good, bad or ugly? on the effects of fiscal rules with creative accounting

www.elsevier.com/locate/econbase

Journal of Public Economics 88 (2003) 377–394

Good, bad or ugly? On the effects of fiscal rules with

creative accounting

Gian Maria Milesi-Ferretti*

International Monetary Fund and CEPR, Room 9-212D, Washington DC 20431, USA

Received 21 September 2000; received in revised form 11 March 2002; accepted 13 March 2002

Abstract

Do fiscal rules lead to fiscal adjustment, or do they encourage the use of ‘creative accounting’?

This question is studied with a model in which fiscal rules are imposed on ‘measured’ fiscal

variables, which can differ from ‘true’ variables because there is a margin for creative accounting.

The probability of detecting creative accounting depends on its size and the transparency of the

budget. The model studies the effects on fiscal policy of a budget rule, separating structural from

cyclical effects, and examines how these effects depend on the underlying fiscal distortion and on the

degree of transparency of the budget.

D 2002 Elsevier B.V. All rights reserved.

JEL classification: E62; H61; H62

Keywords: Fiscal rules; Budget deficits; Creative accounting; Transparency

1. Introduction

Fiscal rules have been at the center of public attention and of the policy debate for

several years. During the second part of the 1990s, respect of the so-called Maastricht

criteria on fiscal deficits and government debt in European Union countries was a central

issue in the run-up to the single currency. In 1997, the proposal for a balanced budget

amendment in the United States failed to win Senate approval by a single vote. More

recently, the numerical budget deficit targets for euro-area countries set forth in Stability

Programs have been under heated discussion because of the difficulty in meeting them

0047-2727/$ - see front matter D 2002 Elsevier B.V. All rights reserved.

doi:10.1016/S0047-2727(02)00076-2

* Tel.: +1-202-623-7441; fax: +1-202-589-7441.

E-mail address: [email protected] (G.M. Milesi-Ferretti).

G.M. Milesi-Ferretti / Journal of Public Economics 88 (2003) 377–394378

during an unanticipated economic slowdown. Constraints on fiscal policy are also

important for developing countries: for example, numerical fiscal targets often play a

pivotal role in the context of conditional lending programs by the International Monetary

Fund.

The arguments in support of legal or regulatory restraints on the ability of governments

to choose their levels of taxation and spending are based on the existence of ‘biases’

inherent in the political system or in the budget decision-making process, that make

discretionary fiscal policy deviate from what is desired by society as a whole (Alesina and

Perotti, 1995). The imposition of numerical rules on budget deficits or public spending is

viewed as one possible way to reduce or eliminate this bias. The shortcomings of such

rules are equally well known. Rules that are not contingent on the macroeconomic cycle

can force governments to tighten fiscal policy during cyclical downturns, thus exacerbat-

ing macroeconomic fluctuations—a concern often expressed in European Union countries.

Furthermore, the imposition of numerical rules may encourage the use of dubious

accounting practices, thereby reducing the degree of transparency in the government

budget. These concerns have gained strength with the use of ‘creative accounting’ by a

number of European countries in order to facilitate meeting the budget deficit ceiling

established in the Maastricht Treaty.

This paper contributes to the literature by exploring how desirability and effectiveness

of fiscal rules depend on the scope for ‘creative accounting’ in the government budget, a

subject on which there is much policy discussion, but no formal modeling. Inevitably,

budgetary rules are imposed on ‘measured’ fiscal aggregates (deficit, public debt, or

government spending), which can differ from the economically meaningful ones (which

matter for government solvency) because of ‘window dressing.’ Numerous studies, such

as Blejer and Cheasty (1991), document pitfalls and shortcomings of traditional indicators

of fiscal policy. Window-dressing measures may imply costs for the government: for

example, they may be deemed inadmissible for the attainment of the required budget

balance, they can imply reputation costs, or cause distortions and economic costs. The

model focuses primarily on costs of the first type (although it can readily incorporate other

types of cost) and assumes that these are likely to depend on the ability of the public to

monitor the government’s budgetary actions (itself a function of the degree of transparency

of the budget), and on the size of the window-dressing measures. Based on these

observations, the model studies how the effects of budget rules on structural and cyclical

fiscal policy depend on the margin for creative accounting. In contrast, the existing

theoretical literature always assumes that fiscal rules are fully binding.1

The title of the paper refers to three possible effects of fiscal rules. Rules can lead to

good outcomes, by inducing profligate governments to engage in virtuous fiscal

behavior. Or they can lead to bad outcomes, as they hinder the use of counter cyclical

1 Several recent studies of fiscal rules examine their potential role in a monetary union (Chari and Kehoe,

1997; Beetsma and Bovenberg, 1999; Beetsma and Uhlig, 1999). Corsetti and Roubini (1997) highlight the trade-

off between deficit bias and margin for stabilization inherent in the choice of fiscal rules; Schmitt-Grohe and

Uribe (1997) show that in a standard neoclassical model a balanced-budget rule can lead to multiplicity of

equilibria; and Peletier et al. (1999) show that a fiscal rule, imposed to address a deficit bias problem, may lead to

a suboptimally low level of public investment.

G.M. Milesi-Ferretti / Journal of Public Economics 88 (2003) 377–394 379

fiscal policy and hamper the operation of automatic stabilizers. Or, finally, they can lead

to ugly outcomes: namely, a lot of creative accounting, but little overall effect on fiscal

policy.

