doi:10.1016/j.jbankfin.2006.05.011

What do deﬁcits tell us about debt?

Empirical evidence on creative accounting with

ﬁscal rules in the EU

rgen von Hagen

a,1

, Guntram B. Wolﬀ

ZEI, University of Bonn, CEPR, and Indiana University, Walter-Flex-Str. 3, 53113 Bonn, Germany

Deutsche Bundesbank, ZEI-University of Bonn and UCIS-University of Pittsburgh,

Wilhelm-Epstein-Straße 14, 60431 Frankfurt am Main, Germany

Available online 24 July 2006

Abstract

Fiscal rules, such as the excessive deﬁcit procedure and the stability and growth pact (SGP), aim

at constraining government behavior. [Milesi-Ferretti, G., 2003. Good, bad or ugly? On the eﬀects of

ﬁscal rules with creative accounting, Journal of Public Economics, 88, 377–394] develops a model in

which governments circumvent such rules by reverting to creative accounting. The amount of this

depends on the reputation cost for the government and the economic cost of sticking to the rule.

We provide empirical evidence of creative accounting in the European Union. We ﬁnd that the

SGP rules have induced governments to use stock-ﬂow adjustments, a form of creative accounting,

to hide deﬁcits. The tendency to substitute stock-ﬂow adjustments for budget deﬁcits is especially

strong for the cyclical component of the deﬁcit, as in times of recession the cost of reducing the def-

icit is particularly large.

JEL classiﬁcation: E62; H61; H62; H63; H70

Keywords: Fiscal rules; Stock-ﬂow adjustments; Debt-deﬁcit adjustments; Stability and growth pact; Excessive

deﬁcit procedure; ESA 95

doi:10.1016/j.jbankﬁn.2006.05.011

Corresponding author. Tel.: +49 69 9566 3353.

E-mail addresses: [email protected] (J. von Hagen), guntram.wolﬀ@bundesbank.de (G.B. Wolﬀ).

Tel.: +49 228 73 9199.

Journal of Banking & Finance 30 (2006) 3259–3279

www.elsevier.com/locate/jbf

1. Introduction

Fiscal rules aim at constraining the behavior of governments. They are introduced to

reduce rent seeking behavior of politicians, to mitigate common pool pro blems, and,

ultimately, to prevent undesired ﬁscal outcomes (von Hagen, 2002). The European

Economic and Monetary Union (EMU) provides an important example. Governments

in a monetary union have an incentive to run excess ive deﬁcits and accumulate exces-

sive debts. High deﬁcits and debt levels increase the pressure on the central bank to

monetize them and create inﬂation. Anticipating this, the private sector adjusts inﬂation

expectations upwards, resulting in higher nominal interest rates.

But since the

(expected) inﬂation from monetizing a given amount of government debt is spread over

all members of the monetary union, the cost of excessive deﬁcits and debts in terms of

higher inﬂation and interest rates is smaller, and the incent ive for proﬂigate ﬁscal

behavior is larger for each individual government than in the case of a national

currency.

Recognizing this problem, the governments of the EMU member states have adopted a

set of rules to strengthen ﬁscal discipline. These ﬁscal rules feature a limit on the annu al

general government budget deﬁcit of 3% of GDP and a limit on general government debt

of 60% of GDP. In practice, the deﬁcit lim it is considered to be the more important one in

the policy debate.

Fiscal rules necessarily refer to speciﬁc budgetary items and data. Governments can

shift ﬁscal expenditures oﬀ the budget, i.e., revert to creative accounting, to circumvent

such rules. Milesi-Ferretti (2003) analyzes the eﬀect of ﬁscal rules on creative accounting

in a model based on von Hagen and Harden (1995, 1996). In this model, the government

has an incentive to circumvent the rule by hiding ﬁscal policies in less visible positions. The

likelihood of creative accounting decreases in the cost the government has to bear if

the cheating is detected. Furthermore, creative accounting is the more likely, the higher

the economic costs of sticking to the rule are. If strict rules prevent the appropriate

response of ﬁscal policy to (business cycle) shocks, the likelihood of creative accounting

increases in the model.

The optimal design of a ﬁscal rule should take the possibility of

creative accounting into account. As a result, an optimal rule is likely to be stricter in

the presence than in the absence of creative accounting.

A number of authors have investigated the eﬀects of ﬁscal rules. The literature

generally assesses the eﬀect on the ﬁscal aggregate constrained by the rules , and, in a

second step, the eﬀect on other, non-constrained ﬁscal positions. von Hagen (1991)

empirically investigates the eﬀects of ﬁscal restraints on state budgets in the US and

shows that they induce substitution out of restricted into non-restricted debt instruments.

Bunch (1991) and Sbragia (1996) show that debt limits on state or local governments

in the US have led to an increased use of non-constrained public authorities to issue

In the absence of perfect international capital mobility, high debt levels may also lead to higher real interest

rates.

Milesi-Ferretti shows that more transparency of the budget is only desirable at very low levels of budget

transparency, since in this case, governments tend to let the budget ﬂuctuate too much. At high levels of

transparency, a further increase of transparency would hinder the working of automatic stabilizers too much and

is therefore not optimal.

3260 J. von Hagen, G.B. Wolﬀ / Journal of Banking & Finance 30 (2006) 3259–3279

debt. Poterba (1994) shows that more restrictive state ﬁscal rules are correlated with

more rapid ﬁscal adjustment to unexpected deﬁcits. Kiewiet and Szakaly (1996)

show that restrictive provisions to limit debt issuance at the state level result in the

devolution of debt issuance to governments at the local level. Bohn and Inman (1996)

ﬁnd for a sample of 47 US states that only balance requirements enforced as constitu-

tional (not statutory) constraints by an independently elect ed (not politically appointed)

state supreme court have signiﬁcant positive eﬀects on a state’s general fund surplus.

Strauch (1998) shows that constitutional expenditure limits in the US induce a shift of

expenditures from the (constrained) current budget to the (unconstrained) investment

budget.

Daﬄon and Rossi (1999) survey the accounting tricks governments used in the run-up

to the Euro. They ﬁnd that the methodological rules of the European system of accounts

are weak and that numerous countries have used tricks to qualify for EMU membership.

Milesi-Ferretti and Moriyama (2004) ﬁnd that in the run up to EMU membership reduc-

tions in government debt were accompanied by strong decumulations of government

assets. These authors argue that, in the run up to the Euro, the ﬁscal rules of Maastricht

led to signiﬁcant ﬁscal operatio ns, which improved the oﬃcial ﬁgures but had no eﬀect on

the actual ﬁscal position of the government. The bottom line of this research is that ﬁscal

rules have an eﬀect on the ﬁscal aggregates to which they refer. However, governments try

to compensate for the loss of ﬂexibility due to the rule by shifting ﬁscal activities from

restricted to non-restricted instruments.

In this paper, we extend this line of research and test the model by Milesi-Ferretti

(2003). In particular, we document stock- ﬂow adjustments in the European Union, which

are computed as the annual changes in debt levels minus the annual budget deﬁcits. Posi-

tive stock-ﬂow adjustments imply that the debt level increases by more than it should given

the deﬁcit. While stock-ﬂow adjustments are a common feature of public ﬁnances due to

accounting issues, they should generally not generate a systematic bias betw een the stock

of debt outstanding and the sum of all budget deﬁcits over time. In many EU states, how-

ever, we ﬁnd that stock-ﬂow adjustments are persistent and that the diﬀerence between

debt stocks and accumulated deﬁcits is large. We interpret this as a ﬁrst indication of a

systematic strategic use of stock-ﬂow adjustment. We then investigate the eﬀect of the ﬁs-

cal rules in Europe, more precisely of the excessive deﬁcit procedure (EDP) and the stabil-

ity and growth pact (SGP). The SGP in particular puts a large weight on the deﬁcit limit in

the EMU, since, in the European public debate, the loss of political reputation is signiﬁ-

cant for countries breaching the deﬁcit limit but not for countries breaching the debt limit.

