Munich Personal RePEc Archive
Bank Earnings Management using
Commission and Fee Income: the Role of
Investor Protection and Economic
Fluctuation
Ozili, Peterson K
2019
Online at https://mpra.ub.uni-muenchen.de/101824/
MPRA Paper No. 101824, posted 15 Jul 2020 09:12 UTC
P.K. Ozili & E.R. Outa
Bank Earnings Management using Commission and Fee Income:
the Role of Investor Protection and Economic Fluctuation
P. K Ozili
Central Bank of Nigeria
Erick R Outa
Charles Darwin University, Australia
Abstract
We investigate whether banks use commission and fee income to manage reported earnings as an
income-increasing or income smoothing strategy. We find that banks use commission and fee income
for income smoothing purposes and this behaviour persist during recessionary periods and in
environments with stronger investor protection. The implication of the findings is that bank non-interest
income which achieves diversification gains to banks is also used to manipulate reported earnings. Our
findings show that real earnings management is prevalent among banks in Africa. Further research into
earnings management should examine real earnings management among non-financial firms in
developing regions. From an accounting standard setting perspective, our evidence suggests the need
for national/international standard setters to adopt strict revenue recognition rules that ensure that banks
or firms report the actual fees they make, and to discourage banks from delaying (or deferring) the
collection of fee income to manage or smooth reported earnings opportunistically.
Keywords: Earnings management, Commission and Fee Income; Non-interest Income; Real Earnings
Management; Income smoothing; Economic Condition; Investor Protection; Banks.
JEL classification - G21 ; G28 ; G34 ; M41.
Page 1 of 21
P.K. Ozili & E.R. Outa
1. Introduction
We examine whether banks use commission and fee income to manage earnings, the incentive to do so,
and the influence of institutional and economic factors on this behaviour. We focus on bank commission
and fee income because commission and fee income is considered to be a significant component of
bank non-interest revenue (Smith et al, 2003; DeYoung and Rice, 2004; Ozili, 2017a). In recent years,
the low interest rate environment is claimed to have led to a decline in bank interest income and has
encouraged banks to rely more on non-interest source of funds to remain profitable (DeYoung & Rice,
2004). Although there are strong arguments for banks’ reliance on non-interest income, non-interest
income is also known to be unstable compared to interest income.1 The unstable nature of banks’ noninterest income can motivate managers to exert some discretion or control on the level of reported noninterest income, and in theory, the variability of income is predicted to create opportunities for managers
to smooth reported earnings to achieve some desired profit levels (Greenawalt and Sinkey, 1988).
However, the extent of this behaviour can be influenced by institutional quality (Leuz et al, 2003), and
by differing economic conditions (Ozili and Thankom, 2018). Therefore, it is important to understand
how variation in the non-interest income component of earnings can affect bank financial reporting.
Given that commission and fee income is a significant component of non-interest income (Smith et al,
2003; DeYoung and Rice, 2004), we argue that bank managers have incentives to influence the
reporting of commission and fee income in an attempt to increase earnings or to smooth out abnormal
fluctuations in earnings. Managers can delay the recognition of commission and fee income to a future
period or increase commission and fee income in the current period to increase earnings to meet some
desired reporting earnings outcomes. We test this prediction using bank data from a region where there
is no uniform regulation or uniform reporting for commission and fee income.
Except for banks in Europe where there is some attempt to regulate and standardise some component
of commission and fee income, there is yet no uniform regulation or reporting for commission and fee
income among banks in Africa. The lack of standardisation in the accounting for bank revenue
recognition among African countries can create opportunities for bank managers in the region to
influence the reporting of commission and fee income to manage reported earnings. The absence of
non-uniform accounting rules for revenue recognition in the region suggests that managerial discretion
will be a significant determinant of revenue recognition for banks in the region; this, therefore, provides
a natural setting to investigate managerial discretion in revenue recognition for earnings management.
In addition to the reasons above, this study is also motivated by the little focus on bank real activitiesbased earnings management compared to the extensive literature on bank accrual earnings management
via loan loss provisions2.
Since we are using dataset from Africa, our study also responds, and provides some insight, to other
issues or questions such as: Do banks in Africa engage in real activities management? What are the
incentives for real activities-based earnings management among banks in the African region? Under
what circumstances do real activities-based earnings management occur among banks in Africa? To
provide some answers to these questions, our study investigates whether bank revenues (in this case,
commission and fee income) are manipulated to influence the level of reported earnings particularly in
an under-researched African region. To date, we are not aware of any study that has examined this
question/topic in the context of banks in developed countries. In a developed country context, Stubben
(2010) show that firms have incentives to manipulate their revenue to manage earnings but his analysis
By ‘unstable’, we mean that clients/customers can quickly change banks to patronise the service of another
bank, which leads to unstable commission and fee during such periods (Smith et al, 2003).
2
E.g. Ahmed et al. (1999) and Fonseca & Gonzalez (2008)
1
Page 2 of 21
P.K. Ozili & E.R. Outa
did not examine banks. In contrast, we examine revenue-based earnings management among banks, and
the banking literature has not considered bank revenue to be a possible earnings management tool.
One common approach used to test for earnings management among banks is to focus on one
component of earnings and its relation to earnings before that component while controlling for factors
that influence that component of earnings (see, McNichols & Wilson, 1988; Ahmed et al., 1999;
McNichols, 2001; Ozili and Thankom, 2018). This is the approach we adopt in this paper. This approach
is considered to provide a more precise estimate of managerial discretion in bank financial reporting
(McNichols, 2001). Accordingly, we model commission and fee income as a function of earnings before
commission and fee income while controlling for economic fluctuation, bank size, investor protection
and other factors. Similar to Ahmed et al. (1999) and Stubben (2010), we model bank commission and
fee income as a function of its discretionary components (i.e., earnings before commission and fee
income) and its proposed non-discretionary components (i.e., the commission and fee income growth
rate, bank size and macroeconomic fluctuation).
Overall, the result indicates that African banks use commission and fee income to smooth earnings and
this behaviour is more pronounced when they are in recessionary periods and in environments with
stronger investor protection. One implication of our findings is that African banks also use real
activities-based techniques to influence the level of earnings not just accruals. Our findings show that
this behaviour is common across banks in most African countries. Our analysis in this paper is useful
to accounting standard setters and bank regulators in the region who want to understand (i) the extent
to which bank managers exercise discretion in earnings, (ii) how they do it, and (iii) the impact of this
behaviour on earnings quality.
Our study makes three contributions to the literature. Our study contributes to the positive accounting
theory (PAT) literature which examines the accounting and non-accounting decisions that influence
managers’ choice of accounting methods in financial reporting (Watts and Zimmermann, 1986). We
show that the need to survive a recession, and the presence of strong investor protection are two nonaccounting decisions that influence bank managers’ choice to engage in real earning management.
Secondly, we provide evidence that banks in Africa use commission and fees to manage (or to smooth)
earnings, a finding which has not been clearly explored in prior literature. Thirdly, by examining
commission and fee income, the study contributes to the literature on the relation between non-interest
income and bank diversification by providing additional insight that non-interest income that achieves
diversification gains is also used to manipulate (or smooth) reported earnings.
The rest of the paper is structured in the following way. Section 2 presents the theory and literature.