In the context of the model in this paper, a measure implying an improvement in the

fiscal balance is considered to be creative accounting if it does not imply an improvement

in the intertemporal budgetary position of the government sector at large (an increase in

the government’s net worth).2 Empirical examples abound. Easterly (1999) argues that

fiscal adjustment in a number of developing countries with World Bank and IMF

programs relied heavily on a decumulation of government assets (mainly through

reductions in public investment and in expenditure on operations and maintenance),

implying that the reduction in government liabilities did not necessarily correspond to an

increase in the government’s net worth. Bunch (1991) presents evidence that US states

with constitutional debt limits use public authorities to circumvent borrowing restrictions,

while von Hagen (1991) and Kiewiet and Szakaly (1996) find that constitutional

limitations pertaining only to guaranteed state debt do not meaningfully affect the total

amount of debt issued by state and local public authorities. A number of well-known

examples which occurred in the run-up to EMU are detailed in the working paper version

of this paper (Milesi-Ferretti, 2001).

The main results of the paper can be summarized as follows. Optimal fiscal rules may

be tighter in the presence of creative accounting, especially if violation of the rule entails

costs for the government but not for its citizens (Section 2). However, if the costs of

creative accounting are ‘large’ even for ‘small’ amounts of window dressing, less

restrictive rules may be met de facto, while tighter rules may induce creative accounting,

and not only fiscal adjustment (Section 3). In the presence of business-cycle fluctuations,

the margin for creative accounting implies that a government can let automatic

stabilizers operate, at least partially, even when constrained by a nominal budget rule

(Section 4).

2. The model

The two-period, reduced-form model borrows from von Hagen and Harden (1996).

The government chooses taxes and spending so as to minimize a convex loss function

which is increasing in the level of taxation and in deviations of spending from its desired

level. The distortion in the conduct of fiscal policy is introduced by assuming that there is

a positive probability that the government discounts the future more heavily than

‘society.’ Fiscal rules can be imposed on the ‘measured’ fiscal balance, which can differ

from the actual (economically meaningful) one, thus generating a margin for creative

accounting. If the public detects that the rule is met de iure but not de facto the

government has to pay the penalty for not meeting the rule, thus implying a disincentive

2 Creative accounting can actually imply a worsening of public accounts; for example, cuts in transfers to

local governments can force them to borrow at rates that are higher than those on central government debt.


to using creative accounting. The probability of detection of creative accounting is

assumed to be increasing in the degree of budget transparency. Formally, the government

chooses the path of taxes (T1; T2) and government spending (G1;G2) so as to minimize the

following quadratic loss function:

Lg ¼ a

2G1 � G*ð Þ2þ 1

2T 21 þ bg a

2G2 � G*ð Þ2þ 1

2T22

� �;

a > 0; 0 < bgV1 ð1Þ

where G* is the desired level of spending (when taxes are not distortionary), subject to the

intertemporal budget constraint:

G1 � T1 þ1

RG2 � T2ð Þ ¼ 0 ð2Þ

where R is the interest factor, taken as given. The preferences of private agents are also

given by (1), with a discount rate equal to 1/R. This loss function can be derived from a

simple endowment model in which consumers’ utility is linear in private consumption and

quadratic in government spending.

The government type g is picked by nature from a binomial distribution and is

unobservable by private agents. The two possible government types differ only in their

discount rate: a social planner (g ¼ p) has the same preferences as private agents (bpR ¼ 1),

while a myopic government (g ¼ m) discounts the future at a rate higher than the rate of

interest; that is, bm ¼ b < 1=R . For example, this higher discount rate can reflect the

possibility that the current policy maker will not be in power in the second period. The

probability that the government is myopic is q<1.

The policies chosen by a myopic government are indicated with the superscript m,

those chosen by a social planner with the superscript p, those undertaken under discretion

with the superscript d and those undertaken under a rule with the superscript r. Given the

assumption that the optimal level of spending G* is the same in both periods, the optimal

policy from the point of view of the public and of the social planner is clearly a balanced

budget:

Gpd1 ¼ T

pd1 ¼ a

1þ aG* ð3Þ

The higher discount rate of the myopic government implies that it will run a deficit in the

first period:

Gmd1 � Tmd

1 ¼ Dmd ¼ 1� bR1þ bR2

G* ð4Þ

The deficit bias is a function of the difference between the discount rate and the rate of

interest.


The fact that the government can be myopic implies that, on average, the economy will

run a budget deficit, thus creating a rationale for fiscal rules. The basic assumptions are the

following:

Assumption 1a. A rule specifies that the budget deficit cannot exceed the level D* .

Violation of the rule, if detected, implies a cost for the government equal to K:

Clearly, for the rule to be binding it must be the case that D*< Dmd. We later consider the

case in which the ‘punishment’ for the detection of a violation of the rule is borne by the

public as well, as would be the case, for example, with a financial penalty (as in Beetsma

and Uhlig, 1999) or with exclusion of the country from a common currency. The next

section discusses the case in which the rule is chosen by an outside agent (such as a super-

national authority).

Assumption 2. The fiscal rule D* is imposed on the ‘measured’ fiscal position #, which

can differ from the actual one (D ) because the government has a margin for creative

accounting.

Assumption 3. The public cannot observe total taxes and spending at the end of period 1.

Assumption 4a. Creative accounting can be detected by the public with probability p,

which is increasing in the degree of transparency of the budget gVand in the size of

creative accounting D� D:

p ¼cVðD� DÞ2 if cVðD� DÞ2 < 1

1 if cVðD� DÞ2z1

ð5Þ

If creative accounting is detected and D > D* , the government bears the cost K of

violating the rule.

