WHY IS FISCAL POLICY OFTEN
PROCYCLICAL?
Alberto Alesina
Harvard University
Filipe R. Campante
John F. Kennedy School of Government,
Harvard University
Guido Tabellini
IGIER Bocconi
Abstract
Fiscal policy is procyclical in many developing countries. We explain this policy failure with a
political agency problem. Procyclicality is driven by voters who seek to “starve the Leviathan”
to reduce political rents. Voters observe the state of the economy but not the rents appropriated
by corrupt governments. When they observe a boom, voters optimally demand more public
goods or lower taxes, and this induces a procyclical bias in fiscal policy. The empirical evidence
is consistent with this explanation: Procyclicality of fiscal policy is more pronounced in more
corrupt democracies. (JEL: E62, D73, D78)
1. Introduction
Most economists agree with the normative prescription that tax rates and dis-
cretionary government spending as a fraction of GDP ought to remain constant
over the business cycle. If governments respected these prescriptions, we should
observe a countercyclical pattern in fiscal policy. Namely, during a boom, (i) total
government spending as a share of GDP should go down because of automatic
stabilizers (if discretionary spending remained constant in real terms, the effect
would be reinforced); (ii) with constant tax rates and some degree of progressiv-
ity, government revenues as a share of GDP should go up (the effect would be
reinforced by tax cuts in recessions and tax increases in boom.); (iii) as a result,
The editor in charge of this article was Jordi Galí.
Acknowledgments: We thank Jordi Galí and three anonymous referees for excellent comments. We
are also grateful to Robert Barro, Eliana La Ferrara, Paolo Manasse, Roberto Perotti, Kenneth Rogoff,
Francesco Trebbi, Aleh Tsivinsky, and participants in a seminar at Harvard for useful suggestions,
Ashoka Mody and Diego Saravia for help with part of the data collection, and Davide Cantoni for
outstanding research assistantship and comments. We gratefully acknowledge financial support from
NSF through a grant from NBER (Alesina) and CIFAR, MIUR, and Bocconi University (Tabellini).
E-mail addresses: Alesina: aalesina@harvard.edu; Campante: filipe_campante@harvard.edu;
Tabellini: guido.tabellini@unibocconi.it
Journal of the European Economic Association September 2008 6(5):1006–1036
© 2008 by the European Economic Association
Alesina, Campante, and Tabellini Why Is Fiscal Policy Often Procyclical? 1007
budget surpluses as a share of GDP should increase. The opposite should occur
in recessions.
1
In practice, in many developing countries fiscal policy has the opposite prop-
erties: it is procyclical. In particular, government spending as a share of GDP
goes up during booms and down in recessions, and deficits increase in booms and
decrease in recessions. In OECD countries, in contrast, fiscal policy is generally
countercyclical.
2
Gavin and Perotti (1997) were the first to point out that in Latin
America fiscal policy is procyclical; then Talvi and Vegh (2005), Catão and Sut-
ton (2002), Manasse (2005), and Kaminski, Reinhart, and Vegh (2004) noted that
this is not a Latin American phenomenon only: Procyclicality of fiscal policy is
common in many—though not all—developing countries.
Why do many countries follow seemingly sub-optimal procyclical fiscal poli-
cies that add to macroeconomic instability? A common answer relies upon the
supply of credit. In bad times, many developing countries cannot borrow, or can
do so only at very high interest rates, therefore they cannot run deficits and have
to cut spending; in booms, they can borrow more easily and choose to do so,
increasing public spending (cf. Gavin and Perotti 1997; Catão and Sutton 2002;
Riascos and Vegh 2003; Kaminski, Reinhart, and Vegh 2004).
3
In our view, however, this argument is incomplete, because it begs two critical
questions. First, why don’t these countries self-insure by accumulating reserves
in good times, so that they are less likely to face binding credit constraints in
recessions? Second, why would lenders not provide funds to countries, even in
recessions, if they were convinced that the borrowing would optimally smooth
out the cycle?