As greater attention is paid to the deﬁcit, we expect that governments try to shift budget

deﬁcits (restricted) to oﬀ-budget deﬁcits (non-restricted) in form of stock-ﬂow adjust-

ments. We ﬁnd that stock-ﬂow adjustments are systematically related to deﬁcits after

the ﬁscal rules became eﬀective. Recorded deﬁcits have been lowered by increasing

stock-ﬂow adjustments. This eﬀect is especially pronounced, when the ﬁscal rule is bind-

ing, i.e., when governments desire to run deﬁcits in excess of the formal limits. In addition,

we conﬁrm the prediction by Milesi-Ferretti (2003) that the use of creative accounting var-

ies over the business cycle.

The remainder of the paper is organized as follows: the next section presents accounting

identities, data and measurement issues and the amount of stock-ﬂow adjustment in the

European Union. In Section 3, we develop our estimation strategy and present the evi-

dence on creative accounting in the EU. Section 4 concludes.

J. von Hagen, G.B. Wolﬀ / Journal of Banking & Finance 30 (2006) 3259–3279 3261

2. Deﬁcits and debt: stock-ﬂow adjustments

2.1. Accounting identities

Standard textbooks in macroeconomics give Eq. (1) as the fundamental relationship

between deﬁcits, D, and debt, B. In this deﬁnition, the deﬁcit is calculated as the diﬀerence

between expenditure and revenue, where expenditure includes interest payments,

¼ B

t1

þ D

; ð1Þ

from this equation, the current debt level is equal to the accumulated past deﬁcits plus the

initial debt level (Eq. (2)),

¼ B

tn

n1

i¼0

ti

: ð2Þ

In practice, Eq. (1) does not always hold, if the deﬁcit is deﬁned as the diﬀerence between

budgetary expenditures and revenues. A residual can be computed according to:

 B

t1

 D

¼ SFA

: ð3Þ

This residual is called stock-ﬂow adjustment, or debt-deﬁcit adjustment. A positive stock-

ﬂow adjustment means that the stock of government debt has increased betw een period t

and (t  1) by more than the budget deﬁcit in period t indicates. The oﬃcial deﬁnition

treats stock-ﬂow adjustments as a statistical residual. As the European Commission states,

stock-ﬂow adjustments ‘‘result primarily from ﬁnancial operations, e.g., debt issuance pol-

icy to manage public debt, privatization receipts, impac t of exchange rate changes on for-

eign denominated debt. In general, these should tend to cancel out over time. However,

large and persistent stock-ﬂow adjustments (especially if they always have a negative im-

pact on debt developments) should give cause for concern, as they may be the result of the

inappropriate recording of budgetary operations and can lead to large ex-post upward

revisions of deﬁcit levels’’ (European Commission, DG for Economic and Financial Af-

fairs, 2003, p. 79).

2.2. Data and measurement issues

Deﬁcit and debt ﬁgures very much depend on their precise deﬁ nition and measurement

(e.g., Blejer and Cheasty, 1991). In this article we use data published in the AMECO data-

base, which is based on Eurostat data and serves as the basis of the EDP and the SGP.

Eurostat follows the ESA 95 accounti ng standard to measure deﬁcits and debt. The data

refer to the general consolidated government sector, which includes the central, state and

local government and the social security sector.

The deﬁnition of debt under the EDP, on

which our data are based, slightly diﬀers from ESA 95, as debt is recorded at face value in

the EDP and not at market value as in the ESA 95.

In general, the diﬀerence between def-

For recent evidence on upward revisions because of persistent stock-ﬂow adjustments, see Balassone et al.

(2004).

For details on the precise deﬁnitions see Eurostat, 2002, p. 8–16.

Debt means total gross debt at nominal value outstanding at the end of the year and consolidated between and

within the sectors of general government (Eurostat, 2002, p. 190).

3262 J. von Hagen, G.B. Wolﬀ / Journal of Banking & Finance 30 (2006) 3259–3279

icits and the change in debt levels results from the fact that public debt is a gross concept,

while deﬁcits are a net concept. For example, if a government issues debt and deposits the

proceeds in a bank acco unt, the eﬀect on the deﬁcit is zero, while gross debt increases.

Stock-ﬂow adjustments result from ﬁve main issues: (1), issuance of zero coupon bonds.

Consider a bond, which is issued for 90 Euro to cover a deﬁcit and has a face value of 110.

This operation is recorded as a deﬁcit of 90 and a stock-ﬂow adjustment of 20 in the year

of issuance, since the debt level at face value increases by 110. In the following four periods

until maturity, an interest of 5 accrues, impac ting on the deﬁcit, the debt level stays con-

stant, the stock-ﬂow adjustment is 5 in each period. As can be seen, stock-ﬂow adjust-

ments of zero coupon issuances should cancel out over time. (2) Revaluation of debt

denominated in foreign currency changes the face value of the debt, without having any

impact on the budget. Revaluation of foreign denominated debt should only matter if a

country has a depreciating currency ov er a long period. Foreign denominated debt does

not play any signiﬁcant role in any of the EU 15 countries. Exchange rate eﬀects are

<0.2% of GDP in general (ECB, 2004, Table 6.3.2). (3) Time-of-recording eﬀects: Deﬁcits

are measured in accrual terms, while debt is a cash concept. For example, when UMTS

licenses are sold, this has an eﬀect on the deﬁcit in the year of selling, so when the receipts

accrue, however, debt is only reduced when the (cash) receipts are used to buy back the

debt. The time of recording eﬀect should usually cancel out after some years.

The only

two remaining issues, where long and persisting positive stock-ﬂow adjustments can be

expected, are (4), equity injections in and privatization of public companies and (5), trans-

actions in ﬁnancial assets. Selling of ﬁnancial assets reduces gross debt, however, it has no

eﬀect on the deﬁcit ac cording to the rules of the ED P and the SGP.

Capital injections into public companies are an important tool to hide deﬁcits and

incur increases in the debt level, resulting in positive stock-ﬂow adjustments. The opera-

tion must be recorded as an equity injection or transaction in shares and other equity,

i.e., the government has to declare to have a lasting economic interest in the public com-

pany in the sense of intending to receive a market interest rate. If the injection is used to

cover recurring losses of the public company, it should be recorded as a current transfer,

which leaves the stock-ﬂow adjustment unaﬀected. In practice, it is very diﬃcult to con-

trol whether injections have been correctly recorded.

Capital injections can therefore be

used to shift public expenditure from the (deﬁcit-relevant) state sector to public compa-

nies. Thereby, the deﬁcit is reduced, while the change in debt level captures the ‘‘true’’

public spending. A positive stock-ﬂow adjustment results.

These mechanisms have also

attracted the attention of the European Commission. Joaqu ı

n Almunia, EU monetary

Interest accrued aﬀects net borrowing/net lending. ‘‘For government debt under EDP (at nominal value, not

including accrued interest) interest due but not paid is to be recorded under Other accounts payable (F.79), as

long as it is not paid (ESA95, §5.131). In the EDP, interest arrears under Other accounts payable are not

accounted for in the government debt’’ (Eurostat, 2002, p. 199). Thus, interest payments are recorded in the

deﬁcit when they accrue, even if they are not paid yet, and should in this case lower stock-ﬂow adjustments as they

are not recorded in the debt according to EDP. In the long-run, interest payments are without eﬀect on stock-ﬂow

adjustment.

The sale of non-ﬁnancial assets reduces the deﬁcit, as it is recorded as negative investment or more precisely

negative public ‘‘gross ﬁxed capital formation’’.

Public-private partnerships can also be used to hide deﬁcits. However, they will not automatically be seen in

the form of stock-ﬂow adjustments.

J. von Hagen, G.B. Wolﬀ / Journal of Banking & Finance 30 (2006) 3259–3279 3263

aﬀairs commissioner, has recently claimed that governments disguise the scale of their

budget deﬁcits.