Section 3 presents the research design, data and method. Section 4 reports the empirical results of the
analysis. Finally, section 5 concludes.
2. Theoretical and Empirical Literature
2.1. Theory
Several hypotheses provide alternative explanations for why firms manage reported earnings. For
instance, the positive accounting theory (PAT)’s bonus plan hypothesis predict that managers of firms
will use accounting techniques or accounting numbers to increase earnings in order to increase the
likelihood of receiving bonuses that depend on the earnings number; while the PAT’s political cost
hypothesis predict that firms will use accounting techniques that lower the size of current earnings if
reported earnings are expected to be too high in order to avoid regulatory scrutiny and political scrutiny
Page 3 of 21
P.K. Ozili & E.R. Outa
of bank earnings by industry regulators (Watts & Zimmerman, 1986). Overall, positive accounting
theory argue that the incentive to manage earnings is driven by the presence of explicit contracts (i.e.,
bonus plans, debt covenant violation and the firm’s sensitivity to regulatory/political scrutiny).
On the other hand, the income smoothing hypothesis predicts that firms will use accounting procedures
or accounting numbers to lower high earnings or to increase low earnings to smooth out the fluctuations
in earnings (Ahmed et al., 1999; Ozili and Thankom, 2018). Also, the information asymmetry
hypothesis suggests that geographically-diversified firms with complex structures have greater
information asymmetry, and managers in such firms may exploit the additional information asymmetry
to manage earnings (Amidu and Kuibo, 2015). Taken together, these hypotheses provide alternative
theoretical explanations for earnings management practices among firms.
2.2 Literature Review
2.2.1. Real Earnings Management
Zang (2011) show that earnings management can occur through two channels: accruals earnings
management (AEM) and real activities-based earnings management (REM). Many studies focus on
earnings management using discretionary accruals (e.g. Dechow & Sloan, 1991; Bens et al., 2002;
Kothari, 2001; Ozili and Outa, 2018) while very few studies investigate banks because of the additional
regulations, disclosure requirements and the difficulties to determine actual accruals in banks.
Regarding earnings management using real techniques, Roychowdhury (2006) define real earnings
management as departures from normal operational practices motivated by managers’ desire to mislead
some stakeholders into believing certain financial reporting goals have been met in the normal course
of operations.
Regarding banks, the literature on real activities-based earnings management among banks is rather
scant, and Barth et al. (2017) confirm this. For instance, Beatty et al. (2002) find evidence that publicly
traded US banks use real techniques e.g. realised securities gains and losses, as well as loan loss
provisions, to eliminate small decreases in earnings. Also, Barth et al (2017) find evidence that banks
use realised gains and losses on available-for-sale securities to smooth earnings. Among developing
country studies, Hamdi and Zarai (2012) show that Islamic banks manage losses to avoid reporting
losses and to avoid earning decreases. Ozili (2017b) investigate the use of accruals (loan loss provisions)
to smooth income by African banks, and observe that African banks, particularly listed banks, use
accruals to smooth income. Additionally, Ozili (2017b) find that accruals are procyclical with economic
fluctuations. Amidu and Kuipo (2015) examine 330 African banks from 29 African countries from 2002
to 2009 and find that more than two-thirds of the 29 countries use discretionary accruals to manage
earnings. Similarly, Ozili (2015) show that banks in Nigeria use loan loss provisions to smooth earnings
over time. These studies do not focus on bank real earnings management via commission and fee
income.
Studies that test for the presence of earnings management among firms commonly use the total accrual
approach that estimate non-discretionary accruals as a linear function of change in revenues (or cash
revenue), change in gross property, plant, and equipment; and the residual is taken as the measure of
discretionary accruals or managerial discretion (Jones, 1991; Dechow et al., 1995). This approach has
been criticised for two reasons. One, it provides noisy and biased estimates of managed earnings
(Bernard and Skinner, 1996; Thomas and Zhang, 2000). Two, the approach do not reveal information
about the component of earnings that is used to manage earnings (Beneish, 2001; McNichols, 2001). In
contrast, banking studies commonly follow the approach of McNichols and Wilson (1988) and Ahmed
et al (1999) that examine one component of earnings and its relation to earnings before the component
Page 4 of 21
P.K. Ozili & E.R. Outa
while controlling for factors that influence that component of earnings. We follow this approach in this
paper to investigate whether a significant component of bank revenue is used to manage earnings.
Because revenues are a positive function of firm earnings, Plummer & Mest (2001), Caylor (2010) and
Stubben (2010) have associated revenue-based earnings management with income-increasing earnings
management, but these studies did not examine banks.
2.2.2. Bank Commission and Fee Income
Commission and fee income is the largest component of bank non-interest income and the second main
source of revenue to banks (Smith et al., 2003; DeYoung & Rice, 2004). To date, the banking literature
focus on how non-interest income/revenue relate to (i) bank diversification benefits (Smith et al., 2003),
and (ii) increase in overall profitability of banks; with little or no attention to whether bank managers
have incentives to influence or delay the recognition of income from fee-based activities to influence
the level of reported earnings. For instance, DeYoung & Rice (2004) suggest that banks engage in noninterest activities to generate non-interest income to boost shortfalls in overall profitability while Stiroh
(2004) and Stiroh & Rumble (2006) argue that banks engage in non-interest activities to generate noninterest income to diversify bank income stream. DeYoung & Roland (2001) show that while income
from fee-based activities increased bank earnings, it also increased the volatility of earnings thus
signalling little or no diversification gains. Overall, there is yet no consensus on whether bank noninterest income achieves its intended diversification benefits. Taken together, prior literature do not
explicitly view bank commission and fee income as a possible earnings management tool for banks,
and whether the presence of institutions that constrain managerial behaviour discourages earnings
management behaviour, if present. Our study explicitly examines this topic, by isolating commission
and fee income component of bank non-interest revenue to examine how bank managers’ reporting for
commission and fee income relate to bank earnings.
2.2.3. Economic Conditions
Some studies show that banks have incentive to use financial/accounting numbers to increase or lower
earnings during upturns and downturns in the economy (e.g. Ozili and Outa, 2017; El Sood, 2012;
Beatty & Liao, 2009; Liu & Ryan, 2006). These studies document that banks use discretionary accruals
to increase earnings during a recession to avoid reporting losses during the period. For instance, El Sood
(2012) find that US banks use accruals to increase earnings to avoid reporting a loss during a recession
(i.e., the 2007-2009 financial crisis period) while Beatty and Liao (2009) find similar evidence for US
banks. Liu & Ryan (2006), on the other hand, find that banks smooth income to reduce high profits
during economic boom. Ozili and Outa (2017) in their survey of literature, demonstrate that the earnings
distribution of banks is directly linked to economic fluctuations - high profits during good times and
low profits during bad times. We complement these studies and investigate whether banks use
commission and fee income to manage/smooth earnings during upturns and downturns in the economy.
