Available online at www.sciencedirect.

com

Journal of Banking & Finance 32 (2008) 217–228

www.elsevier.com/locate/jbf

Cross-country determinants of bank income smoothing

by managing loan-loss provisions
Ana Rosa Fonseca *, Francisco González
University of Oviedo, School of Economics and Business, Spain

Received 27 February 2006; accepted 27 February 2007

Available online 6 July 2007

Abstract

This paper studies the determinants of income smoothing by management of loan-loss provisions in banks around the world. Using a
panel database of 3221 bank-year observations from 40 countries and controlling for unobservable bank effects and for the endogeneity
of explanatory variables, we find that bank income smoothing depends on investor protection, disclosure, regulation and supervision,
financial structure, and financial development. Results suggest there is less bank income smoothing not only with the strength of investor
protection, but also with the extent of accounting disclosure, restrictions on bank activities, and official and private supervision, while
there is more income smoothing with market orientation and development of a country’s financial system.
 2007 Elsevier B.V. All rights reserved.

JEL classification: G34; G38; M41

Keywords: Income smoothing; Loan-loss provisions; Bank regulation; Bank supervision; Institutions

1. Introduction acquire private-control-benefits, which diminishes the

incentives to conceal a firm’s performance.
Bank income smoothing in the form of managing loan- We ask whether Leuz et al.’s findings (2003) on earnings
loss provisions (LLP) varies from country-to-country management in the industrial firms are applicable to highly
depending on variables such as investor protection, disclo- leveraged firms like banks. Our analysis uses a panel data-
sure, regulation, supervision, financial structure, and finan- base of 3221 bank-year observations to analyze the influ-
cial development. Ball et al. (2000) and Fan and Wong ence on income smoothing by managing LLP of investor
(2002) highlight the role of regulation and legal enforce- protection, accounting disclosure, restrictions on bank
ment in explaining international differences in the quality activities, official and private supervision, and financial
of financial statements. The institutional environment has structure and development. We also compare income
a direct effect on earnings management according to Leuz smoothing in publicly traded and non-publicly traded
et al. (2003), who report differences in earnings manage- banks. We focus on bank manipulation of LLP because
ment of publicly traded firms across 31 countries. They most of the empirical literature in banking has analyzed
conclude that earnings management declines with investor LLP for two basic reasons. First, banks have substantial
protection; strong protection limits insiders’ ability to latitude in determining the amount of provisions. Second,
banks’ high leverage makes them quite vulnerable to vola-
tility in asset values, prompting adequate LLP, which
*
become banks’ main accrual; this has important effects
Corresponding author. Address: Department of Business Administra-
tion and Accounting, University of Oviedo, Avenida del Cristo S/N, 33071
on bank stability.
Oviedo, Spain. Tel.: +34 985103694; fax: +34 985103708. High leverage and the safety nets intended to avoid
E-mail address: [email protected] (A.R. Fonseca). industry contagion in the event of a bank run give rise to

0378-4266/$ - see front matter  2007 Elsevier B.V. All rights reserved.
doi:10.1016/j.jbankfin.2007.02.012
218 A.R. Fonseca, F. González / Journal of Banking & Finance 32 (2008) 217–228

the well-known moral hazard problem of risk-shifting. If as the exact nature of bank regulation and supervision as
there are greater incentives for bank insiders to shift risk, well as the structure and development of a country’s finan-
so too are there more incentives to engage in earnings man- cial system. Third, we analyze differences between publicly
agement to hide their risk-shifting. Our primary hypothesis and non-publicly traded banks. Finally, we control for
is therefore that the more efficient bank regulation and individual bank effects that are not explained by the vari-
supervision proves to be in limiting bank risk, the fewer ables explicitly included in the regressions and for the end-
the incentives for bank managers to smooth bank earnings. ogeneity of explanatory variables and the dynamic
This analysis is particularly relevant in the evaluation of behavior of LLP.
the effect of the new Basel Capital Accord (Basel II) on The rest of the paper is structured in the following way.
the reliability of bank financial statements. Basel II empha- Section 2 discusses the hypotheses regarding the differences
sizes the strengthening of regulation (e.g. minimum regula- in income smoothing between publicly and non-publicly
tory capital requirements in Pillar 1) and of supervision by traded banks, and the cross-country determinants of
authorities (Pillar 2), as well as market discipline (Pillar 3) income smoothing. Section 3 describes the database and
as tools to increase bank stability. The approach of the methodology. Section 4 reports the empirical results of
third Pillar of Basel II consists in strengthening market dis- income smoothing in each country and the results of the
cipline by proposing a set of requirements and recommen- cross-country determinants. Finally, Section 5 presents
dations concerning public disclosure practices for banks. our conclusions.
We provide new evidence on the effectiveness of the
requirements set up in the third Pillar of Basel II by analyz-
ing the impact of bank disclosure on income smoothing. 2. Hypotheses
Additionally, when analyzing the influence of official super-
vision on income smoothing, we also provide evidence on There are a number of reasons for income smoothing,
the type of relationship (complements or substitutes?) most of which assume it has negative connotations. Income
between Pillar 2 and Pillar 3. For instance, if official super- smoothing improves the risk perception of a bank for its
vision improves (worsens) the reliability of financial state- investors, regulators, and supervisors. There may also be
ments by reducing (increasing) income smoothing, it also managerial self-interest to smooth earnings. Income
strengthens (weakens) the effectiveness of market discipline smoothing may also be the result of perceived bankruptcy
mechanisms. In this case, Pillar 2 would complement Pillar concerns and/or can be intended to discourage investors
3. from acquiring private information that could then be used
Empirical analysis, by and large US-based, analyzes to trade against uninformed shareholders selling for liquid-
whether earnings before LLP have a positive coefficient. ity reasons. These reasons imply that managers adjust earn-
A positive coefficient would indicate income smoothing, ings figures for subjective reasons, producing some kind of
since it suggests that LLP are high when earnings are high private-control-benefit for insiders that may ultimately
and low when bank earnings are low. Results are mixed for diminish shareholder value (the private-control-benefits
US banks. Greenawalt and Sinkey (1988) and Wahlen hypothesis).1 However, analysis of the uses made of
(1994), among others, find a positive relation between LLP to manage earnings must control for two alterna-
LLP and bank earnings; while Beatty et al. (1995) and tive uses of LLP by bank managers and supervisors: the
Ahmed et al. (1999) find no evidence of earnings smooth- risk-management hypothesis and the capital-management
ing. We extend the study of the LLP-earnings relationship hypothesis.
to an international sample of banks by applying the GMM The risk-management hypothesis emphasizes supervi-
difference estimator to control for unobservable bank sors’ interest in reducing procyclicality of LLP and capital.
effects, and for the endogeneity of explanatory variables It assumes that banks and regulators define a specific level
and the dynamic behavior of LLP. Our results indicate that of protection against credit losses and banks set aside
better investor protection and stricter legal enforcement loan-loss reserves according to the value of expected losses
reduce incentives to smooth earnings in banking. Addi- and raise capital according to unexpected losses. In other
tional evidence shows that incentives to smooth earnings words, credit risk is built up in a boom and materializes
decline with accounting disclosure, restrictions on bank in a downturn, so banks should recognize the underlying
activities, and official and private supervision. Incentives risk and build up loan-loss reserves in good times to be
increase with market orientation and development of a drawn on in bad times. As a result, provisions should there-
country’s financial system. fore move with income (income-smoothing pattern) and
Shen and Chich (2005) have also analyzed earnings with the economic cycle to return the ratio to its ideal (equi-
management in an international bank sample. Their librium) value every time it is modified by a random shock.
research is substantially different from ours in several ways. Seen from this perspective, bank supervisors point out that
First, Shen and Chich look at earnings management in gen- the LLP-earnings link has a positive effect on banks, which
eral, while we focus on the use of LLP, the main bank clashes with the negative connotation suggested by the
accrual, to smooth earnings. Second, we include in the
1
analysis the influence of additional country variables, such See Goel and Thakor (2003) for a more detailed review.
 A.R. Fonseca, F. González / Journal of Banking & Finance 32 (2008) 217–228 219