2.3. Debt vs. accumulated deﬁcits: descriptive evidence

Persistent stock-ﬂow adjustments can be a matter of concern (European Commission,

DG for Economic and Financial Aﬀairs, 2003, p. 79). A natural test for the persistence of

stock-ﬂow adjustments is to compare the debt level (column B of Table 1 ) with the accu-

mulated deﬁcits as described in Eq. (2) , i.e., the sum of the debt level of 1980 and all bud-

get deﬁcits between 1980 and 2003 as a percent of 2003 GDP (column C of Table 1), both

measured in percent of GDP of 2003. Calculating the diﬀerence of actual debt levels and

accumulated deﬁcits in percent of GDP (B–C) shows that most EU countries have regu-

larly had positive stock-ﬂow adjustments. Finland and Greece have 64% and 43% points

of GDP more debt than their budget data suggest, followed by Denmark (30), Luxem-

bourg (29), Germany (15), and Austria (14). The cases of Finland and Luxembourg are

noteworthy as the debt level should be negative if one added the deﬁcits and surpluses.

In both countries, budget surpluses have thus been used in the last years to buy assets

instead of paying back deb t.

Oﬀ-budget debt accumulation in the form of stock-ﬂow adjustments thus plays a con-

siderable role in most EU-15 countries, with substantial variations across countries.

Stock-ﬂow adjustments constitute a signiﬁcant part of the overall debt accumulation in

the member states of the EU. The (unweighted) annual average stock-ﬂow adjustment

in the EU amou nts to 1.56% of GDP in the period 1981–2003. As Fig. 1 shows, it exceeds

3% of GDP in some years. Average stock-ﬂow adjustments had a declining trend until

1996. This negative trend may reﬂect the persistent eﬀorts of European governmen ts to

comply with the debt limit imposed by the Maastricht Treat. Since 1996 the average

SFA has risen again to more than 2% in 2001. A similar pattern can be seen for the aver-

age of three large EU economies, France, Germany and Italy. As argued by Milesi-Ferretti

and Moriyama (2004), EU governments thus reduced their gross debt by selling assets

rather than by improving the net-ﬁnancial position of the government through lower def-

icits. This was a convenient way to fulﬁll the Maastricht criterion of falling debt levels for

the highly indebted countries.

Stock-ﬂow adjustments of selected countries deserve particular attention. In Germany,

the debt level increased by more than 6% of GDP in addition to the deﬁcit in 1995, when

the German federal government oﬃcia lly assumed the debt previously hidden in the

Treuhandanstalt, the holding company of former East German industries.

In Greece,

the stock-ﬂow adjustment reached almost 19% in 1994, when the debt of the Greek gov-

ernment at the Bank of Greece was oﬃcially recorded as public deb t. Finland experienced

http://news.ft.com/cms/s/cd1192f0-35cd-11da-903d-00000e2511c8.html.

Also the US has a signiﬁcantly higher debt level than given by the sum of deﬁcits with a diﬀerence of 9% of

GDP in the period 1980–2003.

In fact, when the Treuhandanstalt was dissolved, the debt of 204 billion DM were carried forward to the

Erblastentilgungsfond. The German statistical oﬃce wanted to classify this as an increase of the debt and of the

deﬁcit. Theo Waigel, the ﬁnance minister at the time, however objected and argued that this debt should not

impact on the deﬁcit according to the Maastricht criteria. Eurostat accepted this view, which explains the large

stock-ﬂow adjustment in this year (Mu

nster, 1997).

3264 J. von Hagen, G.B. Wolﬀ / Journal of Banking & Finance 30 (2006) 3259–3279

a stock-ﬂow adjustment of 12% in 1992 related to the banking crisis it suﬀered in connec-

tion with the currency crisis of the Markka. The negative stock-ﬂow adjustment of

Belgium in 1996 is noteworthy. It reﬂects a booking operation designed to show that Bel-

gium had a declining debt level and, therefore, qualiﬁed for EMU membership (Laughland

1 0 1 2 3 4

Percent of GDP

1980 1985 1990 1995 2000 2005

Year

Average SFA in the EU

Average SFA of France, Germany, and Italy

Fig. 1. Average stock-ﬂow adjustments in percent of GDP. Source: Ameco database, authors’ calculation.

Table 1

Debt and accumulated deﬁcits in percent of GDP

Country Debt, 1980 Debt, 2003 Sum of deﬁcits Diﬀerence

A B C B–C

Austria 36 66 52 14

Belgium 79 103 100 3

Denmark 36 43 13 30

Finland 11 45 19 64

France 20 63 54 9

Germany 31 64 49 15

Greece 25 101 58 43

Ireland 75 33 25 8

Italy 58 106 99 7

Luxembourg 9 5 24 29

Netherlands 46 55 53 2

Portugal 32 58 53 5

Spain 17 51 47 4

Sweden 40 52 50 2

United Kingdom 53 40 39 1

Source: Ameco, own calculations; The accumulated absolute deﬁcits were added to the initial debt level of 1980

(column A) for all countries, except Greece (1988), Luxembourg and Ireland (1990), Sweden (1993), and Spain

(1995) due to data constraints. This cumulative debt measure was divided by GDP of 2003.

J. von Hagen, G.B. Wolﬀ / Journal of Banking & Finance 30 (2006) 3259–3279 3265

and Paul, 1997). In summary, we ﬁnd signiﬁcant evidence for persistent and systematic use

of stock-ﬂow adjustments.

2.4. Stock-ﬂow adjustments and creative accounting

Persistent, positive stock-ﬂow adjustments allow governments to accumulate public

debt in excess of what is implied by the annual budget deﬁcits. But they may also be

the result of net acquisitions of ﬁnancial assets in times of budget surpluses. For example,

a governm ent using budget surpluses to accumulate reserves in pension insurance funds

would have long lasting positive stock-ﬂow adjustments, without hiding away any deﬁcits.

To test for this possibility, we check the correlation between the net acquisition of ﬁnancial

assets and stock-ﬂow adjustments, both in percent of GDP, for the period and countries

for which data were available, namely for the period 1996–2002, and Austria, Belgium,

Finland, France, Germany, Italy, Netherlands, Portugal, and Spain. The resulting coeﬃ-

cient is 0.21 and is not statistically signiﬁcant.

But even if stock-ﬂow adjustments reﬂect asset acquisitions, this could still be a way to

change oﬃcial deﬁcit ﬁgures. The data on asset acquisition do not reﬂect the evolution of

the value of these assets. It is possible, for example, that the asset acquisition is only a hid-

den subsidy, or capital injection into a (public) company. The public company could then

engage in standard public expenditure, driving down the value of its assets without any

impact on the gross debt level of the government, nor on net borrow ing. We therefore con-

clude that the stock-ﬂow adjustments observed in Europe reﬂect to some extent at least

systematic creative accounting.

3. Fiscal rules and stock-ﬂow adjustments

3.1. Approach

Milesi-Ferretti (2003) argues that ﬁscal rules can induce governments to engage in

‘‘bad’’ or even ‘‘ugly’’ creative accounting. To test this proposition, we investigate the rela-

tionship between deﬁcits and stock-ﬂow adjustments. The SGP is a ﬁscal rule with a par-

ticular focus on budg et deﬁcits. It requires the deﬁcit to stay below the 3% reference value

and to have a budget close to balance in the medium term. Following Milesi-Ferretti

(2003), the tendency to use stock-ﬂow adjustments to keep reported deﬁcits lower than

the actual increase in public debt should have increased since the inception of the SGP.

It should be particularly pronounced, when countries are close to breaching the 3% limit

and when the cost of reducing the deﬁcit is high, i.e., in times of recessions.

More speciﬁcally, consider a government allocating expenditures and taxes optimally

over time as in Milesi-Ferretti (2003). Assume that the government has derived an optimal

change in government debt, D



, for period t. If the actual change in government debt devi-

ates from the optimal change, the government suﬀers a cost of K

¼ a

ðB

 B

t1

 D



The parameter a

may vary over time, reﬂecting, e.g., diﬀerent costs of deviating from

the optimal ﬁscal stance at diﬀerent points of the business cycle. In the absence of any con-

The correlation coeﬃcients have to be interpreted carefully as the data on assets refer to the non-consolidated

general government sector, while the deﬁcit and debt data are consolidated across government sectors. The data

source for net asset accumulation is Annual National Financial Accounts (ANFA) dataset.