To answer both questions one needs to consider the political arena, as others
have done and as we do in this paper. In Talvi and Vegh (2005), the presence of
surpluses increases the government propensity to spend; the present article and
Talvi and Vegh (2005) are complementary because they focus on how a politi-
cal distortion interacts with different economic structures (in particular with the
variability of revenues). In this article, we focus on deriving the political distor-
tion. A different but not mutually exclusive political explanation is the “voracity
effect” of Tornell and Lane (1999), Lane and Tornell (1998), and Lane (2003):
1. In light of the careful discussion of Kaminski, Reinhart, and Vegh (2004) we want to be clear
regarding our choice of words. We define as countercyclical a policy that follows the tax-smoothing
principle of holding constant tax rates and discretionary government spending as a fraction of GDP
over the cycle. They define such policy as “acyclical.” Both we and they would define as procyclical
a policy in which tax rates go down in booms and up in recessions, and spending over GDP goes up
in booms. As those authors themselves note, our definition is the most common in the literature.
2. Some countries in both groups have accumulated large amounts of public debt. For a review
of models that explain excessive deficits, see Alesina and Perotti (1995) and Persson and Tabellini
(2000). On the cyclical properties of fiscal policy in OECD countries, see Galí and Perotti (2003).
3. Riascos and Vegh (2003) provide a formalization of the credit channel, whereas most of the
other reports are only empirical.
1008 Journal of the European Economic Association
When more resources are available (i.e., in booms), the common-pool problem is
more severe, and the fight over common resources intensifies, leading to budget
deficits. We consider empirically this possibility as well.
In our article, voters face corrupt governments that can appropriate part of
tax revenues for unproductive public consumption, namely, political rents. Rents
can be thought of as direct appropriation (stealing) of tax revenues by government
officials, but also as favors paid to special interests such as public employees or
“friends” of the government, often identified along ethnic or religious lines, and
so on. Voters can replace a government that abuses its powers, but in equilibrium
they generally cannot push rents all the way to zero. This agency problem inter-
acts with lack of information: Voters observe the state of the economy, but they
cannot observe government borrowing, at least not at the margin; for instance,
the government can accumulate hidden off-balance-sheet liabilities. Hence, when
voters see the economy booming, they demand higher utility for themselves (in
the form of lower taxes or better public goods), in a way that resembles the “starve
the Leviathan” argument. This forces the government to impart a procyclical bias
to fiscal policy, and to borrow too much. Thus, procyclical and myopic fiscal
policy (i.e., an increase in government spending during booms and excessive
government borrowing) arises from voters’ demands. But voters do not demand
irrational policies: Through a reelection constraint on the government, they obtain
a second-best solution to an agency problem in an environment of corruption and
imperfect information. Formally, the model extends to a dynamic environment
with public debt a model of moral hazard and political accountability originally
formulated by Barro (1973) and Ferejohn (1986) and adapted to public finance
by Persson and Tabellini (2000).
4
We then discuss some features of the data. First, we confirm previous evidence
on the widespread procyclicality of fiscal policy. Second, we show a positive corre-
lation between procyclicality and measures of corruption: More corrupt countries
display a more procyclical fiscal policy. Third, the correlation between corrup-
tion and procyclicality is only or mainly present in democracies, confirming the
theoretical idea that procyclicality emerges because voters try to hold corrupt
governments accountable. Finally, we ask how robust is the correlation between
corruption and procyclicality when also taking into account the evidence on bor-
rowing constraints. This is not easy, because more corrupt governments might
also face more binding credit constraints. As a result, many of the same variables
that influence political corruption are also likely to affect the severity of bor-
rowing constraints—indeed, corruption is highly correlated with credit ratings in
the data. (In fact, credit rating agencies may look at corruption as one indicator
4. Note the idea that voters induce debt accumulation to discipline governments that they do not
trust is related to Jensen and Meckling (1976). That seminal contribution shows that debt financing
(as opposed to external equity financing) can mitigate the agency problem inside the firm. Of course,
the mechanism through which this happens in our political context is different.
Alesina, Campante, and Tabellini Why Is Fiscal Policy Often Procyclical? 1009
of countries’ ability to pursue stable macroeconomic policies.) Nevertheless, we
present some suggestive evidence that political agency problems in democracies,
rather than, or at least in addition to, credit market imperfections, are the under-
lying cause of procyclical fiscal policy.
5
Finally, our political agency explanation
can coexist with other political distortions, and in particular with the voracity
effect. In fact, we present some evidence consistent with this interpretation.