private-control-benefits hypothesis for income smoothing The opposite case can also be argued. Smaller banks,
in the industrial sector. which are usually non-publicly traded enterprises, generally
However, empirical evidence fails to show any counter- have fewer diversification opportunities. This could easily
cyclical behavior of LLP. Laeven and Majnoni (2003) and accentuate the risk-shifting conflict caused by high leverage
Bikker and Metzemakers (2005) have analyzed in an inter- and safety nets, thereby increasing the incentives to cover
national sample of banks the relation between LLP and up risk shifting with more income smoothing. Moreover,
economic cycle using both earnings and GDP growth as if supervisors focus more on the bigger banks because they
explanatory variables. Although a positive coefficient for are so important in the case of a banking crisis, they would
the earning variable is consistent with income smoothing, presumably be better able to understand the effects of
the research does not support the countercyclical view, as accounting practices in publicly traded banks. This would
LLP are negatively related to GDP growth. Rather, both reduce incentives to manage financial statements in pub-
papers generally report procyclical behavior when running licly traded banks. Although the argument is not applicable
separate regressions for five regions and seven countries. to every country, differences across countries in official
Following these papers, we include the growth of per capita oversight of publicly and non-publicly traded banks would
GDP as an explanatory variable to control for the potential at least imply changing differences in income smoothing
countercyclical effect of LLP suggested by the risk-manage- between these two groups of banks across countries.
ment hypothesis. The contradictory predictions of hypotheses make dif-
The traditional capital-management hypothesis states ferences in income smoothing between publicly and non-
that bank managers use LLP to reduce expected regulatory publicly traded banks an empirical issue, and banks may
costs associated with violating capital requirements, a neg- have different practices from those in the industrial sector.
ative relationship being predicted between capital ratios Only Beatty and Harris (1999) and Beatty et al. (2002) have
and LLP (Ahmed et al., 1999). However, caution is called shown that publicly traded banks engage in income
for in forecasting this for our sample for several reasons. smoothing more than privately owned banks in the US.
First, the implementation of Basel I in our period of anal- To our knowledge, there is no empirical evidence on the
ysis reduced incentives for low-capital banks to increase differences in the use of LLP to smooth earnings between
LLP, given that provisions cannot be included in TIER 1 publicly traded and private banks outside the US; nor is
capital, and only general provisions can be included in there any empirical evidence on whether these differences
TIER 2 capital and must comply with the 1.25 per cent are stable across countries.
limit. Second, although Basel I was in force in most coun-
tries, according to Barth et al. (2004), countries altered and 2.2. Cross-country determinants of income smoothing
fine-tuned their definitions of regulatory capital, which
could lead to differences across countries in the ability of Possible cross-country determinants of income smooth-
LLP to manage capital and thereby disenable clear fore- ing include: the influence of investor protection, transpar-
casting for the whole set of countries.2 Such differences in ency in accounting disclosures, bank regulation and
capital regulation are not germane to our study as we are supervision, and financial structure and financial develop-
interested in analyzing the use of LLP to manage earnings ment of a country.3
after controlling for it being used to manage capital, but
not in testing the capital-management hypothesis across 2.2.1. Investor protection
countries. By including bank capital as an explanatory var- Investor protection is defined as the power to expropri-
iable of LLP, without specifying a particular empirical rela- ate minority shareholders and creditors within the con-
tionship, we achieve this control. straints imposed by law (La Porta et al., 2002). We use
three variables drawn from La Porta et al. (1998) to repre-
2.1. Publicly traded vs. non-publicly traded banks sent investor protection: rights of minority shareholders
(ANTIDIRECTOR); creditor rights (CREDITOR); and
The literature has traditionally forecasted that publicly legal enforcement (LEGAL). Higher values of these vari-
traded firms engage in more income smoothing. As publicly ables indicate stronger protection of minority shareholders
traded firms have more outsiders, earnings announcements and creditors and stronger legal enforcement.
and financial statements have a greater signaling effect Leuz et al. (2003) have shown that earnings management
(Beatty et al., 2002). Explanations of income smoothing is more pervasive for commercial and industrial firms in
based on managerial self-interest and on trading costs for countries where the legal protection of minority sharehold-
uninformed shareholders would also suggest publicly ers and legal enforcement are weak, because insiders enjoy
traded firms have more incentives for income smoothing.
3
Cavallo and Majnoni (2001) also uncover evidence of the influence of
the legal and regulatory framework on LLP. Although their results can be
2
For instance, Shrieves and Dahl (2003) and Pérez et al. (2004) show interpreted as largely consistent with our evidence, they focus on the
differences in the treatment of LLP in the definitions of regulatory capital influence of institutional variables on the amount of provisions, while we
in Japan, Spain, and the US. focus on their influence on the relation between provisions and earnings.
220 A.R. Fonseca, F. González / Journal of Banking & Finance 32 (2008) 217–228