3. Research Design
3.1. Contextual Framework
Banking systems in African countries vary largely in terms of the level of financial development,
banking concentration, financial deepening, regulation and supervision, corporate governance, investor
protection, banking population, bank transparency, etc. Beck & Cull (2013) points out that banking
systems in Africa are relatively more volatile compared to developed countries. They posit that the
frequent fluctuations in the income stream of firms and households in the region sometimes make it
Page 5 of 21
P.K. Ozili & E.R. Outa
difficult for individuals and firms to repay loans as at when due; hence, contributing to income
instability which can translate to banking system instability in the region. We argue that this claimed
banking instability in the African region can create incentives for banks in the region to use earnings
management techniques to stabilise reported earnings over time when they are in fluctuating banking
environments.
Regarding institutions, an African context to the study of bank real earnings management practices is
important because institutions that constrain bank behaviour across African countries significantly
differ from institutions that constrain bank earnings management behaviour in Europe or the US due to
differences in the level of development, extent of enforcement and so on. Also, the growing need for
African countries to establish institutions that promote increased bank transparency, protection of the
rights of minority shareholder and greater director liability, makes this study relevant. Hence, the need
to understand how real activities-based earnings management is influenced by institutional quality.
3.2 Data
We base our sample on African banking institutions in Bankscope database which contains accounting
information for large number of banks in the region. The sample consists of banks from 18 African
countries during the 2004 to 2013 period. The sample period selected, allows us to focus on the events
occurring within the specified pre-and post-crisis event window, where no significant regional change
in accounting rules had taken place at the time (2004 to 2013)3. The countries in the sample include:
Algeria, Angola, Botswana, Cameroun, Egypt, Ethiopia, Mauritius, South Africa, Nigeria, Kenya,
Togo, Tanzania, Ghana, Morocco, Uganda, Tunisia, Senegal and Zambia.
We use three country-level variables: real gross domestic product growth rate, banking competition and
investor protection. Bankscope database also provides cross-country data for banking competition
(Lerner index) archived in World Bank databank database. We obtain our real gross domestic growth
rate variable from the World Economic forum (see appendix for overview of data sources used for our
empirical analysis). We exclude countries that do not have institutional data relevant to the study. All
banks that report data for commission and fee income for at least 3 years and have the relevant countrylevel data are included in the analysis. Regarding bank type, we did not make a distinction between
types of African banks.
To clean up the data, we eliminated outliers above the 99th percentile and below the 1st percentile, to
minimise outliers and measurement errors. Secondly, we did not eliminate 2008 bank-year observations
to control for the impact of the 2008 financial crisis because we did not have a reason to believe that
the balance sheet of African banks was ‘adversely’ affected by the 2008 crisis. The resulting sample
comprise of 271 banks. Also, because some banks have missing values, the data is an unbalanced panel.
A first look at the sample descriptive statistics in Table 1a reveal that commission and fee income (CF)
for most African countries is around or above the mean CF while CF is much lower for banks in
Mauritius, Morocco and Tunisia. Also, the negative values reported for EBCF for some African
countries indicate that CF is a significant portion of bank earnings, if excluded, would lead to negative
earnings or losses. Finally, the number of observations is large in most columns in Table 1, but the
observations in each column are rather unbalanced across all columns due to missing values for some
variables which are not reported in Bankscope database.
3
Also, the number of available bank years used for this study is 10 years (i.e., 2004 to 2013). A ten-year period
is sufficient for the study because a 10-year is generally considered to reflect a full economic cycle which can
capture both upswings and downturns in an economy
Page 6 of 21
P.K. Ozili & E.R. Outa
Insert Table 1a Here
3.3 Research Design
To test whether African banks use commission and fee income to manage or smooth income, we use a
variation of the models use by prior studies (e.g., Ahmed et al., 1999; Barth et al., 2017; Ozili and
Thakom, 2018), which examine the relation between some bank accounting number and earnings before
the accounting number while controlling for other factors that might influence the magnitude of the
accounting number. Our main modified multivariate regression model is given as:
𝐶𝐹𝑖𝑡 = 𝛼0 + 𝛼1𝐸𝐵𝐶𝐹𝑖𝑡 + 𝛼2𝛥𝐶𝐹𝑅𝑖𝑡 + 𝛼3𝑆𝐼𝑍𝐸𝑖𝑡 + 𝛼4𝛥𝐺𝐷𝑃𝑡 + 𝛼5𝐵𝐴𝑁𝐾𝑑𝑢𝑚𝑚𝑖𝑒𝑠
+ 𝛼6𝐶𝑂𝑈𝑁𝑇𝑅𝑌𝑑𝑢𝑚𝑚𝑖𝑒𝑠 + 𝑒𝑖𝑡(𝐸𝑞𝑢𝑎𝑡𝑖𝑜𝑛 1)
All variables are defined in table 5. CF is the dependent variable measured as net commission and fee
income deflated by bank total asset. The CF variable captures reported commission and fee income
decisions of bank managers that are specific to the bank. EBCF is the earnings management variable of
interest, measured as earnings before tax and net commission and fee income. Barth et al. (2017)
intuitively show that, if firms use a revenue item to increase earnings, a positive relation between the
revenue item and reported earnings is expected while a negative sign is expected if banks use a revenue
item to smooth earnings which can be achieved by reporting fewer revenue items in order to decrease
too high earnings. Accordingly, we predict a positive sign for the EBCF coefficient if African banks
use commission and fee income to increase earnings as an income-increasing strategy and we predict a
negative sign for the EBCF coefficient if African banks use commission and fee income to smooth
reported earnings.
Additionally, we test whether African banks use CF to manage/smooth earnings when they expect
losses or when they are more profitable. To test for this, two dummy variables are introduced: NEG that
take the value 1 if EBCF is negative and zero otherwise; and POS that take the value 1 if EBCF is
above-the-median EBCF and zero otherwise. The POS and NEG variables are then interacted with
EBCF. POS*EBCF test whether banks have incentive to use CF to manage/smooth earnings when they
are more profitable (i.e., above-median EBCF). NEG*EBCF test whether banks have incentive to use
CF to manage/smooth earnings when they expect losses. The expanded model is shown below:
𝐶𝐹𝑖𝑡 = 𝛼0 + 𝛼1𝐸𝐵𝐶𝐹𝑖𝑡 + 𝛼2𝛥𝐶𝐹𝑅𝑖𝑡 + 𝛼3𝑆𝐼𝑍𝐸𝑖𝑡 + 𝛼4𝛥𝐺𝐷𝑃𝑡 + 𝛼5𝑃𝑂𝑆𝑖𝑡 + 𝛼6𝑃𝑂𝑆
∗ 𝐸𝐵𝐶𝐹𝑖𝑡 + 𝛼7𝑁𝐸𝐺𝑖𝑡 + 𝛼8𝑁𝐸𝐺 ∗ 𝐸𝐵𝐶𝐹𝑖𝑡 + 𝑒𝑖𝑡(𝐸𝑞𝑢𝑎𝑡𝑖𝑜𝑛 2)
For the control variables, ΔCFR captures contemporaneous change in the absolute amount of bank net
commission and fee income. This variable controls for the impact of contemporaneous fluctuation in
commission and fee income that may influence bank managers’ decision on the amount of commission
and fee income to be reported in the current period. ΔCFR is change in the absolute value of net
commission and fee income given as [(CFRt – CFRt-1)/CFRt-1]. When banks expect unstable
commission and fee income in the next period, they will have incentives to report more fee income in
the current period to compensate for subsequent periods that will yield lower commission and fee
income. Hence, we predict a positive relation between CF and ΔCFR.