Assumption 3 is clearly necessary to introduce ex-post uncertainty about the conduct of

fiscal policy. Assumption 4a simply states that the government incurs some risks in

meeting the rule with creative accounting: if budgetary tricks come to light (with

probability p) and they entail a violation of the rule, the penalty for not meeting the rule

is paid.3 The parameter cV proxies how difficult it is for the public to ‘pierce the veil’ of

budgetary accounts and infer the true fiscal stance, and is henceforth referred to as the

degree of transparency of the budget.4 This is taken to be a structural feature of the

3 In practice, some creative accounting measures can be ‘disallowed’ by the enforcer of the rule. For

example, in 1997 Eurostat did not allow to count as deficit reduction the proceeds (0.2 percent of GDP) from

taxation of profits resulting from a ‘gold sale’ of the Ufficio Italiano Cambi, a public body related to the Bank of

Italy, to the Bank of Italy itself. Nevertheless, Italy’s deficit remained below the 3 percent Maastricht ceiling.4 Note that under the optimal rule some creative accounting always takes place, but the penalty can only be

imposed if creative accounting is detected.


economy, which cannot be altered by the government.5 Clearly if the government meets

the rule de facto (DVD*), it will choose to do no creative accounting. Since the marginal

cost of creative accounting is zero when D ¼ D , Assumption 4a implies that the

government will always choose to meet the rule, at least de iure, by doing some creative

accounting, because it will pay the cost K with a probability below unity.6 For simplicity,

in the remainder of this paper we shall assume that the configuration of parameters is such

that cVðD� DÞ2 < 1, thus ensuring an interior solution.

The assumptions concerning creative accounting could be altered by assuming either

that policymakers that are caught using accounting tricks bear some other cost unrelated to

K (for example, the Finance Minister loses his/her job), or that creative accounting

imposes distortionary costs on the economy. In both these cases, the government would

face a meaningful choice between meeting the rule only de iure and choosing its

discretionary policy without doing any creative accounting. As shown below, it is easy

to reconcile the model’s formulation with either alternative interpretation. Note also that in

the setup of the model there is no scope for signaling, because the government is not

punished or rewarded with re-election if its preferences become known.7

Given (5), the government’s loss function, conditional on the measured deficit meeting

the rule, becomes:

Lg ¼ a

2G1 � G*ð Þ2þ 1

2T 21 þ IpK þ bg a

2G2 � G*ð Þ2þ 1

2T22

� �ð6Þ

where the indicator function I ¼ 1 when D > D* and 0 otherwise. The loss function (6) is

consistent with the existence of additional ‘reputation’ costs for the government if the rule

is violated, as long as these costs take a quadratic form. If instead creative accounting were

to induce economic distortions, it would be necessary to amend the private agents’ loss

function as well: this is done in Section 3.

2.1. The behavior of the government

From Assumptions 1a, 2, 3, 4a, and Eq. (6), it follows that a myopic government will

set D ¼ D*. Since by assumption D* is smaller than the deficit Dmd a myopic government

would choose in the absence of a rule, there is no reason to overshoot D*.8 The actual

deficit will be determined by the choice of taxes and spending that minimizes (6), with

D ¼ D*, subject to the budget constraint (2). Inspection of (6), taking into account (5),

shows that this setup is identical to one where the fiscal rule has an ‘escape clause,’ with

7 See Rogoff and Sibert (1988) and Rogoff (1990) for models with signaling through fiscal policy.

6 In Section 3 we explore the implications of an alternative specification of the probability of detection,

which can result in no creative accounting taking place in equilibrium.

5 von Hagen and Harden (1994) discuss differences in the transparency of the budget in EU countries, and

Alesina et al. (1999) in Latin American countries. Alesina and Cukierman (1990) show that a government can

rationally choose ‘ambiguous’ procedures in order to make it more difficult for voters to infer its true type.

8 Note that an incentive to overshoot the target may arise if we allowed some scope for signaling, for

example by assuming that the public is more likely to re-elect a ‘social planner.’


the cost of deviating from the rule proportional to the degree of transparency and

increasing in the size of the deviation. The optimal deficit will then be given by:

Dmr ¼ að1� bRÞa 1þ bR2ð Þ þ cð1þ aÞ G* þ cð1þ aÞ

a 1þ bR2ð Þ þ cð1þ aÞ D* ð7Þ

where c ¼ cVK . It can be easily verified that, given a rule D* < Dmd , the deficit Dmr

chosen by a myopic government is such that Dmd > Dmr > D*. That is, when faced with a

budget rule that it is costly to violate, the government will do some fiscal adjustment and

some ‘window dressing’ in order to meet the rule. We are now able to state the first result.

Proposition 1. For a given cost of violating the rule, the size of fiscal adjustment induced

by the rule is increasing in the degree of transparency of the budget cV.

Proof. The proportion of ‘measured’ fiscal adjustment that reflects a change in the actual

deficit is given by ðDmd � DmrÞ=ðDmd � D*Þ ¼ ½cð1þ aÞ�=½að1þ R2Þ þ cð1þ aÞ�;which is increasing in cV¼ c=K.

As budget transparency increases, the amount of actual adjustment increases and the

amount of creative accounting correspondingly decreases, because the rule becomes, in

effect, more binding.

When the government is instead a social planner, its optimal policy is very simple. If

the rule does not require a budget surplus, the budget will be balanced, with taxes and

spending determined by (3). If the rule requires a surplus, a social planner will also resort

to creative accounting so as to meet the rule (see Eq. (A.2) in Appendix A).

2.2. The optimal fiscal rule

In the absence of a margin for creative accounting, a balanced-budget rule would be

optimal for society as a whole. Such a rule would be binding only for a myopic

government, and the resulting policy choice would be the one preferred by private agents.