greater private-control-benefits and hence have stronger bank activities may reduce the opportunities for smoothing
incentives to obfuscate bank performance. In banking, earnings using other discretionary components of bank
there are additional arguments that would suggest investor income such as security gains or losses (Beatty et al.,
protection would have a negative influence on income 1995; Shrieves and Dahl, 2003). The fact that in this case
smoothing. Demirgüc-Kunt and Detragiache (2002) have there are fewer alternatives for smoothing earnings may
shown that a sound legal system with proper enforcement mean that LLP are applied more often toward this end,
of rules reduces the adverse effects of deposit insurance in which case stricter regulations on bank activities would
on bank risk-taking. This lower risk-taking in countries be positively related to income smoothing.
with strong institutional environments would also diminish We use the measure of regulatory restrictions on non-
bank incentives to smooth income which is stable per se. traditional bank activities (securities, insurance, real estate,
Shen and Chich (2005) report a negative relation in bank- and bank ownership and control of non-financial firms)
ing between the rights of the minority shareholders and developed by Barth et al. (2001). Values range from 4 to
earnings management, but do not find a negative influence 16; higher values indicate more restrictions on bank activ-
for the quality of legal enforcement. Given these argu- ities and non-financial ownership and control.
ments, we expect ANTIDIRECTOR and LEGAL to have
a negative influence on income smoothing. We also predict 2.2.4. Bank supervision
that CREDITOR will reduce income smoothing, as stron- We consider both official supervisory power (OFFI-
ger creditor rights against borrower expropriation would CIAL) and private supervision (MONITOR) using the
reduce bank risk in lending activities and thus their incen- indicators developed by Barth et al. (2001). Official super-
tives to smooth earnings. visory power, ranging from 0 to 14, captures the power
of supervisors to take prompt corrective action, to restruc-
2.2.2. Accounting quality ture and reorganize troubled banks, and to declare a trou-
We forecast that a poor accounting system will increase bled bank insolvent. Private oversight, ranging from 0 to 6,
bank income smoothing for two reasons. First, the less measures the intensity of audit requirements and whether
detailed financial statements are, the greater the opportu- subordinated debt is allowable as a part of regulatory cap-
nity to smooth profits reported to investors and supervi- ital. Higher values of OFFICIAL and MONITOR indicate
sors. Second, a poor accounting system that makes it greater power of supervisors and greater private oversight.
difficult for bank lenders to assess the total risk of borrow- If supervisors have greater powers to intervene in banks
ers will increase the problem of asymmetric information to discipline managers and reduce their incentives to under-
between them, thus increasing bank risk. Greater risk take risk, they will also reduce managers’ incentives to use
would create more incentives to smooth earnings. LLP to smooth benefits that are not highly volatile. For
The third Pillar of Basel II assumes that greater disclo- this reason, we expect OFFICIAL to have a negative influ-
sure requirements are effective in strengthening market dis- ence on income smoothing. As there are fewer opportuni-
cipline. Our analysis provides evidence on this issue by ties for banks to smooth their earnings, with increased
directly testing whether greater disclosure increases the reli- private monitoring, we expect MONITOR to have a nega-
ability of bank financial statements by reducing income tive influence on income smoothing.
smoothing. We use the accounting disclosure index (DIS-
CLOSURE) from La Porta et al. (1998) as our indicator 2.2.5. Financial structure
of accounting quality. This index measures the inclusion We incorporate the influence of financial structure by
or the omission of 90 items in the 1990 annual reports, analyzing the comparative importance of stock markets
where higher values indicate more disclosure. Therefore, and banks in a country. The relation between financial
we expect DISCLOSURE to have a negative influence on structure and bank income smoothing may have a number
income smoothing. of root causes. More dispersed bank ownership in market-
oriented financial systems (La Porta et al., 1999; La Porta
2.2.3. Bank regulation et al., 2002) may boost incentives for bank managers to
We incorporate the characteristics of bank regulation in smooth earnings, as the greater number of users of finan-
each country in the analysis through a measure of the cial statements makes accounting figures more important
breadth of activities permitted to banks (RESTRICT). and managers have more reasons to want to influence
There is no clear prediction for the effect of RESTRICT. external perception of the bank’s solvency. This argument
On the one hand, tighter regulations on bank activities predicts a positive influence of market orientation on bank
should reduce both opportunities for taking risk and bank income smoothing. The opposite prediction could be made
competition (Claessens and Laeven, 2004). Reduced bank if financial structure were considered an endogenous
competition should lessen risk-taking incentives to preserve variable. The empirical literature has demonstrated that
the higher charter value of banks in less competitive mar- market-oriented financial systems are more likely to repre-
kets (Keeley, 1990). These arguments would suggest fewer sent high-quality institutional environments with strong
incentives for income smoothing, the tighter regulations on investor protection and good enforceability of contracts
bank activities. On the other hand, stricter limitations on (La Porta et al., 1998). Considered in this light, market
 A.R. Fonseca, F. González / Journal of Banking & Finance 32 (2008) 217–228 221

orientation, like investor protection, ought to be negatively 3.1. Income smoothing

related to income smoothing. However, this negative rela-
tion should disappear when the potential endogeneity of That it is important to use the right methodology to
financial structure is controlled for and only its exogenous test the income-smoothing hypothesis is seen in the contra-
component is considered. dictory outcomes of the many US studies. We apply the
We use the structure-aggregate variable (STRUCT) generalized-method-of-moments (GMM) estimators devel-
described by Beck and Levine (2002), defined as the first oped for dynamic models of panel data by Arellano and
principal component of two variables that measure the Bond (1991). This methodology is designed specifically to
comparative activity and size of markets and banks. Each address three relevant econometric issues: (i) the presence
of the components is constructed so that higher values indi- of unobserved bank-specific effects, which is eliminated
cate more market-oriented financial systems. The first com- by taking first-differences of all variables; (ii) the autore-
ponent is the natural logarithm of the ratio of value traded gressive process in the data regarding the behavior of
to bank credit, and the second component equals the nat- LLP (i.e., the need to use a lagged dependent variables
ural logarithm of the ratio of market capitalization to bank model to capture the dynamic nature of the LLP); and
credit. The data to calculate this variable come from the (iii) the likely endogeneity of the explanatory variables.
Beck et al. (2003) database. The panel estimator controls for potential endogeneity by
using instruments based on lagged values of the explana-
2.2.6. Financial development tory variables. Of the recent empirical studies, only Laeven
As in the case of financial structure, we are unable to and Majnoni (2003) in an international bank sample and
predict a clear influence of financial development on bank Pérez et al. (2004) in a Spanish bank sample use this
income smoothing. Again, more widely dispersed owner- estimator.
ship in more financially developed countries increases the The variables chosen as possibly explanatory of LLP are
number of users of financial statements and gives bank variables traditionally used for the income-smoothing
managers incentives to improve the external perception of hypothesis (see, for example, Greenawalt and Sinkey,
a bank’s solvency by using income smoothing. However, 1988) and we incorporate additional variables to those
if we assume financial development is an endogenous vari- employed in an international bank sample to analyze the
able that increases with market orientation and strong procyclicality of bank provisioning (the risk-management
investor protection (Levine, 1999), we would expect it to hypothesis). Unlike Laeven and Majnoni (2003), we
have the same negative influence on bank income smooth- include bank capital to control for the capital-management
ing as we investor protection, although the negative influ- hypothesis and country dummies to control for differences
ence of financial development should disappear when its in the level of LLP across countries. Unlike both Laeven
endogeneity is controlled for and only its exogenous com- and Majnoni (2003) and Bikker and Metzemakers (2005),
ponent is analyzed. we include the loan-loss reserve to control for non-discre-
We follow Beck and Levine (2002) and use a Finance- tionary components of LLP. The model thus estimated is
Aggregate index (FINAN) to gauge the extent of a country’s      
financial development. This index equals the first principal LLPi;t LLPi;t1 LLPi;t2
¼ b0 þ b1 þ b2
component of two variables that measure the overall activity Ai;t1 Ai;t2 Ai;t3
   
and size of financial intermediaries and markets, where EBTi;t CLOANSi;t
þ b3 þ b4
higher values indicate a more developed financial system. Ai;t1 Ai;t1
The data to calculate this variable come from the Beck    
LLAi;t1 CAPi;t
et al. (2003) database. þ b5 þ b6
Ai;t1 RWAi;t
3. Bank data and methodology X40
þ b7 GDPGR þ b8 Countryj
j¼1
We obtain consolidated bank balance-sheet and income-
X
2002
statement data (in US dollars and in real prices) from þ b9 T t þ mi þ eit ; ð1Þ
Fitch-IBCA Ltd. BankScope Database for 1995–2002. t¼1995
We use information for banks in 40 countries.4
where LLPi,t is the loan-loss provision of bank i at year t.
Lags in the dependent variable (LLPi,t1 and LLPi,t2) cap-
ture adjustment costs that constrain complete adjustment
4
Our initial sample was made up of the 49 countries included in the La to an equilibrium level. We include the first and the second
Porta et al. (1998) database for which information about legal character- lag to take into account a change in the speed of adjust-
istics and investor protection was available. Austria, Belgium, Finland,
ment beyond the first year. We expect positive coefficients
India, Indonesia, Netherlands, Taiwan, and Zimbabwe were excluded
because of lack of data on macroeconomic or bank variables. Finally, we for the lags.
exclude US banks to avoid a potential bias caused by the high percentage EBTi,t, earnings before taxes and LLP, is the most
that they represent of the sample (1325 observations – 29% of the sample). interesting variable in our study, as it measures income
222 A.R. Fonseca, F. González / Journal of Banking & Finance 32 (2008) 217–228