The SIZE variable is included to control for the effect of bank size on commission and fee income.
Anandarajan et al (2003) suggest that large banks are considered to have high level of business activities
and a large client base for which they charge fees and commission in exchange for the services offered.
Page 7 of 21
P.K. Ozili & E.R. Outa
Following this reasoning, we expect banks with high level of business activities to generate more
commission and fee income; that is, large banks should have more fee income, therefore, we expect a
positive sign for the SIZE coefficient. SIZE is measured as the natural logarithm of bank total assets.
Real gross domestic product growth rate, (ΔGDP), controls for the impact of economic cycle fluctuation
on bank commission and fee income. Because bank clients will be able to pay for the services offered
to them during good economic conditions compared to periods of economic downturns, bank
commission and fee income is expected to be relatively substantial during periods of economic
prosperity and lower during economic downturns. Hence, we predict a positive sign for ΔGDP
coefficient.
As an additional test, we check whether banks use commission and fee income to manage earnings
when they are going through periods of economic recession or prosperity. To capture this, we introduce
two dummy variables into the analysis: REC that take the value 1 when ΔGDP is negative and zero
otherwise, and BOOM that take the value 1 when ΔGDP is above-the-median ΔGDP and zero
otherwise. REC and BOOM variables are then interacted with EBCF to test whether the relation
between earnings and commission and fee income depend on transient states of the economy.
𝐶𝐹𝑖𝑡 = 𝛼0 + 𝛼1𝐸𝐵𝐶𝐹𝑖𝑡 + 𝛼2𝛥𝐶𝐹𝑅𝑖𝑡 + 𝛼3𝑆𝐼𝑍𝐸𝑖𝑡 + 𝛼4𝛥𝐺𝐷𝑃𝑡 + 𝛼5𝑅𝐸𝐶𝑖𝑡 + 𝛼6𝑅𝐸𝐶
∗ 𝐸𝐵𝐶𝐹𝑖𝑡 + 𝛼7𝐵𝑂𝑂𝑀𝑖𝑡 + 𝛼8𝐵𝑂𝑂𝑀 ∗ 𝐸𝐵𝐶𝐹𝑖𝑡 + 𝑒𝑖𝑡. (𝐸𝑞𝑢𝑎𝑡𝑖𝑜𝑛 3)
Our country-level variables control for the influence of cross-country investor protection and
competition that might influence the reporting of bank commission and fee income. Fonseca &
Gonzalez (2008) and Ozili (2018a) argue and show evidence that strong investor protection and legal
enforcement discourages bank income smoothing behaviour via discretionary accruals. Similarly, we
use ‘INVPRO’ and ‘LEGAL’ to control for protection of minority shareholder rights and the quality of
legal system across African countries, respectively. Higher values of the two variables indicate stronger
protection of minority shareholders rights and higher legal enforcement quality. We also use the Lerner
index to control for banking competitiveness across countries. Beck et al (2013) also used the Lerner
index to control for cross-country banking competition. Banks in highly competitive banking
environments may charge lower fees for services offered to clients in order to attract new clients and/or
to retain existing clients. Therefore, we expect a negative relation between CF and the Lerner index
variable. Finally, we include the error term. The expanded equation is given as:
𝐶𝐹𝑖𝑡 = 𝛼0 + 𝛼1𝐸𝐵𝐶𝐹𝑖𝑡 + 𝛼2𝛥𝐶𝐹𝑅𝑖𝑡 + 𝛼3𝑆𝐼𝑍𝐸𝑖𝑡 + 𝛼4𝛥𝐺𝐷𝑃𝑡 + 𝛼5𝑅𝐸𝐶𝑖𝑡 + 𝛼6𝑅𝐸𝐶
∗ 𝐸𝐵𝐶𝐹𝑖𝑡 + 𝛼7𝐵𝑂𝑂𝑀𝑖𝑡 + 𝛼8𝐵𝑂𝑂𝑀 ∗ 𝐸𝐵𝐶𝐹𝑖𝑡 + 𝛼9𝐿𝐸𝐺𝐴𝐿 + 𝛼10𝐿𝐸𝐺𝐴𝐿
∗ 𝐸𝐵𝐶𝐹 + 𝛼11𝐼𝑁𝑉𝑃𝑅𝑂 + 𝛼12𝐼𝑁𝑉𝑃𝑅𝑂 ∗ 𝐸𝐵𝐶𝐹 + 𝛼13𝐿𝐸𝑅𝑁𝐸𝑅
+ 𝛼14𝐿𝐸𝑅𝑁𝐸𝑅 ∗ 𝐸𝐵𝐶𝐹 + 𝑒𝑖𝑡. (𝐸𝑞𝑢𝑎𝑡𝑖𝑜𝑛 4)
To test the robustness of the main econometric results, we first run the fixed effects4 OLS estimation to
account for bank and period unobserved heterogeneity between banks and across periods. Also, by
controlling for bank fixed effect, the fixed effect estimation addresses omitted variables bias that may
be associated with the main model in Equation 1. Also, since our explanatory variables and institutional
variables are time-varying, we also find it more appropriate to use the fixed effect estimation rather than
Page 8 of 21
P.K. Ozili & E.R. Outa
pooled OLS. The Hausman test also show that fixed effect estimation is a more appropriate estimation
technique. However, we later use pooled OLS estimation when we introduce two time-invariant
variables.
Finally, we test whether the use of CF to manage/smooth bank earnings exhibit forward-looking
properties. Bushman & William (2012) use this approach and find that managers exploit their discretion
in forward-looking reporting of discretionary accruals to manage earnings. To test for forward-looking
behaviour, we take the lag (or beginning values) of the explanatory variables in Equation 1 except for
EBCF and ΔGDP variables. This approach ensure that the CF coefficient only picks up the extent to
which banks’ reporting of commission and fee income is influenced solely by earnings consideration
and macroeconomic considerations without reference to current information about bank non-interest
income determinants. This lagged approach also allow us to test for the persistence of commission and
fee income over time. The model we adopt for this analysis is similar to Bushman & William (2012),
and is given as:
𝐶𝐹𝑖𝑡 = 𝐶𝐹𝑖𝑡 − 1 + 𝛼1𝐸𝐵𝐶𝐹𝑖𝑡 + 𝛼2𝛥𝐶𝐹𝑅𝑖𝑡 − 1 + 𝛼3𝑆𝐼𝑍𝐸𝑖𝑡 − 1 + 𝛼4𝛥𝐺𝐷𝑃𝑡
+ 𝑒𝑖𝑡
(𝐸𝑞𝑢𝑎𝑡𝑖𝑜𝑛 5)
We estimate the model in Equation 5 by using Arellano & Bond (1991) Generalised-Method-ofMoments (GMM) first difference estimator. This technique address (i) the presence of unobserved
bank-specific effects, which is eliminated by taking first-differences of all variables; (2) the
autoregressive process in the data regarding the persistence of bank commission and fee income and
the (iii) potential endogeneity of the explanatory variables with the error term.