In the presence of a margin for creative accounting, however, a balanced budget rule is not

optimal anymore, because it will give rise, on average, to a fiscal deficit (from Eq. (7),

Dmr > 0 when D* ¼ 0). Given that violation of the rule imposes no direct costs on the

public, the optimal fiscal rule will be ‘tighter’ than the desired fiscal outcome, because

there is a positive probability that the government is myopic and would therefore resort to

cosmetic changes in the deficit, in addition to fiscal adjustment:

Proposition 2. Under Assumptions 1a, 2, 3, and 4a the optimal rule D** requires a budget

surplus. The size of the surplus is positively related to the size of the deficit bias and to the

probability the government is myopic, and negatively related to the degree of transparency.

Proof. See Appendix A.

Proposition (2) states that the optimal rule will ‘overshoot’ the desired fiscal balance. If

the probability that the government is a social planner increases, there is less need for


overshooting. The same happens if the budget is more transparent, since the rule is more

binding (creative accounting is easier to detect). Of course, when the rule requires a

budget surplus a social planner would indeed run a surplus and do some creative

accounting. However, the cost of this deviation from a balanced budget is compensated

by the reduction in the deficit bias for the myopic government. In expected value terms,

the rule described in Proposition 2 implies an actual budget surplus, because of the ‘risk-

aversion’ of private agents implicit in the concave utility function (see Appendix A for

details).

Even if derived from a very stylized model, these results provide interesting insights

concerning the policy debate over the role of fiscal rules and budget transparency. For

example, the model provides a simple justification for imposing a stricter rule on

countries with less transparent budgets and a history of fiscal profligacy (a high q or a

high deficit bias). Conversely, it suggests that adoption of more transparent budgets may

rationally lead to less stringent conditionality requirements in lending by international

organizations.

3. Costs of creative accounting: alternative specifications

In this section we examine first the case in which violation of the rule imposes costs on

the public as well, and then turn to an alternative specification of the probability of

detection which allow for a positive marginal cost of creative accounting even in the

neighborhood of zero creative accounting.

3.1. The public bears the cost of not meeting the rule

We replace Assumption 1a with the following:

Assumption 1b. A rule violation, if detected, imposes costs on both the government and

the public.

This could represent the case of an external enforcer of the rule: if a country does not meet

the fiscal rule, it has to pay a lump-sum financial penalty or bear some other form of cost

(such as the exclusion from monetary union, or the interruption of a program of external

financial assistance). We choose a simple specification of the way in which these costs

enter the private agents’ loss function:

Lg ¼ a

2G1 � G*ð Þ2þ 1

2T21 þ mc

2ðD� D*Þ2

�

þ 1

R

a2

G2 � G*ð Þ2þ 1

2T 22

� ��; mV1 ð8Þ

where we replaced D with D*. The parameter m measures the fraction of the cost K borne

by private agents. Alternatively, we could think that private agents and the government

bear the same cost of violating the rule, and that m captures the fraction of such costs


internalized by an external enforcer.9 To incorporate in the model the costs induced on

other countries by fiscal policy spillovers (for example in the form of higher interest rates,

an argument frequently mentioned to justify the Maastricht criteria), the enforcer’s loss

function can be amended by substituting the term in m with a term capturing deviations of

domestic fiscal policy from its optimal level (here a balanced budget).

If the rule is set by an external enforcer that maximizes private agents’ utility without

internalizing the costs of violating the rule, results would be identical to those of

Proposition 2. If instead the rule is set internalizing the costs that a violation imposes

on private agents, D* would be chosen so as to minimize (8), subject to ðG1;G2; T1; T2Þbeing chosen optimally by the government:

Proposition 3. Under Assumptions 1b, 2, 3 and 4a, the optimal rule requires a budget

surplus which is smaller than under Assumption 1a (Proposition 1). The surplus is

increasing in the deficit bias 1� bR and in the probability q that the government is

myopic, and decreasing in m and the degree of transparency cV.


Proposition 3 states that the result regarding ‘overshooting’ of the budget rule holds even

when creative accounting imposes costs on society at large. When m =1 the optimal rule

implies, in expected value terms, a budget deficit (see Appendix A). Even though the

public (and/or the external enforcer) would like a balanced budget, a tighter rule would

encourage creative accounting and a higher risk of violation, with implied costs that are

increasing in m. Clearly, a higher degree of transparency reduces the incentive to use

creative accounting and therefore reduces the expected deficit. Note finally that if the

budget rule is set by an external enforcer, who does not fully internalize the penalty K, it

would be considered too tight by domestic residents, who have to pay the full penalty if

the creative accounting is detected.

3.2. Positive cost of creative accounting at the margin

We now modify Assumption 4a by allowing for a nonzero marginal cost of creative

accounting even when actual and measured deficit are very close. Let the probability of

detection be:

p ¼c VðD� DÞ2 þ x VI if c VðD� DÞ2 þ x VI < 1

1 if c VðD� DÞ2 þ x VI z1

ð9Þ

with x V> 0 and I ¼ 1 ifD < D and zero otherwise. In this case, even an infinitesimal

amount of creative accounting raises the probability of detection by a discrete amount.

Consider a rule D* and define x ¼ x VK . If a myopic government’s policy choice

9 These costs could also be interpreted as distortions imposed by creative accounting. Assume that these

distortions imply quadratic costs (with weight k in the loss function) and let c* ¼ k þ c. In that case the loss

function would feature the parameter c* (instead of c) and the parameter m would be equal to k=c*.


involves creative accounting, the value of its loss function under this amended probability

of detection will be:

LmðGmr

i ðD*Þ; T mri ðD*Þ;xÞ ¼ L

mðGmri ; T

mri Þþx ð10Þ

where Lm

is defined in (6). Here and below we omit for simplicity to specify the

dependence of G and T on the rule D*.