smoothing; the higher its positive coefficient, the more the beginning balance of total allowance for loan-loss, but
income smoothing there will be. Change in total loans correlates negatively with bank capital.
outstanding (CLOANSi,t) and the beginning balance of
total allowance for loan-losses (LLPi,t) control for non-dis-
3.2. Cross-country determinants of income smoothing
cretionary components of LLP, since these variables are
related to changes in default risk. Following Greenawalt
To test the influence of country variables, we sequen-
and Sinkey (1988) and Wahlen (1994), we expect positive
tially incorporate an interaction term for each country var-
coefficients for both variables.5 All variables (LLP, EBT,
iable and the earnings variable into Eq. (1). The coefficient
CLOANS, and LLA) are normalized by the total bank
of each interaction term thus measures the influence of the
assets at the beginning of year t (Ai,t1) to mitigate poten-
particular political-economy variable on bank income
tial estimation problems with heteroscedasticity.
smoothing. The paucity of instruments, the extensive num-
We include the bank capital normalized by risk-
ber of country variables, and the need to use interaction
weighted assets (CAPi,t/RWAi,t) to control for the poten-
terms with the earnings variable supports incorporation
tial use of capital management. We use TIER 2 in the
of the coefficients separately rather than at the same time.7
reported results, though results do not change using TIER
A major stumbling block when analysis includes several
1. The annual growth of real per capita GDP (GDPGRi,t)
political-economy variables is separating out the effects and
is included to control for the documented procyclical effect
the correlated outcomes. The correlations between the
of provisioning (Laeven P40and Majnoni, 2003; Bikker and political-economy variables in our sample in Table 2 cor-
Metzemakers, 2005). j¼1 Countryj is a set of country roborate the positive relations documented in the literature
dummy variables controlling for specific differences in the
P2002 between LEGAL, STRUCT, and FINAN and also reveal
level of LLP across countries. t¼1995 T t is a set of dummy
that these three aspects correlate positively with DISCLO-
time variables. These dummies capture any unobserved
SURE. Such interrelations and the potential endogeneity
bank-invariant time effects not included in the regression.
of political-economy variables make it difficult to tease
Finally, mi are unobservable bank-specific effects that are
out the specific effect of each variable and to know which
constant over time but vary across banks, while eit is the
of them plays the major role in bank income smoothing.
white-noise error term.
Our empirical analysis uses a number of instruments for
We control for the potential endogeneity of EBT, CLO-
the observed values of each country variable to identify the
ANS, LLA, and CAP in the GMM estimations using two-
exogenous component of the variable and control for
to-four-period lags of the same variables as instruments.
potential simultaneity bias. The instruments are defined
The growth of per capita GDP, the country and the time
following Leuz et al. (2003): the country’s real GDP aver-
dummy variables are the only variables considered exoge-
aged over 1980 to 1989, and four binary variables indicat-
nous.6 As the consistency of the GMM estimator depends
ing English, German, French or Scandinavian legal origin
on the validity of the instruments, we consider two specifi-
according to the classification of La Porta et al. (1998).
cation tests suggested by Arellano and Bond (1991). The
This methodology lets us to focus on the influence of the
first is a Sargan test of over-identifying restrictions, which
exogenous component of each political-economy variable.
tests the overall validity of the instruments. This test con-
Thus correlations between the observed values for the
firms the absence of correlation between the instruments
political-economy variables need not remain when we ana-
and the error term in our models. The second test examines
lyze only their exogenous components.
the hypothesis of absence of second-order serial correlation
in the first-difference residuals (m2). In our models, this
hypothesis of second-order serial correlation is always 4. Results
rejected. Although there is first-order serial correlation
(m1) in the differentiated residuals, it is due to first differ- 4.1. Results of income smoothing
ences in models.
Table 1 reports descriptive statistics and correlations in Before analyzing differences across countries, we first test
our bank sample. The correlations in Panel B show that, on the income-smoothing hypothesis on the complete sample of
average, LLP correlates positively with bank earnings and banks. Results indicate income smoothing, as EBT has posi-
tive coefficients (statistically significant at the 1% level) in all
estimations. Coefficients for the remaining variables are as
5 expected; the two lags of the dependent variable have
Results do not change when we include non-performing loans as an
additional proxy for the non-discretionary components of LLP, but the positive coefficients, indicating that a dynamic specification
smaller sample size due to missing data advises against reporting these to model bank provisioning is recommended. The proxy
results. variables for the non-discretionary components of LLP
6
Results do not vary by applying random-effect estimations. A random- (CLOANS and LLA) have the expected positive coefficients.
effect estimation does not control for the potential endogeneity of
explanatory variables or for adjustment costs. It has the advantage,
7
however, of employing more observations and of also controlling for Barth et al. (2004) use a similar sequential procedure to analyze the
unobservable heterogeneity. influence of regulatory and supervisory practices on bank development.
 A.R. Fonseca, F. González / Journal of Banking & Finance 32 (2008) 217–228 223

