
United States. Despite the stated intention small, but the results do not lend support to
to increase infrastructure and other govern- that view either. Using the estimated equa-
ment purchases through large grants to the tions, a counterfactual simulation of a larger
states, ARRA did not deliver the intended stimulus packagewith the proportions
increase. going to state and local grants, federal pur-
chases, and transfers to individuals the same
as in ARRAwould show little change in
5. Conclusion
government purchases or consumption, as
In sum, this empirical examination of the the temporary funds would be largely saved.
direct effects of the three countercyclical Of course, the story would be different for a
stimulus packages of the 2000s indicates that stimulus program designed more effectively
they did not have a positive effect on con- to increase purchases, but it is not clear that
sumption and government purchases, and such a program would be politically or oper-
thus did not counter the decline in invest- ationally feasible.
ment during the recessions as the basic More generally, the results from the 2000s
Keynesian textbook model would suggest. experience raise considerable doubts about
Individuals and families largely saved the the efficacy of temporary discretionary
transfers and tax rebates. The federal gov- countercyclical fiscal policy in practice. In
ernment increased purchases, but by only this regard, the experience with the stimu-
an immaterial amount. State and local gov- lus packages of the 2000s adds more weight
ernments used the stimulus grants to reduce to the position reached more than thirty
their net borrowing (largely by acquiring years ago by Lucas and Sargent (1978) and
more financial assets) rather than to increase Gramlich (1978, 1979).
expenditures, and they shifted expenditures
away from purchases toward transfers. References
Some argue that the economy would have
Calmes, Jackie, and Michael Cooper. 2009. New Con-
been worse off without these stimulus pack- sensus Sees Stimulus Package as Worthy Step. New
ages, but the results do not support that view. York Times, November 21, p. A1.
According to the empirical estimates of the Cogan, John F., Tobias Cwik, John B. Taylor, and Volker
Wieland. 2010. New Keynesian versus Old Keynes-
impact of ARRA, if there had been no tem- ian Government Spending Multipliers. Journal of
porary stimulus payments to individuals or Economic Dynamics and Control, 34(3): 28195.
families, their total consumption would have Cogan, John F., and John B. Taylor. 2010. What the
Government Purchases Multiplier Actually Multi-
been about the same. And if there had been plied in the 2009 Stimulus Package. National Bureau
no ARRA grants to states and localities, their of Economic Research Working Paper 16505.
total expenditures would have been about Cogan, John F., John B. Taylor, and Volker Wieland.
2009. The Stimulus Didnt Work. Wall Street Jour-
the same. The counterfactual simulations nal, September 17.
show that the ARRA-induced decline in state Congressional Budget Office. 2011. Estimated Impact
and local government purchases was larger of the American Recovery and Reinvestment Act on
Employment and Economic Output from January
than the increase in federal government 2011 through March 2011. Washington, D.C.: Con-
purchases due to ARRA. In terms of the gressional Budget Office.
702 Journal of Economic Literature, Vol. XLIX (September 2011)
Conley, Timothy, and Bill Dupor. 2011. The American Parker, Jonathan A., Nicholas S. Souleles, David S.
Recovery and Reinvestment Act: Public Sector Jobs Johnson, and Robert McClelland. 2011. Consumer
Saved, Private Sector Jobs Forestalled. Ohio State Spending and the Economic Stimulus Payments
University Working Paper. of 2008. National Bureau of Economic Research
Gramlich, Edward M. 1978. State and Local Bud- Working Paper 16684.
gets the Day After It Rained: Why Is the Surplus So Ramey, Valerie A. 2011. Can Government Purchases
High? Brookings Papers on Economic Activity, 1: Stimulate the Economy? Journal of Economic Lit-
191214. erature, 49(3): 67385.
Gramlich, Edward M. 1979. Stimulating the Macro Shapiro, Matthew D., and Joel Slemrod. 2003. Con-
Economy through State and Local Governments. sumer Response to Tax Rebates. American Eco-
American Economic Review, 69(2): 18085. nomic Review, 93(1): 38196.
Johnson, David S., Jonathan A. Parker, and Nicholas Shapiro, Matthew D., and Joel Slemrod. 2009. Did the
S. Souleles. 2006. Household Expenditure and the 2008 Tax Rebates Stimulate Spending? American
Income Tax Rebates of 2001. American Economic Economic Review, 99(2): 37479.
Review, 96(5): 15891610. Taylor, John B. 2000. Reassessing Discretionary Fis-
Lucas, Robert E., and Thomas J. Sargent. 1978. After cal Policy. Journal of Economic Perspectives, 14(3):
Keynesian Macroeconomics. In After the Phillips 2136.
Curve: Persistence of High Inflation and High Unem- Taylor, John B. 2009. The Lack of an Empirical Ratio-
ployment, 4972. Boston: Federal Reserve Bank of nale for a Revival of Discretionary Fiscal Policy.
Boston. American Economic Review, 99(2): 55055.
Oh, Hyunseung, and Ricardo Reis. 2011. Targeted Taylor, John B. 2010. Perspectives on the U.S.
Transfers and the Fiscal Response to the Great Economy: Fiscal Policy Issues. Testimony Before
Recession. National Bureau of Economic Research the Committee on the Budget, U.S. House of
Working Paper 16775. Representatives.
This article has been cited by:
1. 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]