Define ðGmr

i ; Tmr

i Þ as the policy of the myopic government if it satisfies the rule de

facto, so that Gmr

1 � Tmr

1 ¼ D*, and let Lm(Gi

mr, Ti

mr) be the corresponding value of its

loss function. We can now state the following proposition:

Proposition 4. For every level of x >0 there exists a budget rule D(x), BD/Bx <0, such

that a myopic government would be indifferent between meeting the budget rule de facto or

choosing the ‘optimal’ positive amount of creative accounting implied by (7). For any rule

D* such that DmdzD*zD, a myopic government would choose to meet the rule de facto.


Proposition 4 simply states that creative accounting is not worthwhile for a myopic

government unless the rule is ‘sufficiently far’ from its preferences, where the ‘distance’ is

given by x. It also implies that a small tightening of the fiscal rule beyond D would result

in less fiscal adjustment, since a myopic government would find it convenient to start

engaging in creative accounting.

What is the optimal rule from the point of view of the public? Whenever DV0, the

public can obtain its first-best outcome by choosing a balanced-budget rule, since by

Proposition 4 even a myopic government would find it optimal to meet such a rule de

facto. If D>0, the public has two choices: a rule D, in which case a myopic government

would choose ðGmr

i ðDÞ; T mr

i ðDÞÞwhile a social planner would balance the budget, or, as inthe case of Section 2 a rule D** (see Proposition 2 and Eq. (A.4) in the Appendix A), in

which case creative accounting takes place.10 The optimal rule is characterized in the

following proposition:

Proposition 5. Under Assumptions (1a), (2), (3) and Eq. (9), the optimal budget rule is

given by

D**if D > 0 and E LSW ðD**Þ� �

< E LSW ðDÞ� �

D if D > 0 and E LSW ðD**Þ� �

> E LSW ðDÞ� �

0 if DV0

10 The policy choices of the two government types would be determined by Eqs. (A.1) and (A.2) (see

Appendix A).



In summary, when even a minimal degree of creative accounting raises the probability of

detection by a first-order amount and the fiscal rule is not very restrictive, it is possible that

no creative accounting will occur in equilibrium. In choosing the optimal rule the public is

going to face a trade-off between a less restrictive rule, always met de facto, and a tighter

rule which induces creative accounting.

4. Budget rules with output shocks

In the simple model examined so far, there is no stabilization role for fiscal policy;

therefore, given the deficit bias, a budget rule is clearly optimal. However, a numerical rule

can hamper the flexibility of fiscal policy to respond to stochastic shocks. To address this

question, following Corsetti and Roubini (1997), we return to the baseline case of Section

2 and assume that the actual deficit in period 1 has a stochastic component e, uniformly

distributed on the interval [�a; a].11 This component can reflect the operation of automatic

stabilizers in the budget that respond to fluctuations in economic activity, for given tax and

expenditure decisions.12 The underlying output shock is assumed to be observable by the

government before tax and expenditure decisions are taken, but not when the budget rule is

determined. For simplicity, it is also assumed that no stochastic disturbances occur in

period 2. The government’s intertemporal budget constraint thus becomes:

G1 � T1 � e þ 1

RG2 � T2ð Þ ¼ 0 efU ½�a; a� ð11Þ

where for simplicity we assume that a < ð1� bRÞ G* so that the realization of the shock

is never large enough to fully offset the deficit bias.13 Suppose, for example, that the shock

is positive, and that it implies lower public expenditure (in the amount de) and higher tax

revenue (in the amount ð1� dÞ eÞ. The implicit assumption on the expenditure side is that

the desired level of expenditure G* changes by de as well (if output is high and

unemployment benefits fall, so does the optimal level of spending). On the tax side, T

should be interpreted as the tax rate, rather than tax revenue—higher tax revenue when

output is high does not imply more tax distortions. Therefore the loss function remains

unchanged, except for the fact that the (actual) deficit now includes the term e.We first determine the budget response to the shock desired by private agents, and

compare it with the one a myopic government would choose in the absence of a rule. We

13 If this assumption is not satisfied, the algebra is considerably more complex, because of the non-linearity

in the government’s expected policy choice with respect to the rule.

12 In general, the optimal response of taxes and spending to a shock will depend on the nature of the shock.

The purpose of this extension is simply to illustrate the implications of rules for the automatic and discretionary

response of the budget to cyclical fluctuations in activity.

11 Corsetti and Roubini (1997) present a model in which the deficit bias is generated by a disagreement

between parties over future spending priorities, as in Alesina and Tabellini (1990). Given electoral uncertainty, the

government discounts future spending more than private agents. In equilibrium, this can give rise to a deficit bias,

with taxes being too low and spending too high in the first period.


then examine how the imposition of a budget rule modifies the government’s response to

the shock. In this section we focus mainly on the cyclical part of the budget, which has two

components: the actual shock plus the policy response to it. The analysis of the structural

part of the budget is analogous to the one in Section 2.

The public’s desired budget deficit, which reflects smoothing of the shock, is given by:

Dpd ¼ � e1þ R

ð12Þ

A positive shock would hence lead to a budget surplus and a negative shock to a deficit.

The ‘discretionary’ deficit choice of a myopic government is obtained by minimizing (1)

subject to (11):

Dmd ¼ 1� bR1þ bR2

G* � 1

1þ bR2e ð13Þ

An interesting result emerges when comparing (12) and (13). Namely, a myopic govern-

ment’s budget responds ‘too much’ to the shock with respect to what is optimal; because

of the government’s higher discount rate, the response of fiscal policy to the shock does

not achieve sufficient ‘smoothing.’