4. Result
4.1. Main Result
The main result is reported in Column 1 of Table 2. The EBCF coefficient is negative and significant
at 1% level and indicates that banks in the African region use commission and fee income to smooth
earnings. This is consistent with the income smoothing hypothesis and is consistent with Barth et al
(2017) who find that banks use real techniques to smooth income. The result implies that African banks
report fewer commission and fee income to lower high earnings and report higher commission and fee
income to increase low earnings so that reported earnings are never too high or too low, to achieve
income smoothing. Ozili (2015) also find evidence for income smoothing among Nigerian banks, and
Ozili and Thankom (2018) find evidence for income smoothing among European systemic banks.
The control variables report the predicted signs except for SIZE. ΔCFR report the expected positive
sign but is insignificant while SIZE coefficient is negatively significant, indicating that large banks
report fewer commission and fee income. ΔGDP coefficient reports the predicted positive sign but is
insignificant, implying that reported commission and fee income by African banks do not exhibit
significant cyclical behaviour in response to changing economic conditions in the African region.
4.2. Additional Analysis: Transient Effect
Column 2 and 3 of Table 2 show that the POS*EBCF and NEG*EBCF coefficients are insignificant.
Column 4 report a negative sign for REC*EBCF coefficient and is significant at 5% level, indicating
that the use of commission and fee income to smooth income by African banks is more pronounced
during economic downturns/recessions. Beatty & Liao (2009) and El Sood (2012) document similar
Page 9 of 21
P.K. Ozili & E.R. Outa
evidence for accruals. BOOM*EBCF, on the other hand, report a negative but insignificant sign and is
inconsistent with Liu & Ryan (2006).
Regarding investor protection and banking competition, INVPRO*EBCF coefficient report a negative
sign and is significant at 1% level. This indicates that bank income smoothing via commission and fee
income is more pronounced in environments that have stronger protection of minority shareholders
rights. LEGAL*EBCF is negatively significant at the 5% level, indicating that bank income smoothing
via commission and fee income is also pronounced in environments with higher legal enforcement
quality. Taken together, these findings indicate that African bank managers are more likely to use real
techniques to smooth bank earnings when they are in strong legal and investor protection environments.
Also, LERNER*EBCF coefficient report a positive but insignificant sign. Overall, the results indicate
that African banks use commission and fee income to smooth earnings and this behaviour is more
pronounced when they are in recessionary periods and in environments with stronger investor
protection.
Insert Table 2 Here
4.3. Cross-Country Analysis
Next, we undertake country-specific analysis to control for the bias that international analysis ignores
national aspects that differ by country. We re-run the model for each country and include real GDP
growth rate but exclude the institutional variables from the model. EBCF and ΔGDP are the variables
of interest here. Table 3 report the results. As can be observed, EBCF coefficient reports a negative sign
for 14 countries (Algeria, Angola, Botswana, Cameroun, Egypt, Ethiopia, Ghana, Kenya, Nigeria,
Senegal, South Africa, Tanzania, Tunisia, and Zambia). Of these, EBCF coefficient is negatively
significant for banks in 8 African countries (Algeria, Cameroun, Ethiopia, Ghana, Nigeria, Senegal,
South Africa and Tanzania), indicating evidence for earnings smoothing via commission and fee
income. Also, EBCF coefficient is positively significant for banks in Mauritius, indicating evidence for
income-increasing earnings management. ΔGDP coefficient is negatively significant in Zambia, Togo
and Morocco, indicating a counter-cyclical relation between CF and economic cycle fluctuations. Also,
procyclical commission and fee income behaviour is observed in Cameroun and Ethiopia as indicated
by the positively significant ΔGDP coefficient. Overall, the result suggests that earnings smoothing is
common among countries in our sample. Also, the link between commission and fee income and the
economic cycle across countries in the sample is mostly weak (insignificant). This weak link provides
some justification for banks’ involvement in non-interest activities as income generated from such
activities are not significantly correlated with business cycle fluctuations.
Insert Table 3 Here
4.4. Pre- and Post-Financial Crisis
Next, we test whether earnings management is pronounced in the post-financial crisis period relative to
the pre-financial crisis period. To do this, we create a financial crisis dummy variable (CRISIS) and
assign a value ‘1’ for the post-crisis period (2009-2013) and assign a value of ‘0’ for the pre-financial
crisis period (2004-2007)5; thereafter, we interact the financial crisis variable with the earnings
5
The year-2008 data is excluded from the analysis. This is because most banks had significant write-offs in the
crisis-years and including such crisis-data into the models often constitute outliers which can bias the empirical
results due to the extreme or large values for some variables. Furthermore, African banks experienced
significant write-offs during the crisis-years due to their heavy exposure to fluctuating oil prices in 2008. The
financial crisis made oil prices volatile and transmitted huge losses on the balance sheet of African banks that
Page 10 of 21
P.K. Ozili & E.R. Outa
management variable (EBCF). The result is reported in Column 3 of Table 4. The EBCF coefficient is
significant but the interaction of EBCF with CRISIS is insignificant, indicating that there is no evidence
for bank earnings management via commission and fee income in the post-crisis period. More so, the
CRISIS variable is not statistically significant, indicating that the post-crisis period did not have a
significant effect on bank earning management using commission and fee income in Africa, after the
financial crisis.
Insert Column 3 of Table 4 Here
4.5. Robustness
First, the correlation matrix in the appendix show that the correlation among the variables is sufficiently
low and suggests that multicollinearity is not an issue in the analysis. Second, we re-estimate the models
using the natural logarithm of real GDP growth rate as an alternative measure to capture non-negative
fluctuations in the economic cycle instead of real GDP growth rate. Taking the natural log drops out
the negative values. We then interact the new measure with EBCF and re-run the model and the results
remain insignificant. We also modify the BOOM variable to take the value 1 for all positive values of
real GDP growth rate while negative values take zero. The result is not significantly affected by this
change. Hence, we did not report these analyses due to lack of space in the manuscript. Further,
regarding the high earnings dummy variable ‘POS’, we use an alternative measure where the POS
dummy variable take the value 1 when EBCF is positive and zero otherwise. The result was not
significantly affected by this change.
Third, with respect to the sample size, we used (i) active banks in the region, and (ii) use all banks that
have data for three consecutive years in any order in the time series. The latter allows us to include
active banks that do not have full reporting data on commission and fee income, therefore, we are
confident that survivorship bias is not an issue in the analysis. Fourth, we test whether the use of CF to
smooth earnings is achieved when banks do not consider current information about the structure of
commission and fee income. The result is derived from the model in Equation 5. Column 1 and 2 of
Table 4 show that the CF coefficient is negative but not significant indicating that bank managers do
not use CF to smooth earnings when they do not take into account current information about non-interest
income or commission and fee income (or non-interest income structure). The observed negative sign
further confirms the main result that bank managers use CF to smooth earnings. Also, CFt-1 is positively
significant in Column 1 and 2, indicating that previous information about commission and fee income
is a major determinant of reported commission and fee income in the current period. Also, we check
whether listed and unlisted African banks use CF to smooth or manage earnings and the results are not
significant.
Finally, we address concerns that the large number of sample banks for South Africa may affect our
inference. We excluded South African banks from the sample and the results do not change significantly
as can be observed in Column 7 of Table 4.
Insert Table 4 Here
5. Conclusion
had significant exposure in the oil sector. Much of the losses were written-off in their year-2008 financial
statement, hence the need to exclude 2008 year-observations.