3266 J. von Hagen, G.B. Wolﬀ / Journal of Banking & Finance 30 (2006) 3259–3279

straint on the government budget deﬁcit, the actual change in government debt equals the

budget deﬁ cit, D

, plus an exogenous, random stock-ﬂow adjustment, e

, due to timing

eﬀects etc., with E

t1

) = 0, where E

t1

denotes the expected value based on information

available in period t1. Thus, B

 B

t1

¼ D

þ e

¼ D



þ e

Note that, by assumption, the

budget deﬁcit and the observed stock-ﬂow adjustment are uncorrelated.

Assume now that there is a deﬁcit limit, DL, applying to the observed budget deﬁcit. If

the government violates the deﬁcit limit, it suﬀers a cost increasing in the size of the vio-

lation, K

= b

+ b

 DL)

for D

>DLandK

= 0 otherwise. Finally, assume that

the government can use stock-ﬂow adjustments strategically to increase government debt

deliberately by more than the budget deﬁcit. Let the strategic component of stock-ﬂow

adjustments be S

. Thus, the total stock-ﬂow adjustment observed ex-post is SFA

+ e

. With a given probability p, the strategic use of stock-ﬂow adjustments will become

known to the public. If that happens, the government suﬀers a reputational cost

= c

+ c

for S

> 0, and K

= 0 otherwise.

The government now has two instruments to implement the optimal change in govern-

ment debt, budget deﬁcits and strategic stock-ﬂow adjustments. Formally, it minimizes the

cost function K

= K

+ K

subject to the constraint B

 B

t1

= D

+ S

+ e

and

< DL. As long as the deﬁcit limit is not binding, the government will not use stock-ﬂow

adjustments strategically, and choose D

¼ E

t1

ðB

 B

t1

Þ¼D



. Assuming that the cost of

deviating from the optimal change in government debt is known when D

and S

are cho-

sen, we obtain the following solutions for the budget deﬁcit and the strategic stock-ﬂow

adjustment:



ðD



 S



Þþb

þ a

; ð4Þ



ðD



 D



þ a

. ð5Þ

Thus, if the cost of deviating from the optimal change in government is small, a

< b  c,

the change in government debt will be smaller than in the absence of the debt limit, and the

actual deﬁcit lies between the unconstrained deﬁcit and the deﬁcit limit. Furthermore, if

the deﬁcit limit is binding, the actual deﬁcit and the strategic part of the stock-ﬂow adjust-

ment are negatively correlated. That is, the larger the strategic stock-ﬂow adjustment, the

smaller the observed budget deﬁcit. This correl ation increases, when the economic cost of

deviating from the optimal change in government debt becomes large. Subsequently, we

focus on this correlation to test for the strategic use of stock-ﬂow adjustments to hide gov-

ernment deﬁcits under EMU.

Based on these considerations, we do not expect a systematic relationship between

stock-ﬂow adjustments and the deﬁcit before 1998, when the SGP became binding. After

1998, however, governments may have used stock-ﬂow adjustments actively to control the

deﬁcit. SFA becomes a policy variable, inﬂuencing the deﬁcit level. Thus, we expect a

signiﬁcant negative relation between stock-ﬂow adjustments and the budg et deﬁcit for

the period 1998–2003, when EU countries had to comply with the SGP. In a ﬁrst step

we calculate the correlation coeﬃcients between stock-ﬂow adjustments and deﬁcits in

the two periods for all EU 15 countries. In the ﬁrst period (1980–1997), the correlation

is 0.03 and insigniﬁcant, in the second period (1998–2003) it is 0.53 and statistically

signiﬁcant.

J. von Hagen, G.B. Wolﬀ / Journal of Banking & Finance 30 (2006) 3259–3279 3267

This simple correlation analysis allows neither for country-speciﬁc eﬀects nor for auto-

correlation in the variables caused by business cycle ﬂuctuations. We therefore employ the

following more elaborate panel econometric approach. From identity (3), we know that

the change of the total debt level in percent of GDP in country i at time

t ðDb

B

i;t1

Þ is the sum of stock-ﬂow adjustment in percent of GDP (sfa

) and the

deﬁcit in percent of GDP (d

), i.e., Db

= sfa

+ d

. Estimating the following equation:

¼ a

þ a

sfa

þ e

; ð6Þ

gives a

as:

covðDb

; sfa

varðsfa

covðsfa

þ d

it;

sfa

varðsfa

Þþcovðd

; sfa

varðsfa

¼ 1 þ

covðd

; sfa

varðsfa

: ð7Þ

If a

= 1, we know that the covariance between deﬁcits and stock-ﬂow adjustments is zero.

A coeﬃcient smaller (larger) one implies a negative (positive) covariance between sfa and

d. The following regression allows to estimate the impact of the ﬁscal rule,

¼ a

þ a

sfa

þ a

sfa

 T

þ l

þ e

; ð8Þ

where T

is a dummy that takes a value of 1 for the years 1998–2003 and zero otherwise.

The coeﬃcient a

measures the eﬀect of the dummy (the ﬁscal rule) on the level of the

change in debt levels. a

Measures the eﬀect of the ﬁscal rule on the relationship between

sfa and the change in debt levels. The coeﬃcient a

gives the impact of the ﬁscal rule on the

covariance between the deﬁcit and sfa. Given that our hypothesis of no relation betw een d

and sfa before the introduction of the rule holds true, i.e., a

= 1, the coeﬃcient a

then

directly measures the covariance between deﬁcits and stock-ﬂow adjustments after 1997.

A negative coeﬃcient a

implies, that the covariance between deﬁcits and stock-ﬂow

adjustments became negative in the second period. An increase in the stock-ﬂow adjust-

ment (sfa

) would therefore result in a lower deﬁcit.

In order to separate the eﬀects of structural from cyclically adjusted deﬁcits, we specify

an alternative regression model:

¼ b

þ b

 T

þ l

þ e

; ð9Þ

where T

is again a dummy that takes a value of 1 for the years 1998–2003 and zero for the

years before. The eﬀect of the ﬁscal rules (the treatment eﬀect) can then be calculated

accordingly. A negative coeﬃcient b

implies, that an increase in the deﬁcits (d

) results

in a lower stock-ﬂow adjustment as a consequence of the introduction of the ﬁscal rule.

The coeﬃcients b

and a

should be of the same sign as they reﬂect the same covariance.

To capture the eﬀect of the structural and the cyclical part of the deﬁcit, we augment Eq.

(9) and estimate,

¼ b

þ b

 T

þ b

 T

þ l

þ e

; ð10Þ

where d

speciﬁes the cyclically adjusted (structural) deﬁcit, whereas d

is the cyclical part

of the deﬁcit. We expect the coeﬃcient b

to be of similar size as the coeﬃcient a

since the

largest part of the deﬁcit is structural. For b

the model by Milesi-Ferretti (2003) predicts a

larger coeﬃcient, i.e., we expect creative accounting to be used strongly with the business

cycle.

3268 J. von Hagen, G.B. Wolﬀ / Journal of Banking & Finance 30 (2006) 3259–3279

To test whether creative accounting is most prevalent, when the rule is binding, we fur-

ther augment the approach. We separate empirically the eﬀect of the introduction of the

rule captured by the time dummy from the eﬀect the rule has on governments, whose pol-

icy objective is to spend more than 3% of GDP in excess of their revenu es. Since in this

case, we expect coun tries to employ SFA systematically to lower the deﬁcit, we cannot

identify the policy objective by looking at the deﬁcit. Instead, we identify the intention

of the government to br each the 3% limit by

¼ 3% if the ﬁscal rule is in place.

If the rule applies only to those countries, which breach the 3% criterion, then only those

countries intending to increase their debt level by more than 3% should engage in creative

accounting. Countries below that limit should show no particular attempt to engage in cre-

ative accounting.

This means that the correlation between stock-ﬂow adjustments and

deﬁcits should be negative when the rules are in place (second period) and the change

of the debt level is above 3%. It should be zero, when the change of the expected debt level

is below 3%.