Analogously to the previous section, consider now the case in which a fiscal rule is

imposed on the first-period budget balance. Throughout this section we assume that this

rule is not state-contingent, in line with the way fiscal rules are often designed in

practice.14 Clearly, a state-contingent rule would not hamper the automatic response of

the budget to shocks. The government chooses spending, taxes and ‘window-dressing’ so

as to minimize the loss function (6), subject now to the budget constraint (11). Note in

particular that D is now equal to G1 � T1 � e. Appendix A reports the ‘structural’ and

‘cyclical’ components of the chosen levels of taxes and spending (Eqs. (A.1) and (A.10),

respectively). The deficit chosen by a myopic government is given by:

Gmr1 � T

mr

1 �e ¼ Dmr¼ að1� bRÞ G* þ cð1þ aÞ D* � ae

a 1þ bR2ð Þ þ cð1þ aÞð14Þ

The ‘structural’ part of the budget deficit is clearly the same as in (7), and the same goes

for the levels of spending and taxation. The rule implies that the budget deficit can change

by less in response to cyclical disturbances with respect to the case in which there is no

rule. Note, however, that this response is not entirely prevented, because of the margin for

creative accounting.

According to conventional wisdom, numerical rules on the budget deficit impose costs

because they hinder a cyclical response to the budget. Does this imply that, from the point of

view of cyclical ‘smoothing,’ less transparency implies a more desirable cyclical response

of the budget? This is not necessarily the case, as the following proposition clarifies.

14 For example, the budget targets in euro-area countries’ Stability Programs are defined in nominal terms,

even though the Stability and Growth Pact emphasizes a medium-term objective of a balanced budget over the

cycle. Switzerland recently approved a fiscal rule based on the cyclically-adjusted budget.


Proposition 6. At low levels of transparency, an increase in transparency implies a more

desirable cyclical response of the budget to shocks when the government is myopic. At

high levels of transparency, an increase in transparency implies a less desirable response

of the budget to shocks.

Proof. Compare the cyclical components of Eqs. (12), (13) and (14).

The budget response to the shock is decreasing in the degree of transparency c, but inthe absence of a rule the budget would respond ‘too much’ to the shock (compare (12)

with (13) above). Therefore, at low levels of c, more transparency implies a response of

the budget to shocks which is closer to agents’ preferences. This implies that, at low

levels of transparency, a rule can moderate a deficit bias, and also ensure a more

appropriate response to shocks. As the degree of transparency rises and creative

accounting becomes more costly, the margins for smoothing are eroded. The result in

Proposition 6 reflects the fact that if discretionary fiscal policy is subject to a bias there

is no reason to presume that the government would choose an ‘optimal’ response to

cyclical shocks.15

The importance of the result of Proposition 6 depends on the likelihood that the

government is indeed myopic. Since the discretionary deficit choice of a social planner

would coincide with the choice of the public, a binding budget rule necessarily implies a

less desirable response of the budget to shocks (compare (12) with (14), setting b ¼ 1=R).While it is still true that at low levels of c an increase in transparency raises welfare (in

expected value terms, the above-mentioned negative effect would be offset by the more

desirable response obtained in the case the government is myopic) the intensity of the

effect and the range of values of c over which this occurs depend on the value of q. This

reflects the fact that as q decreases the logic for the imposition of a budget rule is

weakened, since discretionary fiscal policy gets closer and closer to the one desired by

private agents.

The working paper version of this paper compares the effects of a budget rule with

those of a spending rule, which leaves more margin for a budgetary response to shocks

through taxation. In that context, which rule is preferable will depend on the relative

weight of ‘structural’ and ‘cyclical’ factors. In order to improve the responsiveness of

the budget to cyclical shocks, a rule could establish that the budget has to be balanced

over the economic cycle, but not year by year. Since the intertemporal budget

constraint requires a balanced budget on average over two periods, analyzing such a

rule would require extending the model to more than two periods. This type of rule

allows a better response to economic shocks than a period-by-period rule. However, the

existence of a deficit bias implies that the government will still initially run budget

deficits. Furthermore, this type of rule could raise problems when there is a change in

15 A more extreme example of this point can be found in Tornell and Lane (1998); the authors show in the

presence of a ‘common pool’ problem in fiscal policy conduct a ‘voracity’ effect can induce spending to rise more

than proportionately following of a temporary positive shock. Clearly the ‘cyclical advantage’ provided by lack of

transparency depends on the fact that the rule is not state-contingent.


government, because one party can leave a legacy of deficits that must be offset by its

successor.

5. Concluding remarks

This paper has studied the effects of fiscal rules when the government has a

margin for ‘creative accounting’. In addition to the trade-off between deficit bias and

margin for cyclical stabilization, it has emphasized the existence of a second trade-off

between (costly) window-dressing and real fiscal adjustment, relating it to the degree

of transparency of the budget. Ceteris paribus, a rule imposed when the budget is not

transparent yields more creative accounting and less fiscal adjustment. For a given

level of budget transparency, the chosen rule will be tighter the higher the probability

that the government is fiscally profligate. The existence of a margin for creative

accounting also implies that the budget retains some ability to respond to cyclical

shocks even in the presence of a numerical budget rule. It was shown in Proposition

6 that the effects of increased transparency on the cyclical response of the budget to

shocks are ambiguous.