Page 11 of 21
P.K. Ozili & E.R. Outa
Earnings management among banks in emerging and developing countries is an emerging area in the
literature and has received considerable attention among researchers, regulators and analysts in the
banking sector. This study re-examines the question on earnings management focusing on the African
banking sector. We focus on how banks use commission and fee income to influence reported earnings.
Using African bank data, over a 10-year period 2004 to 2013, the result and conclusions indicate that
African banks use commission and fee income to smooth reported earnings and this behaviour is more
pronounced when they are in recessionary periods and in environments with stronger investor
protection.
From a prudential perspective, research on bank commission and fee income is important to banking
supervisors who have concerns that banks in the region charge high fees to clients but disguise this
behaviour by understating earnings to avoid reporting too high earnings possibly to evade scrutiny of
bank profits. Hence, our evidence shed some light into this issue and underline the need for sound
prudential guidelines to supervise and monitor the reporting of commission and fee income and other
revenue items by African banks. From an accounting standard setting perspective, our findings stress
the need for national/international standard setters to adopt strict revenue recognition rules that ensure
that banks/firms report the actual fees they make, and to discourage banks from delaying (or deferring)
the collection of fee income to manage or smooth reported earnings opportunistically.
One limitation of the study is that recent developments in African countries could alter the results,
particularly in the post 2014 era. Another limitation is that the years after 2008 could also be affected
by the crisis. Therefore, future research should explore the potential for revenue management as an
earnings management strategy in the post-crisis period.
A natural direction for future research is the need for future studies to undertake an in-depth analysis of
specific factors, including accounting and regulatory practices in individual countries, that influence
this behaviour in the region. Future research can replicate this study to developed country contexts
where the reporting of revenue is not regulated or standardised. Finally, future research could also
investigate whether Basel capital regulation have any influence on banks reported commission and fee
income. For instance, banks with more regulatory capital can have incentives to engage in risky
activities for which they can charge higher fees and commission.
Page 12 of 21
P.K. Ozili & E.R. Outa
Reference
Ahmed, A. S., Takeda, C., & Thomas, S. (1999), “Bank loan loss provisions: a reexamination of
capital management, earnings management and signaling effects”, Journal of Accounting and
Economics, Vol 28 No. 1, pp, 1-25.
Amidu, M. & Kuipo, R. (2015), “Earnings management, funding and diversification strategies of
banks in Africa”, Accounting Research Journal, Vol 28 No 2, pp. 172-194.
Anandarajan, A., Hasan, I., & Lozano-Vivas, A. (2003), “The role of loan loss provisions in earnings
management, capital management, and signaling: The Spanish experience”, Advances in International
Accounting, 16, 45-65.
Arellano, M. & Bond, S. (1991), “Some tests of specification for panel data: Monte Carlo evidence
and an application to employment equations”, The Review of Economic Studies, Vol 58 No. 2, pp.
277-297.
Barth, M. E., Gómez Biscarri, J., Kasznik, R., & López-Espinosa, G. (2017), “Bank earnings and
regulatory capital management using available for sale securities”. Review of Accounting Studies,
Vol 22, No 4, 1761-1792.
Beatty, Anne L., Bin Ke, & Kathy R. Petroni. (2002), “Earnings Management to Avoid Earnings
Declines across Publicly and Privately Held Banks”, The Accounting Review, Vol 77, No 3 pp. 547–
570.
Beatty, A., & Liao, S. (2009), “Regulatory capital ratios, loan loss provisioning and procyclicality”,
Columbus, United States: Ohio State University. Mimeographed document.
Beck, T., De Jonghe, O. & Schepens, G. (2013), “Bank competition and stability: cross-country
heterogeneity” Journal of Financial Intermediation, Vol 22 No. 2, pp. 218-244.
Beck, T., & Cull, R. (2013) “Banking in Africa.” World Bank Policy Research Working Paper, No.
6684.
Beneish, M. D. (2001) “Earnings management: A perspective. Managerial Finance, Vol 27 No. 12,
pp. 3-17.
Bens, D. A., Nagar, V., & Wong, M. H. (2002) “Real investment implications of employee stock
option exercises”, Journal of Accounting Research, Vol 40 No. 2, pp. 359-393.
Bernard, V. L., & Skinner, D. J. (1996), “What motivates managers' choice of discretionary
accruals?” Journal of Accounting and Economics, Vol. 22 No. 1, pp. 313-325.
Bushman, R. M., & Williams, C. D. (2012), “Accounting discretion, loan loss provisioning, and
discipline of banks’ risk-taking”. Journal of Accounting and Economics, Vol 54 No.1, pp. 1-18.
Caylor, M. L (2010), “Strategic revenue recognition to achieve earnings benchmarks”, Journal of
Accounting and Public Policy, Vol 29 No.1, pp. 82-95.
Dechow, P. M., Sloan, R. G., & Sweeney, A. P. (1995), “Detecting earnings management”, The
Accounting Review, Vol 70 No 2, pp. 193-225.
Dechow, P., & R. Sloan. (1991), “Executive incentives and the horizon problem: An empirical
investigation”, Journal of Accounting and Economics, Vol 14 No.1, pp. 51–89.
Page 13 of 21
P.K. Ozili & E.R. Outa
DeYoung, R., & Rice, T. (2004), “How do banks make money? The fallacies of fee income”,
Economic Perspectives - Federal Reserve Bank of Chicago, Vol. 28 No.4, pp. 34.
DeYoung, R., & Roland, K. P. (2001), “Product mix and earnings volatility at commercial banks:
Evidence from a degree of total leverage model”, Journal of Financial Intermediation, Vol 10 No 1,
pp. 54-84.
El Sood, H. A. (2012) “Loan loss provisions and income smoothing in US banks pre and post the
financial crisis”, International Review of Financial Analysis, Vol 25, pp. 64-72.
Fonseca, A. R. & González, F. (2008), “Cross-country determinants of bank income smoothing by
managing loan-loss provisions”, Journal of Banking and Finance, Vol 32 No 2, pp. 217-228.
Greenawalt, M. B., & Sinkey, J. F. (1988). Bank loan-loss provisions and the income-smoothing
hypothesis: an empirical analysis, 1976–1984. Journal of financial services research, Vol 1 No 4, pp.
301-318.
Hamdi, M. & Zarai, M. (2012), “Earnings Management to Avoid Earnings Decreases and Losses:
Empirical Evidence from Islamic Banking Industry”, Research Journal of Finance and Accounting.
Vol 3 No.3, pp. 88.
Jones, J. (1991), “Earnings management during import relief investigations”, Journal of Accounting
Research, Vol. 29, No 2, pp.193-228.
Kothari, S. P. (2001), “Capital markets research in accounting”, Journal of Accounting and
Economics, Vol. 31 No. 1, pp. 105-231.
Leuz, C., Nanda, D., & Wysocki, P. D. (2003). Earnings management and investor protection: an
international comparison. Journal of financial economics, Vol. 69 No.3, pp. 505-527.
Liu, C. C., & Ryan, S. G. (2006), “Income smoothing over the business cycle: Changes in banks'
coordinated management of provisions for loan losses and loan charge-offs from the pre-1990 bust to
the 1990s boom”, The Accounting Review, Vol 81 No.2, pp. 421-441.