The following model is estimated to test this hypothesis,

¼ c

þ c

sfa

þ c

 sfa

þ c

 R

þ c

 sfa

þ c

 R

 sfa

þ l

þ e

; ð11Þ

where R

¼ 1ifE(Db

t1

) > 3 is a dummy taking the value of 1 if the expected change in

debt is larger than 3% of GDP. In this view, governments will engage in creative account-

ing if they expect their newl y accumulated debt to be above 3%. We therefore expect c

be signiﬁcantly negative, implying a strong negative correlation between sfa and deﬁcits,

when the rules are in place (second period) and binding (expected debt level change larger

3). The eﬀect of the introduction of the rule, given by c

, should become insigniﬁcant, if we

assume that the rule is not binding when countries have a deﬁcit below 3%. Finally, as

given by the accounting identity, we expect sfa to contribute as above one to one to an

increase of the debt level, i.e., c

= 1 and c

=0.

An obvious problem of these approaches is the simultaneous equation bias, which ren-

ders the least square estimator inconsistent (Gujarati, 1995, pp. 642; Greene, 2000, pp.

652). We therefore ran two stage least square instrumental variable estimators and instru-

mented with the lag of the variables. To address the endogeneity of R, we instrumented the

realized R

with the appropriate lags. However, in this approach we cannot take account

of serial correlation. Serial correlation can be expected as the change in the debt level Db

depends on the business cycle.

We theref ore specify a dynamic panel model with the lagged dependent variable

included as a regressor. We use the dynamic panel estimator by Arellano and Bond

(1991), restricting the number of lagged levels to ﬁve in the instrument set.

To address

the simultaneous equation bias and endogeneity bias, we explicitly allow sfa, T Æ sfa,

and for the extended approach in addition R Æ sfa, R Æ T Æ sfa, R,andR Æ T to be endoge-

nous variables. This means that all possible lags until t  1 of these variables in levels

are included as instruments for these endogenous variables.

In fact, if

< 3 and SFA > 0, we can be sure, that the government meets the criteria set by the pact.

An extension of the instrument set to all possible lags did not change any of our results. A reduction of lags

also has robust results.

J. von Hagen, G.B. Wolﬀ / Journal of Banking & Finance 30 (2006) 3259–3279 3269

3.2. The eﬀect of ﬁscal rules

The basic empirical results are shown in Table 2. Stock-ﬂow adjustments, as the

accounting identity suggests, contribute to the change in the debt level with a coeﬃcient

close to one, the 95% conﬁdence interval for regression 1 is [0.819, 0.992]. Increasing the

stock-ﬂow adjustment per GDP by one percentage point results in roughly one additional

percentage point debt level per GDP. However, this changes in the second period, when an

increase in the stock-ﬂow adjustment results in a

+ a

additional debt, and stock-ﬂow

adjustments do not translate into higher debt on a one to one basis. As the coeﬃcient

is statistically not diﬀerent from 1, the estimated coeﬃcient a

represents the covariance

between stock-ﬂow adjustments and the deﬁcit in the second period, which we ﬁnd to be

signiﬁcantly negative. In regression (1) of Table 2, an increase of sfa by one percentage

point results in a 0.25 percentage point lowering of the deﬁcit. This suggests that

stock-ﬂow adjustments have become a policy variable to control the deﬁcit in the time per-

iod when the ﬁscal rule was in place. In the earlier period, the regression results do not

imply any correlation between stock-ﬂow adjustments and deﬁcits. Thus, our results indi-

cate that the introduction of the ﬁscal rule led governments to systematically use stock-

ﬂow adjustments to lower deﬁcits.

To check the robustness of our results, we omit a number of countries and observa-

tions. Finland and Sweden are dropped, as Finland had positive stock-ﬂow adjustment

because of budget surpluses invested into assets, and so did Sweden in some years. We also

drop Finland, Sweden, and Luxembourg, as all three countries had positive stock-ﬂow

Table 2

Measuring the impact of ﬁscal rules

12 3 4 5 6 7

sfa 0.91 0.95 0.97 0.94 0.88 0.89 0.95

0.04 0.04 0.04 0.05 0.05 0.05 0.05

T 1.55 1.11 1.31 1.16 0.88 1.84 0.91

0.33 0.31 0.32 0.31 0.37 0.37 0.42

T · sfa 0.25 0.40 0.38 0.33 0.32 0.18 0.36

0.08 0.09 0.10 0.10 0.09 0.08 0.08

Cons 0.04 0.02 0.01 0.00 0.00 0.03 0.14

0.03 0.03 0.03 0.03 0.03 0.03 0.05

LDV 0.51 0.45 0.44 0.46 0.53 0.50 0.41

0.03 0.03 0.03 0.03 0.03 0.03 0.04

Obs 263 233 221 220 212 221 183

Sargan p 0.88 0.98 0.99 1.00 1.00 1.00 0.99

Autocorr 2, p 0.50 0.86 0.80 0.75 0.75 0.43 0.87

Omitted SE, FI SE, FI, LU SE, FI, GR SE, DK, UK DE,FR <1991

Note: Dependent variable: change in the debt relative to GDP. T = 1 if year > 1997. LDV refers to the lagged

dependent variable. Last line refers to which observations were omitted. Standard errors are reported below the

coeﬃcients. Method: Arellano bond dynamic GMM panel estimator.

It is possible that strong negative shocks to the budget induce the government to increase deﬁcits and stock-

ﬂow adjustments, thereby causing a positive correlation. Our result is strengthened, since we ﬁnd the negative

relationship to prevail. The systematic use of creative accounting thus outweighs possible shocks (e.g., resulting

from control errors) aﬀecting the deﬁcit and stock-ﬂow adjustments in the same direction.

3270 J. von Hagen, G.B. Wolﬀ / Journal of Banking & Finance 30 (2006) 3259–3279

adjustments because of asset purchases (regression 2). Also some of the Greek ﬁgures

might be distorted in the early to mid-nineties, and we also know that the data in the later

years were wrongly reported (regression 3). Then we also drop the three non-Euro coun-

tries, which are oﬃcially subject to the ﬁscal rules, however, without being subject to ﬁnes

in case of non-compliance (regression 4). In a further regression, we drop Germany and

France, as it might be diﬃcult to enforce sanctio ns against them. They might therefore

be less constrained by the ﬁscal rule (regression 5). We also exclude the observations from

the 1980s, as in this period, the emergence of any set of rules was not discussed (regression

6). None of these control regressions changes our results.

The results based on regression Eq. (9) conﬁrm this ﬁnding and are given in Appendix,

Table A.1. In the ﬁrst period, there is no systematic relationship between stock-ﬂow

adjustment and deﬁcits, while in the second period a negative co-variance emerges.

increase in the deﬁcit by one percentage point is associated with a lowering of the

stock-ﬂow adjustments by 0.32 percentage point. This ﬁgure is also robust to changes

in the sample.

We then separate the eﬀect of the cyclically adjusted deﬁcit and the cyclical component

of the deﬁcit (Table 3). We use the two oﬃcial measures of cyclically adjusted balances

provided by the European Commission. The ﬁrst is based on the output gap in a structural

model, the second is based on an HP-ﬁltered trend.

The estimation results based on

potential output are presented in the left part of the table, while trend output results

are given on the right side of the table. In the regression, we include the cyclically adjusted

deﬁcit (CAD) and the cyclical part of the deﬁcit (CD). Again the coeﬃcient for the ﬁrst

period is close to 1 as we expect for the structural deﬁcit and for the cyclical deﬁcit. Thus,

in the ﬁrst period we do not ﬁnd a signiﬁ cant correlation between deﬁcits an d stock-ﬂow

adjustments. For the second period, however, there is a clear negati ve correlation between

the structural deﬁcit and stock-ﬂow adjustments for both calculation methods similar in

magnitude to the previously estimated coeﬃcient.

The cyclical component of the deﬁcit

and stock-ﬂow adjustments in the second period are very strongly negatively correlated. In

fact, an increase in the cyclical deﬁcit in the second period is almost completely oﬀset by

reductions in stock-ﬂow ad justments, indicating that stock-ﬂow adjustments are used to

weaken the impact of the cycle on the deﬁcit.