Our idea that political agency can lead to excessive debt accumulation when
voters are uninformed also differs from two other political models of government
borrowing in the literature. The strategic debt argument (Persson and Svensson
1989; Alesina and Tabellini 1990; Tabellini and Alesina 1990) does not rely on
an agency problem: Voters are not uninformed about fiscal policy, and the results
are driven by different preferences among political parties or groups of voters
alternating in government. In the rational budget cycles literature (Rogoff and
Sibert 1988; Rogoff 1990), voters face an adverse selection problem, and this
leads to distorted fiscal policy before the election. The assumption about voters’
information is similar to ours, but here the incentive problem is one of moral
hazard, not adverse selection. Moreover, those papers do not discuss the reaction
of economic policy to external shocks, nor do they allow for a state variable like
government debt.
The article is organized as follows: In Section 2 we lay out the model; in
Section 3 we derive the economic and political equilibrium; Sections 4 and 5
discusses the empirical evidence; and the last section concludes.
2. The Model
2.1. The Economy
Consider a small open economy with an infinite horizon. The private sector con-
sists of a representative consumer that maximizes the present discounted value of
expected utility from private and public consumption:
E
∞
t=0
β
t
[u(c
t
) + h(g
t
)], (1)
where c
t
and g
t
denotes private and public consumption respectively in period t,
E is the expectations operator, and u(·) and h(·) are smooth and strictly concave
increasing functions. For simplicity, we neglect the intertemporal choices of the
5. Satyanath and Subramanian (2004) show empirical evidence that democratic failure explains
macroeconomic instability. However, they focus on the distinction between democracies versus. non-
democracies, whereas we argue that procyclical fiscal policy stems from the interaction of democratic
accountability and political corruption.
1010 Journal of the European Economic Association
private sector and only focus on its political role of controlling the government
agency problem. Thus, we assume that private consumption in each period is
just given by endowment income (y) net of taxes (τ ): c
t
= y
t
(1 − τ
t
). The
model is meaningful only if government debt is non-neutral and there is a role
for countercyclical fiscal policy, and this is the simplest way to get that property.
Income is an i.i.d. random variable, drawn each period from a distribution with
bounded support over [y
, ¯y]. All variables are expressed in per capita terms.
6
Besides spending in “useful” government consumption that provides utility
to the consumer, g
t
, the government can also appropriate rents, r
t
≥ 0, that benefit
the government but not the consumer. In period t the government can issue public
debt, b
t+1
, at a market price β. Government debt is bought by foreign residents and
there is full repayment of debt next period.
7
Thus, we can write the government
budget constraint as
g
t
+ r
t
+ b
t
≤ τ
t
y
t
+ βb
t+1
. (2)
We assume that there is a limit to how much resources a government can
appropriate for its own exclusive benefit: r
t
≤ q
t
. The upper bound q
t
denotes
what the government can steal from the public coffers without ending up in jail.
We consider two alternative assumptions about q
t
. In the simplest case, it is
a linear and increasing function of current per capita income: q
t
=¯q + ρy
t
,
ρ>0. Thus, as the tax base rises, the government has more opportunities to grab
rents. Alternatively, we assume that the upper bound is a decreasing and concave
function of public debt outstanding: q
t
= Q(b
t
), with Q
b
< 0, Q
bb
< 0. Thus,
if the previous legislature accumulated a large amount of government debt, there
is less room to steal today. As discussed in the next section, debt is only observed
by the public at large in the subsequent period, when it has to be repaid. Thus,
this second assumption says that, if the government accumulated large liabilities
in the previous legislature, it is under more careful scrutiny today, both from the
domestic voters and international organizations, and as a result the upper bound
on rents is more severe. As we shall see, the assumption that there is an upper
bound on rents plays a role even if this constraint is not binding in equilibrium,
because it determines the strength of out-of-equilibrium threats. But the policy
response to income shocks is similar irrespective of whether the upper-bound Q
t
does or does not depend on government debt outstanding.
Finally, we assume that government debt can be issued only up to a maximum
amount
¯
b. Up to this amount, debt is always repaid in full and there is no default
risk nor any credit market imperfection. This upper limit on government debt
6. Alternative but more complicated assumptions would be to allow the consumer to borrow or
lend in an economy with tax distortions, or to model explicitly a liquidity constraint on private
consumption.
7. The assumption of a small open economy is appropriate for our empirical work, in which we
consider this kind of country. Without default risk there is no risk premium, but in the empirical
analysis we allow for the effects of risk premia on government-issued liabilities.