This simple model provides reasonable implications of the effects of rules on fiscal

policy decisions, as well as a rationale for the imposition of tighter rules (a stricter

interpretation of rules) on countries that have a history of fiscal profligacy and less

transparent budgets. Several extensions are possible. For example, it would be

interesting to explore the incentives for creative accounting over longer horizons,

allowing for some accumulation of nonrecorded liabilities over time. The model could

also be enriched by endogenizing transparency and by a more realistic information

structure. In the current setting, a myopic government’s choice of creative accounting

is merely dictated by the cost of violating the rule, with signaling playing no role. In

a more complex setting, in which the government is up for re-election and

government preferences matter in the second period as well, a ‘good’ government

would have an incentive to signal its type (as, for example, in Rogoff, 1990). Unlike

in that paper, however, here the signaling would imply a larger surplus. Finally,

empirical evidence could shed light on whether the size of creative accounting (as

measured, for example, by the difference between budget deficits and the change in

public debt) is higher in the presence of fiscal rules and whether it is related to

indices of budget transparency.

Acknowledgements

I am grateful to Alberto Alesina, Zaohui Chen, Tito Cordella, Gian Paolo Dell’Ariccia,

Pietro Garibaldi, Nissan Liviatan, Torsten Persson, Alessandro Prati, Jorge Roldos, the co-

editor Lans Bovenberg, three anonymous referees, and seminar participants at the IMF, the

CEPR ESSIM conference in Vouliagmeni and the Political Economy of European

Integration meeting in Berkeley for useful suggestions. I am responsible for all remaining

cases of ‘creative algebra.’ The opinions expressed are mine and do not necessarily reflect

those of the International Monetary Fund.


Appendix A

Proof of Proposition 2. The optimal tax and expenditure policy from the point of view of

a myopic government under a budget rule D* is given by:

ðGmr1 ; T

mr1 Þ ¼

a1þ a

G* 1þ 1� bRC

� �þ c

CD*;

�a

1þ aG* 1� að1� bRÞ

C

� �� ac

CD*

�

ðGmr2 ; T

mr2 Þ ¼

a1þ a

G* 1� Rð1� bRÞC

� �� cR

CD*;

�

a1þ a

G* 1þ aRð1� bRÞC

� �þ acR

CD*

�ðA:1Þ

where C ¼ að1þ bR2Þ þ cð1þ aÞ. For a social planner, the optimal policy in response to

a rule requiring a surplus (D� < 0) is given by:

ðG pr

1; T

pr

1 Þ¼a

1þ aG*þ c

að1þ RÞ þ cð1þ aÞ D*;

�

a1þ a

G* � acað1þ RÞ þ cð1þ aÞ D*

�

ðG pr

2 ;Tpr

2 Þ ¼a

1þ aG* � cR

að1þ RÞ þ cð1þ aÞ D*;

�

a1þ a

G* þ acRað1þ RÞ þ cð1þ aÞ D*

�ðA:2Þ

The public chooses the optimal rule D* to minimize its social welfare (SW) loss function:

LSW ¼ Ea2

G1 � G*� �2

þ 1

2T21 þ 1

R

a2

G2 � G*� �2

þ 1

2T22

� �� ðA:3Þ

subject to (G1;G2; T1; T2) being chosen optimally by the government (Eqs. (A.1)–(A.2)).

This yields the following expression for the optimal rule:

D** ¼ � aqð1� bRÞ G*cð1þ aÞ qþ ð1� qÞ C2

1

ðA:4Þ

where C1 ¼ ½að1þ bR2Þ þ cð1þ aÞ�=½að1þ RÞ þ cð1þ aÞ� < 1: Using (A.1) and (A.2),

we can find the rule Dbb* that ensures, in expected value terms, a balanced budget:

D*bb ¼ � aqð1� bRÞ G*cð1þ aÞ qþ ð1� qÞ C1½ � ðA:5Þ


It is straightforward to see that the surplus implied by (A.4) is larger than the one implied

by (A.5). Note also that D** is decreasing in q (a higher q implies a larger surplus).

Proof of Proposition 3. The optimal rule D*** is the level of D* that minimizes (8),

subject to ðG1;G2; T1; T2Þ being chosen optimally by the government. This yields the

following expression:

2

�2#

ðA:6Þ

D*** ¼ � aqð1� bRÞ 1þ R� mð1þ bR2Þ½ � G*

cð1þ aÞð1þ RÞ qþ ð1� qÞ C21

þ am 1þ bR2ð Þ2 qþ ð1� qÞ C1

1þ R1þ bR

�"

The rule entails a budget surplus since mV1, proving Proposition 3. As m increases, the

required surplus falls. Plugging this expression into the one for the expected budget

balance (obtained using (A.1) and (A.2)), it is straightforward to prove that for m=1 the

rule implies an expected budget deficit.

Proof of Proposition 4. We first characterize the policies Gmri ðD*Þ; T mr

i ðD*Þ followed by

a myopic government that chooses to meet the budget rule de facto:

ðGmr1 ; T

mr1 Þ ¼

a1þ a

G* þ 1

1þ aD*;

a1þ a

G* � a1þ a

D*� �

ðG mr2 ; T

mr2 Þ ¼

a1þ a

G* � R

1þ aD*;

a1þ a

G* þ aR1þ a

D*� �

Plugging (A.1) and (A.7) into the loss function (2) and using (10), one obtains:

Lm Gmri ðD*Þ; T mr

i ðD*Þ� �

� Lm Gmri ðD*Þ; T mr

i ðD*Þ� �

¼ 2x

� a2

1þ a1þ bR2ð Þ2

a 1þ bR2ð Þ þ cð1þ aÞ ðDmd � D*Þ2 ðA:8Þ

where Dmdis defined in Eq. (4). Given that D*VDmd, this implies that for every x > 0

there exists a deficit rule D ¼ Dmd �ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi2xð1þ aÞ½að1þ bR2Þ þ cð1þ aÞ�

pa 1þ bR2ð Þ , AD=Ax < 0

such that LmðGmri ðDÞ; T mr

i ðDÞÞ ¼ LmðGmri ðDÞ; T mr

i ðDÞÞ þ x. Without loss of generality,

assume that when the above equality holds, the myopic government chooses the policy

ðGmri ðDÞ; T mr

i ðDÞÞ.