Table 3 also provides robustness checks for the impact of cyclical and structural deﬁcits

on the use of stock-ﬂow adjustments. Similar to the other robustness checks, we drop var-

ious countries from the sample. Finland, Luxembourg and Sweden are dropped because of

their surpluses, Greec e is dropped as data quality is problematic, Denmark, Sweden and

the UK are dropped as they do not belong to the Eurozone, and ﬁnally Germany and

France are dropped as sanctions and warning letters are diﬃcult to enforce against them.

The estimation results conﬁrm the previous results. For the second period, there is a strong

negative correlation between the cyclical part of the deﬁcit and stock-ﬂow adjustments for

With the robustness check, we show that the signiﬁcance of our regression coeﬃcients does not depend on the

choice of countries and methods. It is not possible, however, to compare the magnitude of the regression

coeﬃcients, since the standard errors are too large.

The 95% conﬁdence interval for b

is [0.705, 0.993]. The H

that

1PLDV

¼ 1 cannot be rejected with a p-value

of 0.71. The coeﬃcient b

for the interacted term thus again measures the covariance.

For details on the de-trending methodologies of the EU, see (European Commission, 2004, p. 79).

In most EU countries, the structural deﬁcit represents the largest part of the total deﬁcit.

J. von Hagen, G.B. Wolﬀ / Journal of Banking & Finance 30 (2006) 3259–3279 3271

Table 3

Measuring the impact of ﬁscal rules on the cyclical component of the deﬁcit

CAD1 0.82 0.89 0.81 0.78 0.75 CAD2 0.80 0.91 0.85 0.81 0.79

0.07 0.09 0.07 0.08 0.08 0.08 0.09 0.07 0.08 0.08

T · CAD1 0.20 0.16 0.23 0.24 0.23 T · CAD2 0.21 0.15 0.21 0.23 0.20

0.12 0.17 0.11 0.13 0.14 0.12 0.17 0.11 0.14 0.14

CD1 1.01 1.41 1.18 1.03 1.28 CD2 1.15 1.34 1.03 0.88 1.14

0.13 0.19 0.12 0.15 0.15 0.13 0.19 0.12 0.14 0.14

T · CD1 0.83 0.91 0.87 0.64 1.05 T · CD2 0.88 0.89 0.81 0.53 1.03

0.25 0.31 0.22 0.27 0.27 0.33 0.24 0.22 0.29 0.29

T 1.44 1.53 1.63 1.30 1.79 T 1.42 1.59 1.67 1.32 1.80

0.49 0.57 0.43 0.52 0.54 0.48 0.57 0.43 0.54 0.54

Cons 0.19 0.16 0.20 0.18 0.25 Cons 0.18 0.18 0.21 0.18 0.26

0.03 0.04 0.03 0.04 0.04 0.03 0.04 0.03 0.04 0.04

LDV 0.17 0.15 0.15 0.16 0.13 LDV 0.17 0.14 0.15 0.17 0.13

0.05 0.06 0.05 0.06 0.06 0.05 0.06 0.05 0.06 0.06

Obs 263 221 250 212 221 263 221 250 212 221

Sargan 1 1 0.79 0.99 0.99 1 1 0.78 0.99 0.99

Autocorr 2, p 0.59 0.31 0.94 0.65 0.67 0.59 0.34 0.92 0.66 0.66

Omitted FI, LU, SW GR DK, SW, UK DE, FR FI, LU, SW GR DK, SW, UK DE, FR

Note: Dependent variable: change in the debt relative to GDP. T = 1 if year > 1997. LDV refers to the lagged dependent variable. CAD1 = cyclically adj. deﬁcit,

CD1 = cyclical component both based on potential output. CAD2 = cyclically adj. deﬁcit, CD2 = cyclical component both based on HP ﬁltered output trend.

Standard errors are reported below the coeﬃcients. Method: Arellano bond dynamic GMM panel estimator. Deﬁcit, and interaction term speciﬁed as endogenous

variables.

3272 J. von Hagen, G.B. Wolﬀ / Journal of Banking & Finance 30 (2006) 3259–3279

both calculation methods. Any change in the cyclical deﬁcit is almost completely oﬀset by

a corresp onding change of sfa.

To show the robustness of our results to changes in the methodology, we also report the

results of a non-dynamic model, neglecting the simultaneous equation bias (Tables A.2–

A.4). OLS and ﬁxed eﬀect regressions yielded similar results.

To control for heterosced-

asticity, we also run generalized least squares. However, Monte-Carlo simulations by Beck

and Katz (1995) show that GLS provides over-optimistic standard errors in panels of our

size, therefore, we present the panel corrected standard error results in the Tables. Overall,

the model ﬁts the data reasonably well. The size of the coeﬃcients is slightly larger than in

the Arellano bond GMM estimator, as expected. In the entire investigated period, the

average debt acc umulation per year given stock-ﬂow adjustment s of 0% of GDP was

roughly 4.5% of GDP, in the second period it went however down by almost 3% points.

In this sense the ‘‘treatment’’, the introduction of ﬁscal rules is successfully reducing debt

accumulation, especially the recorded deﬁcit. a

remains statistically signiﬁcant and nega-

tive. In the second period, an increase in the deﬁcit by 1% point resulted in roughly 0.3

lower deﬁcits. The coeﬃcients for the cyclical and structural component of the deﬁcit are

similar to the benchmark regressions. Especially the cyclical part of the deﬁcit is oﬀset by

an equall y strong movement of stock-ﬂow adjustments.

The regression coeﬃcients robustly indicate that governments have used stock-ﬂow

adjustments systematically to hide deﬁcits since the ﬁscal rules are introduced in Europe.

This is especially relevant for the cyclical component of the deﬁcit.

3.3. The eﬀect of binding ﬁscal rules

Our regression analysis so far shows that after the introduction of the ﬁscal rules, stock-

ﬂow adjustments have become a policy instrument to control the evolution of the deﬁcit.

In this section we want to extent this point and separate empirically the eﬀect of the intro-

duction of the rule from the eﬀect the rule has on governments, when it actually constrains

their intentions.

As a ﬁrst test, whether the sfa increases when the constraint of the SGP becomes bind-

ing, we compare the average sfa in the group of observations with Db > 3 before and after

the SGP was in place with the average of sfa for the other group (Db <3)sfa increased

more in the ﬁrst group than in the second group (0.36 = (2.99  2.19)  (0.94  0.5)). This

indicates, that the use of sfa as a way to incur new debt has especially increased in the

group of countries, for which the rule on deﬁcits is binding.

The econometric results aimed at testing whether creative accounting increases in

importance, when the ﬁscal rules become binding, are presented in Table 4. We present

panel ﬁxed eﬀect regressions and dynamic Arellano bond estimation results.

The ﬁrst, somewhat surprising result concerns the coeﬃcient on sfa and the deﬁcit. It is

now only 0.5 and 0.43, respectively, for the whole sample. However, we also ﬁnd a signif-

icant interaction term R. sfa of 0.38 and 0.35, wher e we expected a zero coeﬃcient.

This means that cou ntries with a lower change in debt show a negative correlation between

deﬁcits and sfa, while for debt level changes above 3% we ﬁnd a coeﬃcient close to the

The F-test on the ﬁxed eﬀects indicates that country speciﬁc eﬀects are signiﬁcant.

In the Arellano bond estimation procedure we addressed the endogeneity issues discussed above.

J. von Hagen, G.B. Wolﬀ / Journal of Banking & Finance 30 (2006) 3259–3279 3273

previously estimated 1 implying no correlation between sfa and deﬁcits. This eﬀect is dri-

ven by budget surpluses as we conclude from a look at the scatterplots. Countries with

budget surpluses in general do not use these surpluses to reduce their debt levels but

instead have a tendency to buy assets. This observation is in line with our descriptive evi-

dence of Section 2.3, where the largest discrepancy between debt and accumulated deﬁcits

are observed for Finland, a country with signiﬁcant surpluses.

We, therefore, estimate the proposed model for only those observations, where the def-

icit is positive (i.e., we dropped obs ervations of budget surpluses in columns 4 and 5 of

Table 4). Now, the coeﬃcient on sfa has the expecte d value, i.e., 1, stock-ﬂow adjustment

contribute to increases of the debt level on a one to one basis, no negative correlation

between sfa and deﬁcits exists in general.