Table 1
Summary statistics
Panel A: Descriptive statistics
Country Median LLP Median EBT Median CLOANS Median LLA Median CAP Median GDPGR # Observations # Banks
Argentina 0.014 0.014 0.002 0.028 0.241 0.240 158 58
Australia 0.016 0.013 0.071 0.001 0.070 6.225 87 26
Brazil 0.007 0.035 0.002 0.017 0.248 8.972 289 103
Canada 0.002 0.005 0.038 0.006 0.131 5.478 23 10
Chile 0.006 0.022 0.028 0.013 0.142 5.580 68 23
Colombia 0.014 0.023 0.002 0.020 0.164 7.888 71 23
Denmark 0.003 0.013 0.031 0.019 0.064 3.474 66 38
Ecuador 0.013 0.024 0.125 0.062 0.169 1.739 22 21
Egypt 0.009 0.025 0.026 0.062 0.141 7.028 97 27
France 0.003 0.012 0.025 0.028 0.080 4.040 340 117
Germany 0.003 0.006 0.004 0.015 0.051 2.019 15 6
Greece 0.005 0.019 0.093 0.013 0.150 7.722 30 10
Hong Kong 0.004 0.019 0.002 0.012 0.217 0.699 115 33
Ireland 0.001 0.015 0.053 0.007 0.066 14.637 33 11
Israel 0.004 0.012 0.047 0.019 0.075 8.808 47 13
Italy 0.040 0.015 0.054 0.016 0.108 4.539 251 90
Japan 0.003 0.004 0.006 0.018 0.080 0.551 115 102
Jordan 0.008 0.017 0.032 0.027 0.135 4.075 13 5
Kenya 0.013 0.034 0.038 0.032 0.265 7.414 57 25
Korea 0.011 0.015 0.045 0.021 0.048 8.123 44 16
Malaysia 0.010 0.025 0.020 0.026 0.118 6.186 104 34
Mexico 0.005 0.015 0.017 0.017 0.190 6.140 72 29
New Zealand 0.001 0.018 0.088 0.003 0.052 5.140 26 7
Nigeria 0.015 0.062 0.085 0.044 0.278 15.307 88 34
Norway 0.003 0.013 0.066 0.014 0.074 8.913 35 10
Pakistan 0.004 0.021 0.119 0.024 0.111 8.550 74 21
Peru 0.018 0.026 0.068 0.037 0.125 5.875 50 18
Philippines 0.010 0.017 0.009 0.030 0.118 11.045 72 22
Portugal 0.004 0.013 0.088 0.010 0.083 6.957 63 22
Singapore 0.002 0.016 0.027 0.038 0.170 2.407 27 11
South Africa 0.009 0.026 0.024 0.023 0.190 10.685 42 13
Spain 0.003 0.015 0.052 0.012 0.098 7.016 59 19
Sri Lanka 0.009 0.016 0.031 0.022 0.096 11.776 19 7
Sweden 0.0001 0.009 0.020 0.005 0.047 4.426 23 7
Switzerland 0.001 0.014 0.016 0.001 0.187 2.316 215 84
Thailand 0.008 0.004 0.031 0.061 0.060 4.201 25 10
Turkey 0.003 0.026 0.004 0.010 0.154 18.415 12 10
United Kingdom 0.002 0.013 0.029 0.008 0.184 5.085 239 80
Uruguay 0.008 0.022 0.097 0.009 0.097 2.481 13 7
Venezuela 0.014 0.038 0.041 0.028 0.148 21.084 22 11
Mean 0.011 0.023 0.043 0.035 0.285 7.316
Median 0.005 0.017 0.025 0.017 0.139 5.314
Standard deviation 0.041 0.054 0.216 0.220 1.158 18.547

Panel B: Correlations
VARIABLES LLP EBT CLOANS LLA CAP
LLP 1
EBT 0.285*** 1
CLOANS 0.017 0.086*** 1
LLA 0.664*** 0.371*** 0.107*** 1
CAP 0.044*** 0.029 0.103*** 0.020 1
GDPGR 0.030 0.036** 0.019 0.018 0.008
Panel A reports descriptive statistics. The bank sample consists of 1213 banks in 40 countries. All data are in real US dollar prices and are reported on an
annual basis over 1995–2002. LLP is the loan-loss provision, EBT is earnings before taxes and LLP, CLOANS is the change in total loans outstanding
estimated as the difference of total bank loans between year t and year t  1, and LLA is the beginning balance of the total allowance for loan-losses. All
these variables are normalized by the total bank assets at the beginning of year t (Ai,t1). CAP is bank capital divided by risk-weighted assets. GDPGR is
the growth of real per capita GDP in the bank’s country. Panel B reports the correlation matrix. *** and ** represent significance at the 1% and 5% level,
respectively.

Capital has positive coefficients, contrary to the predictions capita GDP has negative coefficients, confirming the procyc-
of the capital-management hypothesis. Growth of real per lical effect of LLP.
224 A.R. Fonseca, F. González / Journal of Banking & Finance 32 (2008) 217–228

Table 2
Correlations of country variables
ANTIDIRECTOR CREDITOR LEGAL DISCLOSURE RESTRICT OFFICIAL MONITOR STRUCT
ANTIDIRECTOR 1
CREDITOR 0.114 1
LEGAL 0.246 0.052 1
DISCLOSURE 0.435*** 0.193 0.665*** 1
RESTRICT 0.087 0.184 0.368** 0.160 1
OFFICIAL 0.098 0.056 0.276 0.396** 0.182 1
MONITOR 0.184 0.070 0.498*** 0.444*** 0.344** 0.307 1
STRUCT 0.340** 0.128 0.416*** 0.627*** 0.109 0.158 0.212 1
FINAN 0.370** 0.038 0.663*** 0.678*** 0.189 0.130 0.425*** 0.821***
Values of ANTIDIRECTOR, CREDITOR, LEGAL, and DISCLOSURE are from La Porta et al. (1998). Values of RESTRICT, OFFICIAL, and
MONITOR are from Barth et al. (2001). All these variables are measured at a specific point in time. In contrast, STRUCT and FINAN are calculated
annually over 1995–2002 using the Beck et al. (2003) database. *** and ** represent significance at the 1% and 5% level, respectively.

To analyze differences across countries, we estimate Column 2 shows the coefficients of the PT · EBT interac-
regression (1) for each country in the sample. Column 1 tion variable for each country. Since some countries have
of Table 3 provides the EBT variable coefficients by coun- a limited number of publicly traded banks, this coefficient
try, which are a measure of bank income smoothing in each cannot be calculated for Canada, Korea, New Zealand,
country. To save space, we report only countries with sta- Turkey, and Uruguay.
tistically significant coefficients. To compare the income Our results confirm different patterns of income smooth-
smoothing of publicly and non-publicly traded banks, we ing across countries. A positive relation between LLP and
incorporate into Eq. (1) an interaction term of the earnings bank earnings in 13 countries (Brazil, Chile, Denmark,
variable (EBT) and a dummy variable that takes the value Egypt, Italy, Kenya, Korea, Peru, Philippines, Portugal,
of 1 if the bank was publicly traded and 0 otherwise (PT). Spain, Sweden, and Venezuela) is consistent with the