McNichols, M. F. (2001), “Research design issues in earnings management studies”, Journal of
accounting and public policy, Vol. 19 No.4, pp. 313-345.
McNichols, M., & Wilson, G. P. (1988), “Evidence of earnings management from the provision for
bad debts”, Journal of Accounting Research, Vol 26, pp.1-31.
Ozili, P. K. (2015), “Loan Loss Provisioning, Income Smoothing, Signaling, Capital Management and
Procyclicality: Does IFRS Matter? Empirical Evidence from Nigeria”, Mediterranean Journal of
Social Sciences, Vol 6 No.2, pp. 224-232.
Ozili P. K. (2017a), “Bank Earnings Management and Income Smoothing using Commission and Fee
Income: A European Context”, International Journal of Managerial Finance, Vol 13, No 4, pp. 419439.
Ozili P. K. (2017b), “Bank earnings smoothing, audit quality and procyclicality in Africa: The case of
loan loss provisions”, Review of Accounting and Finance, Vol 16 No 2, pp 142-161.
Ozili P. K. and Thankom, A. G. (2018), “Income Smoothing among European systemic and nonsystemic banks.” British Accounting Review, Vol 50, No 5, 539-558.
Page 14 of 21
P.K. Ozili & E.R. Outa
Ozili, P. K., & Outa, E. (2017). Bank loan loss provisions research: A review. Borsa Istanbul Review.
Vol 17, No 3, 144-163.
Ozili, P. K. (2018a). Bank loan loss provisions, investor protection and the macroeconomy.
International Journal of Emerging Markets, 13(1), 45-65.
Ozili, P. K., & Outa, E. R. (2018). Bank earnings smoothing during mandatory IFRS adoption in
Nigeria. African Journal of Economic and Management Studies.
Petersen, M. (2009), “Estimating Standard Errors in Finance Panel Data Sets: Comparing
Approaches”, Review of Financial Studies, Vol 22, No 1 pp. 435-448.
Plummer, E., & D. Mest. (2001), “Evidence on the management of earnings components”, Journal of
Accounting, Auditing & Finance, Vol 16, No 4, pp 301–323.
Roychowdhury,S. (2006). “Earnings management through real activities manipulation”, Journal of
Accounting and Economics, Vol 42, No 3 pp. 335-370
Smith, R., Staikouras, C. & Wood, G. (2003), “Non-interest income and total income stability”, Bank
of England Working Paper No 198. Available at:
http://www.bankofengland.co.uk/archive/Documents/historicpubs/workingpapers/2003/wp198.pdf
(Accessed: 4th January, 2016).
Stiroh, K. (2004), “Diversification in banking: Is non-interest income the answer?” Journal of Money,
Credit and Banking, Vol. 36, No 5 pp. 853-882.
Stiroh, K. J. & Rumble, A. (2006), “The dark side of diversification: The case of US financial holding
companies”, Journal of Banking & Finance, Vol 30, No.8, pp. 2131-2161.
Stubben, S. R. (2010), “Discretionary revenues as a measure of earnings management”. The
Accounting Review, Vol. 85 No. 2, pp. 695-717.
Thomas, J., & Zhang X. J. (2000), “Identifying unexpected accruals: a comparison of current
approaches”. Journal of Accounting and Public Policy, Vol. 19, No 4-5, pp.347-376.
Watts, R. L., & Zimmerman, J. L. (1986), Positive accounting theory. Prentice-Hall, New Jersey.
Zang, A. Y. (2011), “Evidence on the trade-off between real activities manipulation and accrual-based
earnings management”. The Accounting Review, Vol. 87 No. 2, pp. 675-703.
Page 15 of 21
P.K. Ozili & E.R. Outa
Tables
Table 1: Descriptive Statistics (Number of Observations per country)
The final number of observations for each country are reported in parenthesis (below the mean values)
Country
Mean
Mean
Mean
Mean
Mean
Mean
Mean
Mean
#
CF
ΔCFR
EBCF
SIZE ΔGDP LERNER INVPRO LEGAL Banks
Algeria
0.017
1.39
0.004
14.32
3.1
0.57
5
-0.71
15
(132)
(115)
(132)
(142)
(150)
(120)
(135)
(120)
Angola
0.015
2.49
0.008
13.88
10.8
0.43
5.3
-1.34
13
(99)
(86)
(99)
(102)
(130)
(104)
(117)
(104)
Botswana
0.015
0.31
0.034
12.90
7.6
0.206
5.4
0.63
12
(99)
(84)
(99)
(102)
(120)
(84)
(108)
(96)
Cameroun
0.025
-0.04
-0.006
13.09
3.5
0.388
4.3
-1.12
6
(36)
(33)
(39)
(54)
(60)
(48)
(54)
(54)
Egypt
0.013
0.087
0.001
14.83
4.6
0.135
3.6
-0.12
16
(345)
(304)
(345)
(345)
(360)
(240)
(324)
(288)
Ethiopia
0.014
1.33
0.018
13.28
11.0
0.537
3.3
-0.72
10
(82)
(72)
(82)
(82)
(100)
(80)
(90)
(80)
Ghana
0.027
0.20
0.009
13.11
7.5
0.348
6.3
-0.07
15
(109)
(94)
(109)
(109)
(150)
(120)
(135)
(120)
Kenya
0.018
0.189
0.007
12.47
5.3
0.318
5
-0.96
24
(234)
(209)
(234)
(237)
(240)
(192)
(216)
(192)
Mauritius
0.006
0.153
0.009
13.68
3.9
0.475
7.7
0.91
14
(121)
(106)
(121)
(124)
(140)
(112)
(126)
(112)
Morocco
0.007
0.116
0.015
16.01
4.4
0.293
3.4
-0.18
13
(99)
(86)
(99)
(104)
(130)
(104)
(117)
(104)
Nigeria
0.023
0.153
0.185
15.73
8.8
0.185
5.7
-1.22
16
(59)
(43)
(59)
(63)
(160)
(128)
(144)
(128)
Senegal
0.015
1.37
-0.003
12.70
3.8
0.313
3
-0.25
10
(80)
(68)
(80)
(92)
(100)
(80)
(90)
(80)
South Africa
0.032
0.309
-0.002
14.90
3.3
0.264
8
0.10
29
(269)
(239)
(269)
(272)
(290)
(232)
(261)
(232)
Tanzania
0.020
0.223
-0.004
12.21
6.7
0.312
4.9
-0.41
16
(147)
(131)
(147)
(147)
(160)
(128)
(144)
(128)
Togo
0.017
0.17
0.001
12.41
3.5
0.244
3.7
-0.94
7
(55)
(48)
(55)
(59)
(70)
(56)
(63)
(56)
Tunisia
0.008
0.12
0.002
13.74
3.9
0.250
4.8
0.13
20
(191)
(166)
(191)
(191)
(200)
(160)
(180)
(180)
Uganda
0.025
0.338
-0.006
11.98
7.1
0.332
4.7
-0.45
21
(138)
(117)
(136)
(180)
(136)
(105)
(189)
(168)
Zambia
0.034
0.246
-0.027
11.77
7.8
0.279
5.3
-0.51
14
(108)
(94)
(108)
(114)
(140)
(112)
(126)
(112)
TOTAL
271
Mean
0.019
0.265
0.003
13.51
5.74
0.322
5.23
-0.36
Median
0.014
0.11
0.005
13.24
5.17
0.302
5.00
-0.39
S.D.