In addition, the eﬀect of sfa on the change in debt is the same for R = 1 and R = 0. The

coeﬃcient on sfa interacted with the dummy for the second period is insigniﬁcant. This

means, that no general negative correlation between sfa and deﬁcits can be observed in

the second period.

In contrast, we ﬁnd a strongly signiﬁcant and negative eﬀect for the coeﬃcient on sfa

interacted with the dummy for the period when the SGP was in place and a dummy for

those observations, where countries intend to increase their debt levels by more than

3%. This means, that for those observations, for which the rule is expected to become

Table 4

Measuring the impact of binding ﬁscal rules

Fixed eﬀect Arellano bond Fixed eﬀect Arellano bond

sfa 0.40 0.50 1.06 0.95

0.12 0.08 0.21 0.14

R · sfa 0.54 0.38 0.09 0.03

0.13 0.09 0.22 0.14

T · sfa 0.12 0.03 0.23 0.11

0.15 0.10 0.42 0.26

T · R · sfa 1.14 0.51 0.89 0.62

0.28 0.18 0.49 0.29

R 3.67 2.24 2.83 1.46

0.39 0.27 0.45 0.30

T 1.56 0.63 1.28 0.24

0.39 0.31 0.54 0.40

T · R 0.95 0.22 0.87 0.24

0.96 0.61 1.02 0.63

Cons 2.03 0.02 3.31 0.06

0.28 0.02 0.38 0.03

LDV 0.41 0.37

0.03 0.03

Obs 293 263 225 180

0.74 0.77

Sargan 0.32 0.95

Autocorr 2, p 0.79 0.68

Omitted observations None None Budget surplus Budget surplus

Variable: change in the debt relative to GDP. T = 1 if year > 1997. R =1ifDb > 3, LDV refers to the lagged

dependent variable. Standard errors are reported below the coeﬃcients. Methods: panel ﬁxed eﬀect FE and

Arellano bond dynamic GMM panel estimator AB.

3274 J. von Hagen, G.B. Wolﬀ / Journal of Banking & Finance 30 (2006) 3259–3279

binding, a negative regression coeﬃcient for sfa of 0.62 is estimated, implying a negative

correlation between sfa and deﬁcits in case of a binding ﬁscal rule.

Again, we check the robustness of the estimation results by dropping a variety of coun-

tries in the estimation. The results are presented in Table A.5 in Appendix. The results

conﬁrm the result. The estimated coeﬃcient on the simple interaction term for the second

period becomes insigniﬁcant. This means, that the imposition of the SGP does not result in

a signiﬁcant general use of sfa to improve the budget. We ﬁnd a signiﬁcantly negative coef-

ﬁcient for the interaction term R · T · sfa. This means that those governments that intend

to run deﬁcits beyond the 3% limit heavily resort to stock-ﬂow adjustments instead of the

deﬁcit to comply with the legal limits.

The extended estimation gives a more de tailed view on the empirical eﬀect of ﬁscal

rules.

The results show that any intended increase of the debt level beyond 3% under

the SGP regime was almost exclusively done by increasing stock-ﬂow adjustments instead

of oﬃcially published deﬁcits. Stock ﬂow adjustments were, however, not used as a general

mean to reduce the deﬁcit to comply with the pact’s provision to balance the budget in the

medium term. These results are in line with evidence taken from political debates. The

debate among policy makers almost exclusively focussed on the ‘‘magic’’ 3% threshold.

At this threshold, we also ﬁnd creative accounting to become relevant. The robustness

tests further conﬁrm that governments in general heavily resort to creative accounting

in the EU, when their desired ﬁscal stance does not comply with the binding ﬁscal rule.

4. Conclusions

Fiscal rules are introduced to constrain governments. EU countries have adopted a set

of rules to constrain deﬁcits with the goal to keep debt levels sustainable. We have given

evidence that deﬁcits in Europe provide only limited informat ion on the evolution of debt

levels in the past.

We then have tested the hypothesis that governments try to circumvent ﬁscal rules by

means of creative accounting. Our empirical evidence indicates that the introduction of the

stability and growth pact and the excessive deﬁcit procedure in Europe have resulted in

creative accounting. While stock-ﬂow adjustments have signiﬁcantly contributed to debt

accumulation in the last 20 years in Europe, only after the introduction of the ﬁscal frame-

work in Europe a systematic relationship between these adjustments and deﬁcits can be

detected. Furthermore, this use of creative accounting is especially responsive to cyclical

parts of the deﬁcit, where the associated costs of the non-state-contingent ﬁscal rule are

high. The use of creative accounting is especially measurable, when the ﬁscal rules become

binding. Our results conﬁrm the vulnerability of ﬁscal rules due to creative accounting.

5. Acknowledgments

We thank the anonymous referees, Jo

rg Breitung, Kirsten Heppke-Falk, Heinz Herr-

mann, Jana Kremer, Wolfgang Lemke, Rolf Strauch, and the ﬁscal policy departments

The same results are obtained when using deﬁcits as the explanatory variable.

A separate estimation of the cyclical eﬀects for binding vs. non-binding ﬁscal rules was not possible because

the relatively short sample period yields only few observations.

J. von Hagen, G.B. Wolﬀ / Journal of Banking & Finance 30 (2006) 3259–3279 3275

Table A.2

Robustness check: measuring the impact of ﬁscal rules with diﬀerent methodologies

OLS FE PCSE Db

OLS FE PCSE PCSE PCSE

sfa 0.96 1.07 0.98 Deﬁcit 0.98 1.04 1.01

0.09 0.07 0.05 0.05 0.06 0.08

T 3.56 3.63 2.37 T 0.4 0.16 0.2 0.53 0.52

0.54 0.38 0.78 0.4 0.39 0.57 0.61 0.59

T · sfa 0.53 0.41 0.32 T · deﬁcit 0.46 0.46 0.47

0.19 0.13 0.11 0.12 0.11 0.13

Cons 4.34 4.15 4.13 Cons 1.73 1.48 1.07 1.11 1.24

0.3 0.21 0.93 0.28 0.29 0.66 0.52 0.53

CAB 1 0.95

0.069 0.07

T · CAB 0.34 0.34

0.15 0.15

CD 1.21 1.39

0.17 0.17

T · CD 0.98 1.12

0.36 0.34

0.43 0.61 0.71 R

0.69 0.67 0.71 0.79 0.79

Obs 293 293 293 Obs 293 293 293 293 293

Country dummies No No Yes Country dummies No No Yes Yes Yes

Output Trend Potential

Note: Dependent variable: change in the debt relative to GDP. T = 1 if year > 1997. In the panel corrected

standard error (PCSE) regressions, we took account of possible autocorrelation in the error term. FE refers to

standard ﬁxed eﬀect regressions.

Table A.1

Measuring the impact of ﬁscal rules

12 3 4 5 6 7

Deﬁcit 0.85 0.88 0.91 0.91 0.81 0.84 0.88

0.07 0.08 0.09 0.08 0.08 0.08 0.10

T 1.81 1.77 1.87 1.91 1.50 2.14 1.38

0.46 0.48 0.52 0.41 0.49 0.53 0.65

T · deﬁcit 0.32 0.27 0.23 0.27 0.28 0.31 0.36

0.10 0.12 0.15 0.11 0.11 0.12 0.12

Cons 0.19 0.16 0.17 0.16 0.17 0.23 0.18

0.03 0.04 0.04 0.03 0.04 0.04 0.08

LDV 0.17 0.20 0.18 0.22 0.18 0.17 0.08

0.05 0.06 0.06 0.05 0.06 0.06 0.07

Obs 263 233 221 220 212 221 183

Sargan p 1 1.00 1.00 1.00 1.00 1.00 1.00

Autocorr 2, p 0.76 0.34 0.34 0.44 0.68 0.83 0.80

Omitted SE, FI SE, FI, LU SE, FI, GR SE, DK, UK DE, FR <1991

Note: Dependent variable: change in the debt relative to GDP. T = 1 if year > 1997. LDV refers to the lagged

dependent variable. Last line refers to which observations were omitted. Standard errors are reported below the

coeﬃcients. Method: Arellano bond dynamic GMM panel estimator.