Table 3
Income smoothing across countries
Income smoothing (1) Sargan test Difference between publicly and Sargan test # Observations # Banks
non-publicly traded banks (2)
Brazil 0.1018*** (3.25) 90.10 0.1374 (0.74) 91.61 289 103
Chile 0.0516 (1.00) 58.23 0.5809** (2.01) 55.96 68 23
Colombia 0.6804*** (8.82) 60.78 0.8094*** (3.86) 58.59 71 23
Denmark 0.6138*** (4.47) 44.96 0.5082 (1.59) 57.13 66 38
Egypt 0.4347*** (6.58) 19.80 0.5585*** (4.79) 100.13 97 27
Greece 0.0386 (0.32) 21.51 0.3069*** (4.17) 30.45 30 10
Italy 0.0640** (2.12) 79.43 0.1168** (1.92) 72.99 251 90
Kenya 0.1750 (1.21) 47.03 0.5725* (1.79) 54.88 57 25
Korea 0.2582*** (2.82) 32.42 – – 44 16
Malaysia 0.3122** (2.14) 81.17 0.2899 (0.35) 81.35 104 34
Pakistan 0.1840** (2.46) 53.54 0.0655 (0.44) 52.78 74 21
Peru 0.6300*** (4.72) 36.62 0.3074* (1.68) 37.72 50 18
Philippines 0.5237*** (5.46) 55.61 0.1924 (1.59) 53.48 72 22
Portugal 0.0675** (2.57) 68.63 0.2635** (2.00) 67.98 63 22
Spain 0.0250 (0.66) 47.40 0.1738** (2.20) 48.07 59 19
Sweden 0.3861* (1.72) 9.00 0.1061 (0.21) 8.29 23 7
Thailand 2.4646** (2.54) 14.30 3.4513*** (2.94) 13.39 25 10
United Kingdom 0.1734*** (3.56) 74.91 0.2787 (0.88) 70.47 239 80
Venezuela 0.4446*** (3.99) 11.88 0.3015 (1.46) 9.18 22 11
These results are for countries with statistically significant coefficients. The measure of income smoothing across countries shown in Column 1 is obtained
using the Arellano and Bond (1991) GMM difference estimator for panel data with lagged dependent variables. In each country regression, the dependent
variable is the ratio of LLP over lagged total assets. As explanatory variables, we include two lags of the dependent variable, profit before taxes and LLP
over lagged total assets (EBT), the change in total loans outstanding over lagged total assets (CLOANS), the beginning balance of the total allowance for
loan-loss over lagged total assets (LLA), bank capital over risk-weighted assets (CAP), real growth in per capita GDP (GDPGR), bank-specific fixed
effects, and year country dummies. The coefficient of EBT is the measure of income smoothing. Column 2 shows the coefficients of the interaction variable
PTxEBT when this variable is added to the regression in Column 1; it captures the difference in the income smoothing between publicly and non-publicly
traded banks. PT is a dummy variable that takes the value of one for publicly traded banks and zero otherwise. Regressions are estimated for each country
for 1995–2002. Year dummy variables are included in all estimations, but are not reported. Coefficients for Argentina, Australia, Canada, Ecuador,
France, Germany, Hong Kong, Ireland, Israel, Japan, Jordan, Mexico, New Zealand, Nigeria, Norway, Singapore, South Africa, Sri Lanka, Switzerland,
Turkey, and Uruguay are not statistically significant and are not reported in the table. T-statistics are in parentheses. ***, **, and * represent significance at
the 1%, 5%, and 10% level, respectively.
 A.R. Fonseca, F. González / Journal of Banking & Finance 32 (2008) 217–228 225

Table 4
Bank income smoothing and investor-protection variables
Predicted (1) (2) (3) (4) (5)
sign
LLPt-1 + 0.0139*** (2.74) 0.0121** (2.34) 0.0159*** (3.23) 0.0158*** (3.19) 0.0187*** (3.93)
LLPt-2 + 0.0382*** (9.40) 0.0371*** (9.08) 0.0382*** (9.71) 0.0391*** (9.80) 0.0399*** (10.36)
EBT + 0.0548*** (4.91) 0.0589*** (5.57) 0.0602*** (4.65) 0.0496*** (4.75) 0.0523*** (4.15)
CAP +/ 0.0014*** (3.15) 0.0015*** (3.32) 0.0020*** (4.33) 0.0017*** (3.71) 0.0022*** (4.81)
CLOANS + 0.0187*** (11.00) 0.0190*** (11.24) 0.0196*** (11.14) 0.0192*** (11.26) 0.0196*** (11.07)
LLA + 0.0422*** (4.20) 0.0432*** (4.37) 0.0417*** (4.13) 0.0415*** (4.11) 0.0394*** (3.76)
GDPGR  0.0001** 0.0001* (1.78) 0.0001 (1.31) 0.0001* (1.74) 0.0001 (1.29)
(1.96)
LEGAL 0.0003*** (3.89) 0.0002 (0.28) 0.0005** (2.13)
EBT · LEGAL  0.0054***
(3.51)
ANTIDIRECTOR 0.0007*** (3.99) 0.0006** (2.14)
EBT · ANTIDIRECTOR  0.0150***
(4.00)
CREDITOR 0.0009 (0.89) 0.0006 (0.86)
EBT · CREDITOR  0.0182***
(2.79)
EBT · LEGAL · ANTIDIRECTOR  0.0017***
(3.23)
EBT · LEGAL · CREDITOR  0.0020**
(2.15)
Year dummies Yes Yes Yes Yes Yes
Country dummies Yes Yes Yes Yes Yes
m1 2.13** 2.13** 2.14 ** 2.14** 2.14**
m2 0.74 0.75 0.75 0.73 0.74
Sargan test 113.04 112.16 111.80 111.83 112.14
# Observations 3221 3221 3186 3221 3186
# Banks 1213 1213 1197 1213 1197
# Countries 40 40 38 40 38
Regressions are estimated using the Arellano and Bond (1991) GMM difference estimator for panel data with lagged dependent variables. A dependent
variable is the ratio of LLP over lagged total assets. As explanatory variables, we include two lags of the dependent variable, bank-specific fixed effects,
year and country dummies. EBT is profit before taxes and LLP over lagged total assets. CLOANS is the change in total loans outstanding over lagged
total assets. LLA is the beginning balance of the total allowance for loan-loss over lagged total assets. CAP is bank capital over risk-weighted assets.
GDPGR is real growth in per capita GDP. LEGAL is the measure of legal enforcement. ANTIDIRECTOR measures the protection of minority
shareholders and CREDITOR measures creditor rights. Regressions are estimated for 1995–2002. Year and country dummy variables are included in all
estimations, but are not reported. T-statistics are between parentheses. ***, **, and * represent significance at the 1%, 5%, and 10% level, respectively.

income-smoothing hypothesis. In Chile, Kenya, and Spain, that takes the value of 1 for developed countries and 0
income smoothing is detected only in publicly traded banks otherwise.9 Although not reported, the negative coefficient
but not in non-publicly traded banks.8 In six countries of the interaction term is consistent with a lower income
(Colombia, Greece, Malaysia, Pakistan, Thailand, and smoothing in developed countries.
the United Kingdom) results contradict the income- Results also indicate that publicly traded banks engage
smoothing hypothesis, with negatively related LLP and in income smoothing more than non-publicly traded banks
EBT. For Asian banks, Laeven and Majnoni (2003) find in eight countries (Chile, Colombia, Egypt, Kenya, Peru,
a significant negative association between earnings and Portugal, Spain, and Thailand). In Greece and Italy, how-
provisioning. We find no statistically significant LLP– ever, publicly traded banks smooth income less than non-
EBT relation in the remaining 21 countries. publicly traded banks. Such a difference across countries
To compare income smoothing between developed and suggests that more external users of publicly traded banks’
developing countries, we also estimate regression (1) in
the complete sample of banks incorporating an interaction
9
term of the earnings variable (EBT) with a dummy variable We classify countries as developed or developing following the World
Bank classification (‘‘Beyond Economic Growth: An Introduction to
Sustainable Development’’. World Bank Learning Resources Series,
8
Interestingly, in Spain non-publicly traded banks do not seem to second ed., 2004, Washington, DC). The countries for which our dummy
follow an income-smoothing pattern (GDP growth has a non-significant variable takes the value 1 are Australia, Canada, Denmark, France,
coefficient) even though the Bank of Spain in the mid-1990s introduced a Germany, Hong Kong, Ireland, Israel, Italy, Japan, New Zealand,
requirement for procyclical provisions for reasons of risk management Norway, Portugal, Singapore, Spain, Sweden, Switzerland and United
(the ‘‘statistical provision’’). Kingdom.
 226
Table 5
Bank income smoothing and regulation and supervision
Predicted sign (1) (2) (3) (4) (5) (6)
LLPt-1  0.0128** (2.51) 0.0074 (1.40) 0.0093* (1.78) 0.0116** (2.24) 0.0040 (0.76) 0.0030 (0.57)
LLPt-2 + 0.0375*** (9.22) 0.0345*** (8.34) 0.0355*** (8.66) 0.0369*** (8.99) 0.0329*** (8.01) 0.0289*** (7.08)
EBT + 0.0573*** (5.34) 0.0708*** (6.64) 0.0654*** (6.07) 0.0609*** (5.56) 0.0910*** (7.16) 0.1083*** (8.76)