ðA:7Þ

Proof of Proposition 5. If DV0 the public can obtain its preferred policy (given by (3))

under either possible government by choosing D* ¼ 0 (see Proposition (4) and Eq. (A.7)).

Consider now the case D > 0. If the public chooses a rule D* < D, its best choice is D**

(Eq. (A.4)). The alternative is a rule D* ¼ D, since any rule D* > D yields higher losses


than D* ¼ D (no creative accounting but higher deficits under a myopic type). The

optimal rule is determined by comparing EðLSW ðD**ÞÞ (with policies determined by (A.1)

for a myopic type and (A.2) for a social planner), and EðLSW ðDÞÞ (with policies

determined by (A.7) for a myopic type and (3) for a social planner). This comparison

yields:

E LSW ðD**Þ� �

> E LSW ðDÞ� �

ZD <að1� bRÞ

ffiffiffiffiffiffiffiffiffiffiffi1� q

p

að1þ RÞ þ cð1þ aÞ½ � qþ ð1� qÞ C21

In words, the optimal rule is D if D is sufficiently close to zero, and D** otherwise.

Budget rule (cyclical response). The public’s desired response to shocks is given by:

Gp1e; T

p1e

� �¼ G

p2e; T

p2e

� �¼ 1

1þ aR

1þ Re;� a

1þ aR

1þ Re

� �ðA:9Þ

so that there is perfect ‘smoothing’ of the shock. The optimal government response is

instead:

Gmr1e ; T

mr1e

� �¼ abR2 þ cð1þ aÞ

ð1þ aÞ a 1þ bR2ð Þ þ cð1þ aÞ½ � e;�aGmr1e

� �

Gmr2e ; T

mr2e

� �¼ aR

abR2 þ cð1þ aÞ Gmr1e ;�

a2RabR2 þ cð1þ aÞ Gmr

1e

� � ðA:10Þ

where the superscript b indicates a budget rule.

References

Alesina, A., Cukierman, A., 1990. The politics of ambiguity. Quarterly Journal of Economics 105 (3), 829–850.

Alesina, A., Hausmann, R., Hommes, R., Stein, E., 1999. Budget institutions and fiscal performance in Latin

America. Journal of Development Economics 59 (2), 253–273.

Alesina, A., Perotti, R., 1995. The political economy of budget deficits. IMF Staff Papers 42 (1), 1–31.

Alesina, A., Tabellini, G., 1990. A positive theory of fiscal deficits and government debt. Review of Economic

Studies 57 (3), 403–414.

Beetsma, R.M.W.J., Bovenberg, A.L., 1999. Does monetary unification lead to excessive debt accumulation?

Journal of Public Economics 74 (3), 299–325.

Beetsma, R.M.W.J., Uhlig, H., 1999. An analysis of the stability pact. Economic Journal 109 (458), 546–571.

Blejer, M.I., Cheasty, A., 1991. The measurement of fiscal deficits: analytical and methodological issues. Journal

of Economic Literature 29 (4), 1644–1678.

Bunch, B.S., 1991. The effect of constitutional debt limits on state governments’ use of public authorities. Public

Choice 68 (1), 57–69.

Chari, V.V., Kehoe, P.J., 1997. On the need for fiscal constraints in a monetary union. Mimeo. Federal Reserve

Bank of Minneapolis, September.

Corsetti, G., Roubini, N., 1997. Politically motivated fiscal deficits: policy issues in closed and open economies.

Economics and Politics 9 (1), 27–54.

Easterly, W., 1999. When is fiscal adjustment an illusion? Economic Policy 28 (1), 57–76.


Kiewiet, D.R., Szakaly, K., 1996. Constitutional limitations on borrowing: analysis of state bonded indebtedness.

Journal of Economics, Law and Organization April, 62–97.

Milesi-Ferretti, G.M., 2001. Good, bad or ugly? On the effects of fiscal rules with creative accounting. CEPR

Discussion Paper 2663, January.

Peletier, B.D., Dur, R.A.J., Swank, O.H., 1999. Voting on the budget deficit: comment. American Economic

Review 89 (5), 1377–1381.

Rogoff, K., 1990. Equilibrium political budget cycles. American Economic Review 80 (1), 21–36.

Rogoff, K., Sibert, A., 1988. Elections and macroeconomic policy cycles. Review of Economic Studies 55 (1),

1–16.

Schmitt-Grohe, S., Uribe, M., 1997. Balanced-budget rules, distortionary taxes, and aggregate instability. Journal

of Political Economy 106 (5), 976–1000.

Tornell, A., Lane, P.R., 1998. Are windfalls a curse? A non-representative agent model of the current account.

Journal of International Economics 44 (1), 83–112.

von Hagen, J., 1991. A note on the empirical effectiveness of formal fiscal restraints. Journal of Public Eco-

nomics 44 (2), 199–210.

von Hagen, J., Harden, I., 1994. National budget processes and fiscal performance. European Economy, Reports

and Studies 3, 311–418.

von Hagen, J., Harden, I., 1996. Budget processes and commitment to fiscal discipline. IMFWorking Paper 96/78,

July.

good, bad or ugly? on the effects of fiscal rules with creative accounting

Documents