3276 J. von Hagen, G.B. Wolﬀ / Journal of Banking & Finance 30 (2006) 3259–3279

of the ECB and the Deutsche Bundesbank for many, very helpful discussions. Research

assistance by Sascha Heise is gratefully acknowledged. Remaining errors are ours. The

opinions expressed in this paper do not necessarily reﬂect the views of the Deutsche Bun-

desbank or its staﬀ.

Appendix

See Tables A.1–A.5.

Table A.4

Robustness check: measuring the impact of ﬁscal rules

PCSE PCSE PCSE PCSE PCSE PCSE

Deﬁcit 1.11 1.14 1.18 0.97 1 1.06

0.09 0.10 0.08 0.08 0.08 0.11

T 0.11 0.23 0.32 0.36 0.23 0.61

0.53 0.66 0.51 0.51 0.68 0.6

T · deﬁcit 0.44 0.38 0.49 0.41 0.48 0.5

0.12 0.17 0.11 0.13 0.15 0.16

Cons 0.64 0.46 0.35 1.24 1.11 0.46

0.64 0.69 0.63 0.63 0.73 0.61

0.75 0.75 0.75 0.74 0.71 0.74

Obs 259 245 244 236 247 189

Omitted SE, FI SE, FI, LU SE, FI, GR DE, FR SE, DK, UK <1990

Note: Dependent variable: change in the debt relative to GDP. T = 1 if year > 1997. All regressions include

country dummies. In the panel corrected standard error (PCSE) regressions we took account of possible auto-

correlation in the error term.

Table A.3

Robustness check: measuring the impact of ﬁscal rules

PCSE PCSE PCSE PCSE PCSE PCSE

sfa 1.00 1.02 1.09 0.94 0.98 1.03

0.06 0.06 0.07 0.06 0.06 0.06

T 2.31 2.56 2.13 2.18 2.94 2.82

0.69 0.70 0.64 0.69 0.82 0.79

T · sfa 0.39 0.39 0.41 0.33 0.3 0.44

0.12 0.14 0.14 0.13 0.12 0.11

Cons 4.08 4.25 3.92 4.03 4.45 4.35

0.84 0.85 0.84 0.87 0.88 0.78

0.75 0.76 0.71 0.71 0.72 0.77

Obs 259 245 244 236 247 189

Country dummies Yes Yes Yes Yes Yes Yes

Omitted SE, FI SE, FI, LU SE, FI, GR DE, FR SE, DK, UK <1990

Note: Dependent variable: change in the debt relative to GDP. T = 1 if year > 1997. In the panel corrected

standard error (PCSE) regressions we took account of possible autocorrelation in the error term.

J. von Hagen, G.B. Wolﬀ / Journal of Banking & Finance 30 (2006) 3259–3279 3277

References

Arellano, M., Bond, S., 1991. Some tests of speciﬁcation for panel data: Monte Carlo evidence and an application

to employment equations. Review of Economic Studies 58, 277–297.

Balassone, F., Franco, D., Zotteri, S., 2004. EMU Fiscal indicators: a misleading compass? WP.

Beck, N., Katz, J., 1995. What to do (and not to do) with time-series and cross-section data. American Political

Science Review 89, 634–647.

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

Economic Literature 24, 1644–1678.

Bohn, H., Inman, R.P., 1996. Balanced budget rules and public deﬁcits: evidence from the US states. NBER

Working Paper, 5533.

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

Choice 68, 57–69.

Daﬄon, B., Rossi, S., 1999. Public accounting fudges towards EMU: a ﬁrst empirical survey and some public

choice considerations. Public Choice 101, 59–84.

ECB, 2004. Monthly Bulletin March. European Central bank, Frankfurt am Main.

European Commission, 2004. Public Finances in EMU 2004. The European Commission, Brussels.

European Commission, DG for Economic and Financial Aﬀairs, 2003. Public Finances in EMU 2003. The

European Commission, Brussels.

Eurostat, 2002. ESA95 Manual on Government Deﬁcits and Debt. Oﬃce for Oﬃcial Publications of the

European Communities, Luxembourg.

Greene, W.H., 2000. Econometric Analysis. Prentice Hall International, London.

Table A.5

Robustness check: measuring the impact of binding ﬁscal rules

FE AB FE AB FE AB FE AB

sfa 0.80 0.74 0.40 0.50 0.38 0.46 0.39 0.49

0.20 0.13 0.12 0.08 0.13 0.09 0.12 0.09

R · sfa 0.21 0.12 0.53 0.34 0.58 0.43 0.53 0.36

0.21 0.13 0.13 0.09 0.14 0.10 0.13 0.09

D · sfa 0.29 0.21 0.12 0.04 0.09 0.07 0.20 0.04

0.28 0.18 0.15 0.10 0.16 0.11 0.16 0.11

R · D · sfa 0.83 0.52 0.95 0.24 1.18 0.65 0.96 0.43

0.39 0.25 0.35 0.22 0.32 0.21 0.32 0.21

R 3.04 1.99 3.65 2.26 3.37 1.89 4.29 2.69

0.44 0.29 0.39 0.27 0.44 0.31 0.44 0.31

D 1.79 0.47 1.55 0.61 1.40 0.13 1.70 0.79

0.44 0.35 0.38 0.31 0.46 0.35 0.45 0.35

R · D 0.43 0.19 0.70 0.19 0.87 0.22 2.30 1.07

0.99 0.61 0.97 0.62 1.01 0.65 1.24 0.79

Cons 2.76 0.01 1.76 0.03 2.38 0.02 1.91 0.03

0.35 0.02 0.27 0.02 0.33 0.03 0.32 0.03

LDV 0.39 0.42 0.43 0.39

0.03 0.03 0.03 0.03

Obs 245 221 278 250 236 212 247 221

0.77 0.73 0.73 0.77

Sargan 0.86 0.85 0.99 0.94

Autocorr 2 0.67 0.90 0.78 0.82

Omitted FI, LU, SW GR DK, SW, UK DE, FR

Note: Dependent variable: change in the debt relative to GDP. T = 1 if year > 1997. R =1ifDb > 3, LDV refers

to the lagged dependent variable. Standard errors are reported below the coeﬃcients. Methods: panel ﬁxed eﬀect

(FE) and Arellano bond dynamic GMM panel estimator (AB).

3278 J. von Hagen, G.B. Wolﬀ / Journal of Banking & Finance 30 (2006) 3259–3279

Gujarati, D.N., 1995. Basic Econometrics, third ed. McGraw-Hill Book Co., New York.

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

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

Laughland, J., Paul, J.-M., 1997. Belgium cooks its EMU books. The Wall Street Journal Europe (January 9).

Milesi-Ferretti, G., 2003. Good, bad or ugly? On the eﬀects of ﬁscal rules with creative accounting. Journal of

Public Economics 88, 377–394.

Milesi-Ferretti, G., Moriyama, K., 2004. Fiscal Adjustment in EU Countries: A Balance Sheet Approach. IMF

Working Paper, 04/143.

nster, W., 1997. Das Spiel mit den Deﬁziten., Su

ddeutsche Zeitung, 23. August. In: Deutsche Bundesbank,

Auszu

ge aus Presseartikeln.

Poterba, J.M., 1994. State responses to ﬁscal crises: the eﬀects of budgetary institutions and politics. Journal of

Political Economy 102 (4), 799–821.

Sbragia, A., 1996. Debt Wish. University of Pittsburgh Press, Pittsburgh.

Strauch, R.R., 1998. Budget processes and ﬁscal discipline: evidence from the US states. ZEI working paper.

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

44, 199–210.

von Hagen, J., 2002. Fiscal rules, ﬁscal institutions, and ﬁscal performance. The Economic and Social Review 33

(3), 263–284.

von Hagen, J., Harden, I.J., 1995. Budget processes and commitment to ﬁscal discipline. European Economic

Review 39, 771–779.

von Hagen, J., Harden, I.J., 1996. Budget Processes and Commitment to Fiscal Discipline. IMF Working Paper,

96/78.

J. von Hagen, G.B. Wolﬀ / Journal of Banking & Finance 30 (2006) 3259–3279 3279