A.R. Fonseca, F. González / Journal of Banking & Finance 32 (2008) 217–228

CAP +/ 0.0015*** (3.24) 0.0010** (2.36) 0.0012*** (2.75) 0.0013*** (2.96) 0.0007* (1.88) 0.0004 (1.44)
CLOANS + 0.0189*** (11.19) 0.0189*** (11.52) 0.0190*** (11.46) 0.0187*** (11.15) 0.0188*** (11.02) 0.0196*** (11.63)
LLA + 0.0427*** (4.29) 0.0482*** (5.03) 0.0468*** (4.81) 0.0436*** (4.41) 0.0536*** (5.74) 0.0641*** (7.29)
GDPGR  0.0001* (1.84) 0.0001*** (1.89) 0.0001* (1.73) 0.0001* (1.90) 0.0002** (2.38) 0.0002** (2.42)
DISCLOSURE 0.00004*** (3.85)
EBT · DISCLOSURE  0.0007*** (3.78)
RESTRICT 0.0004*** (3.56)
EBT · RESTRICT +/ 0.0058*** (4.92)
OFFICIAL 0.0003*** (3.55)
EBT · OFFICIAL  0.0045*** (4.52)
MONITOR 0.0005*** (3.82)
EBT · MONITOR  0.0085** (4.01)
STRUCT 0.0008* (0.77)
EBT · STRUCT +/ 0.0732*** (6.27)
FINAN 0.0008* (1.80)
EBT · FINAN +/ 0.0465*** (7.95)
Year dummies Yes Yes Yes Yes Yes Yes
Country dummies Yes Yes Yes Yes Yes Yes
m1 2.13** 2.13** 2.13*** 2.13*** 2.12** 2.12**
m2 0.74 0.77 0.76 0.75 0.78 0.83
Sargan test 112.76 112.68 112.69 113.17 114.92 110.41
# Observations 3003 3177 3221 3221 3221 3221
# Banks 1123 1197 1213 1213 1213 1213
# Countries 34 39 40 40 40 40
Regressions are estimated using the Arellano and Bond (1991) GMM difference estimator for panel data with lagged dependent variables. A dependent variable is the ratio of LLP over lagged total
assets. As explanatory variables, we include two lags of the dependent variable, bank-specific fixed effects, year and country dummies. EBT is profit before taxes and LLP over lagged total assets.
CLOANS is the change in total loans outstanding over lagged total assets. LLA is the beginning balance of the total allowance for loan-loss over lagged total assets. CAP is bank capital over risk-
weighted assets. GDPGR is real growth in per capita GDP. DISCLOSURE is the accounting disclosure index. RESTRICT is the measure of regulatory restrictions on bank activities, OFFICIAL
measures the power of official bank supervision. MONITOR is an index of private bank monitoring. STRUCT measures the market-orientation of the financial system. FINAN measures the country’s
financial development. Regressions are estimated for 1995–2002. Year and country dummy variables are included in all estimations, but are not reported. T-statistics are between parentheses. ***, **, and
* represent significance at the 1%, 5%, and 10% level, respectively.
 A.R. Fonseca, F. González / Journal of Banking & Finance 32 (2008) 217–228 227

financial statements fails to fully explain differences that target the credit and deposit markets reduce incentives
between publicly traded and non-publicly traded banks. to smooth earnings. Furthermore, this effect is greater than
We advise caution when interpreting these results, as eight the effect of substitute accruals reduction, such as capital
years of data cannot do full justice to the measurement of gains and losses, when banks are unable to operate in the
income-smoothing patterns across economic cycles. securities, insurance, and real estate markets.
Differences across countries have two primary mean- The positive coefficients of EBT · STRUCT and
ings. First, differences show how important it is to study EBT · FINAN, however, indicate that the exogenous com-
national conditions that affect bank incentives to smooth ponents of market orientation and development of the
earnings and obviate a common behavior pattern of provi- financial system are positively associated with bank income
sioning. Second, they point to the bias of estimations using smoothing. Greater income smoothing in market-oriented
international data that fail to control for national variables and more developed financial systems is consistent with
that may influence bank manager incentives to smooth the idea that bank managers have incentives to report more
earnings. stable profits, the more external users of financial state-
ments there are.10 Results also highlight the limited eco-
4.2. Results of cross-country determinants nomic significance of political-economy variables for
LLP, despite statistically significant, except at the level of
We report results using the GMM difference estimator market orientation and financial development. For
with country and time dummy variables in Tables 4 and instance, using the coefficients of the interaction terms of
5. Table 4 shows that legal variables measuring investor Table 5, a standard increase in STRUCT and FINAN
protection have the expected negative influence on bank would result in an enhanced relation between EBT and
income smoothing, as the EBT · LEGAL, EBT · ANTI- LLP that represents, respectively, 0.5 and 1.12 times the
DIRECTOR, and EBT · CREDITOR coefficients are neg- standard deviation of LLP.
ative. Moreover, as real protection of shareholders and Tests to control for a set of four macroeconomic vari-
creditors depends not only on legally established rights ables (inflation, loan growth, GDP growth, and stock-mar-
but also on their enforcement, LEGAL may be seen as a ket volatility) revealed no statistical significances, either as
complement of ANTIDIRECTOR and CREDITOR. To a group or individually including potential interactions
test this complementary effect, we interact LEGAL with with the earnings variable.
ANTIDIRECTOR and CREDITOR, respectively, in col-
umns 4 and 5. The negative coefficients of both interaction 5. Conclusions
terms confirm the complementary nature of legal enforce-
ment. This indicates that the greater the degree of law We have used a panel database of banks from 40 coun-
enforcement, the more investor protection reduces income tries to analyze bank income smoothing by management of
smoothing. These results are consistent with those of Leuz loan-loss provisions. We apply the GMM difference esti-
et al. (2003) for industrial companies. mator to control for unobservable heterogeneity and
The positive coefficients of LEGAL and ANTIDIREC- potential endogeneity of the explanatory variables. Results
TOR indicate that stronger minority shareholder protec- indicate that neither income smoothing nor different
tion and legal enforcement have a positive effect on the income smoothing between publicly and non-publicly
amount of LLP. The other variables mimic the coefficients traded banks is stable across countries.
observed when political-economy variables are not The root causes of differences in the pattern of income
included; i.e., the two lags of LLP, CAP, CLOANS, and smoothing across countries are many and varied. When
LLA have positive coefficients, and the growth of per we sequentially incorporate potential country determinants
capita GDP has negative coefficients. into the GMM difference estimations, our results indicate
Table 5 shows that better accounting disclosure, stricter that investor protection and legal enforcement reduce
regulations on bank activities, stricter official supervision, incentives to smooth earnings in banks. Incentives to
and more private monitoring reduce the use of LLP to smooth earnings decline with accounting disclosure,
smooth earnings. The negative influence of these four vari- restrictions on bank activities, and official and private
ables is consistent with expectations. The negative coeffi- supervision; incentives increase with market orientation
cient of EBT · DISCLOSURE suggests that stringent and development of the financial system.
accounting disclosure requirements are effective in improv-
ing the reliability of financial reports and reducing income
smoothing. The negative coefficients of EBT · OFFICIAL
and EBT · MONITOR indicate that official and private 10
When we control for the endogeneity of political-economy variables
supervision is effective in reducing bank risk, thereby using legal origin as an instrument, we focus only on their exogenous
component, isolating correlations caused by the legal origin of each
dampening incentives for managers to smooth income to
country. Analysis of the exogenous component alone explains why
reduce the volatility of bank income. Finally, although STRUCT and FINAN have a different effect on income smoothing from
we had forecasted contradictory effects for RESTRICT, LEGAL and DISCLOSURE, despite the correlations between the values
its negative influence indicates that the lower risks of banks observed for these four variables shown in Table 2.
228 A.R. Fonseca, F. González / Journal of Banking & Finance 32 (2008) 217–228

The basic implications are that any bank regulation and Barth, J.R., Caprio, G., Levine, R., 2004. Bank regulation and supervi-
supervision that reduces bank risk-taking will also diminish sion: What works best? Journal of Financial Intermediation 13, 205–
248.
a bank’s incentives to smooth earnings. Thus, stringent Beatty, A.L., Harris, D.H., 1999. The effects of taxes, agency cost and
restrictions on bank activities that reduce opportunities information asymmetry on earnings management: A comparison of
to assume risks in undertaking non-traditional bank activ- public and private firms. Review of Accounting Studies 4, 299–326.
ities also diminish the benefits of managing earnings to Beatty, A., Chamberlain, S., Magliolo, J., 1995. Managing financial
reduce a volatility, which is low in the first place. This is reports of commercial banks: The influence of taxes, regulatory capital
and earnings. Journal of Accounting Research 33, 231–262.
a similar scenario to the income smoothing observed in Beatty, A., Ke, B., Petroni, K.R., 2002. Earnings management to avoid
countries with more stringent official and private bank earnings declines across publicly and privately held banks. The
supervision. Accounting Review 77, 547–570.
The place of banking regulation and supervision in Beck, T., Levine, R., 2002. Industry growth and capital allocation: Does
explaining differences across countries in bank income having a market or bank-based system matter? Journal of Financial
Economics 64, 147–180.
smoothing is consistent with evidence that the financial Beck, T., Demirgüc-Kunt, A., Levine, R., 2003. Financial structure and
statement reliability varies across countries. We expand economic development Database. World Bank, Available from:
this evidence to show that bank regulation and supervision, <http://econ.worldbank.org>.
intended to enhance financial stability, also make financial Bikker, J.A., Metzemakers, P.A.J., 2005. Bank provisioning behaviour
statements more reliable. Our results support the usefulness and procyclicality. Journal of International Financial Markets, Insti-
tutions and Money 15, 141–157.
of the new banking regulation contained in Basel II in Cavallo, M., Majnoni, G., 2001. Do banks provision for bad loans in good
strengthening market discipline, because not only greater times. World Bank WP No. 2619.
disclosure requirements (Pillar 3) but also a more stringent Claessens, S., Laeven, L., 2004. What drives bank competition? Some
official supervision (Pillar 2) reduces income smoothing international evidence. Journal of Money, Credit and Banking 36,
and increases the reliability of a bank’s financial state- 585–592.
Demirgüc-Kunt, A., Detragiache, E., 2002. Does deposit insurance
ments. In this respect, Pillar 2 and Pillar 3 are complements increase banking system stability? An empirical investigation. Journal
in providing for stronger market discipline. of Monetary Economics 49, 1373–1406.
Fan, J., Wong, T., 2002. Corporate ownership structure and the
Acknowledgements informativeness of accounting earnings in East Asia. Journal of
Accounting and Economics 33, 401–426.
Goel, A.M., Thakor, A., 2003. Why do firms smooth earnings? Journal of
We thank Ana Isabel Fernández, Fernando Gascón, Business 76, 151–192.
Vı́ctor González, participants in the ACEDE Conference Greenawalt, M.B., Sinkey, J.F., 1988. Bank loan-loss provisions and the
at La Laguna (2005), the AECA Conference at Oviedo income-smoothing hypothesis: An empirical analysis. Journal of
(2005), and two anonymous referees for their helpful com- Financial Services Research 1, 301–318.
ments and suggestions. Financial support provided by the Keeley, M.C., 1990. Deposit insurance, risk and market power in banking.
American Economic Review 80, 1183–1200.
Spanish Education and Science Ministry – ERDF, Project La Porta, R., Lopez-de Silanes, F., Shleifer, A., Vishny, R.W., 1998. Law
SEC2006-15040 and by the Regional Government of Astu- and finance. Journal of Political Economy 106, 1113–1155.
rias, Project IB05-183, is gratefully acknowledged. A previ- La Porta, R., Lopez-de Silanes, F., Shleifer, A., 1999. Corporate
ous version of this paper has been published as the working ownership around the world. Journal of Finance 52, 471–517.
paper n 226/2005 of working paper series of the Funda- La Porta, R., Lopez-de Silanes, F., Shleifer, A., Vishny, R.W., 2002.
Investor protection and corporate valuation. Journal of Finance 52,
ción de las Cajas de Ahorro (FUNCAS) and was awarded 1147–1170.
the 2005 Sanchı́s Alcover Award at the ACEDE Confer- Laeven, L., Majnoni, G., 2003. Loan loss provisioning and economic
ence in La Laguna. slowdowns: Too much, too late? Journal of Financial Intermediation
12, 178–197.
Leuz, Ch., Nanda, D., Wysocki, P., 2003. Earnings management and
References investor protection: An international comparison. Journal of Financial
Economics 69, 505–527.
Ahmed, A.S., Takeda, D., Thomas, S., 1999. Bank LLP: A reexamination Levine, R., 1999. Law, finance and economic growth. Journal of Financial
of capital management, earnings management and signaling effects. Intermediation 8, 8–35.
Journal of Accounting and Economics 28, 1–25. Pérez, D., Salas, V., Saurina, J., 2004. Principles versus rules and the
Arellano, M., Bond, S., 1991. Some test of specification for panel data: definition of regulatory bank capital. Evidence from a unique
Monte Carlo evidence and application to employment equations. environment. Mimeo, Bank of Spain.
Review of Economic Studies 58, 227–297. Shen, C., Chich, H., 2005. Investor protection, prospect theory, and
Ball, R., Kothari, S., Robin, A., 2000. The effect of international earnings management: An international comparison of the banking
institutional factors on properties of accounting earnings. Journal of industry. Journal of Banking and Finance 29, 2675–2697.
Accounting Economics 29, 1–52. Shrieves, R.E., Dahl, D., 2003. Discretionary accounting and the behavior
Barth, J.R., Caprio, G., Levine, R., 2001. The regulation and supervision of Japanese banks under financial duress. Journal of Banking and
of banks around the world: A new database. World Bank Working Finance 27, 1219–1243.
Paper No. 2588. The updated version is available from: <http:// Wahlen, J.M., 1994. The nature of information in commercial bank loan
econ.worldbank.org>. loss disclosures. The Accounting Review 69, 455–478.