0.02
1.97
0.26
1.94
3.91
0.650
1.48
0.58
Observation
2215
1914
2213
2328
2710
2045
2439
2168
CF = net commission and fee income to total asset ratio. EBCF = earnings before tax and commission and fee
income to total assets. SIZE = natural logarithm of total asset. ΔCFR is change in commission and fee income
outstanding. ΔGDP is real gross domestic product growth rate. INVPRO = minority shareholder rights
protection. LERNER = banking competition. LEGAL = Quality of legal enforcement. Number of
observations are reported in parenthesis
Page 16 of 21
P.K. Ozili & E.R. Outa
C
ΔCFR
EBCF
(1)
0.105***
(4.24)
0.0002
(1.27)
-0.136***
(-3.72)
SIZE
-0.006***
(-3.43)
ΔGDP
0.00003
(0.41)
POS
NEG
REC
BOOM
Table 2: Main Regression (Fixed Effect)
(2)
(3)
(4)
(5)
0.103*** 0.103*** 0.105*** 0.108***
(4.35)
(4.38)
(4.17)
(4.14)
0.0002
0.0002
0.0002
0.0002
(1.25)
(1.25)
(1.24)
(1.28)
-0.103*
0.152***
(-1.94)
0.133*** 0.130***
(-3.07)
(-3.58)
(-2.92)
0.006*** 0.006*** 0.006*** 0.007***
(-3.52)
(-3.58)
(-3.43)
(-3.42)
0.00002
0.00002
0.0001
-0.0001
(0.30)
(0.22)
(0.88)
(-1.51)
-0.0004
(-0.51)
0.001*
(1.67)
0.001
(1.04)
0.002**
(2.15)
INVPRO
(6)
0.101***
(4.22)
0.0002
(1.21)
0.266**
(2.12)
0.006***
(-3.74)
0.00007
(0.88)
(7)
0.129***
(4.01)
0.0002
(1.26)
0.230***
(-3.58)
0.008***
(-3.50)
0.00006
(0.81)
REC*EBCF
BOOM*EBCF
INVPRO*EBCF
LEGAL*EBCF
LERNER*EBCF
0.00003
(0.42)
-0.006
(-1.61)
LERNER
NEG*EBCF
-0.008***
(-3.37)
0.001***
(3.22)
LEGAL
POS*EBCF
(8)
0.133***
(3.95)
0.0002
(1.20)
-0.322**
(-2.49)
-0.005
(-1.01)
0.043
(0.65)
-0.038
(-0.57)
-0.175**
(-1.99)
-0.011
(-0.25)
0.070***
(-2.60)
-0.169**
(-2.43)
0.494
(1.39)
Adjusted R²
76.35
76.36
76.40
76.50
76.43
77.86
81.63
81.19
F-statistic
22.81
22.66
22.71
22.83
22.74
24.65
24.39
23.05
P-value
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
Observation
1912
1912
1912
1912
1912
1910
1406
1365
All estimations include robust standard errors clustered by bank and year. Bank and year fixed effects are included. Tstatistics are reported in parentheses with ***, **, and * indicating 1%, 5%, and 10% significance level, respectively. CF =
net commission and fee income to total asset ratio. POS = dummy variable that take the value 1 when EBCF ratio is abovethe-median EBCF ratio and zero otherwise; NEG = dummy variable that take the value 1 when EBCF is negative and zero
otherwise. REC = dummy variable that take the value of 1 during periods of economic downturns, that is, periods with
negative ΔGDP growth rate, and zero otherwise; BOOM = dummy variable that take the value of 1 for periods of economic
prosperity, that is, periods with above-the-median ΔGDP growth rate, and zero otherwise. LEGAL = quality of legal systems
across countries. INVPRO = protection of minority shareholders rights. LERNER = banking competitiveness.
Page 17 of 21
P.K. Ozili & E.R. Outa
Table 3: County-Specific (Pooled OLS) Regression
Model: CFit = α0 + α1EBCFit + α2ΔCFRit + α3SIZEit + α4ΔGDPt + eit
Country
β0
EBCF
ΔCFR
SIZE
ΔGDP
Algeria
-0.060**
-0.293***
-0.0004***
0.006***
0.0003
(-2.44)
(-2.77)
(-3.54)
(3.29)
(0.54)
Angola
0.042**
-0.009
0.0001***
-0.002*
0.00006
(2.51)
(-0.38)
(4.51)
(-1.71)
(0.55)
Botswana
0.135***
-0.016
0.0005
-0.009**
-0.0002
(2.87)
(-0.63)
(1.24)
(-2.54)
(-1.14)
Cameroun
0.159***
-0.547**
-0.0009
-0.012***
0.005***
(4.44)
(-2.36)
(-1.41)
(-4.14)
(3.97)
Egypt
0.331***
-0.042
0.005
-0.022***
-0.0004
(4.02)
(-0.23)
(0.57)
(-3.91)
(-0.91)
Ethiopia
-0.085***
-0.343***
-0.0001**
0.007***
0.0009**
(-3.17)
(-5.17)
(-2.54)
(3.98)
(1.96)
Ghana
0.028
-0.045*
0.004
-0.0001
0.00002
(0.99)
(-1.84)
(1.17)
(-0.06)
(0.08)
Kenya
0.047***
-0.038
0.003**
-0.002***
-0.00008
(4.91)
(-1.13)
(2.38)
(-2.98)
(-0.64)
Mauritius
0.016
0.068*
0.0009***
-0.0008
-0.0001
(1.38)
(1.74)
(2.85)
(-0.93)
(-0.64)
Morocco
0.028
0.031
0.005**
-0.001
-0.0003*
(1.59)
(0.31)
(2.40)
(-1.17)
(-1.84)
Nigeria
0.128***
-0.043**
0.002
-0.007***
-0.0005
(3.52)
(-1.97)
(1.27)
(-3.01)
(-1.02)
Senegal
-0.021
-0.146**
-0.0001
0.003*
-0.0002
(-0.92)
(-2.47)
(-0.87)
(1.71)
(-0.33)
South Africa
0.382***
-0.423***
0.0009
-0.025***
0.0004
(5.42)
(-3.75)
(0.66)
(-5.09)
(0.91)
Tanzania
0.034**
-0.119***
-0.00001*
-0.001
0.0005
(1.90)
(-2.76)
(-1.88)
(-0.97)
(1.19)
Togo
0.036***
0.002
0.006
-0.001*
-0.0001***
(3.87)
(1.60)
(0.43)
(-1.67)
(-2.77)
Tunisia
-0.001
-0.011
0.0009**
0.0007
0.00002
(-0.12)
(-0.88)
(2.37)
(0.91)
(0.29)
Uganda
0.066
0.015
-0.0008
-0.004
0.0001
(1.07)
(0.33)
(-0.68)
(-0.72)
(0.37)
Zambia
0.239***
-0.061
-0.0003*
-0.016***
-0.002**
(3.64)
(-0.40)
(-1.92)
(-2.98)
(-2.32)
Note: robust standard error correction is applied.
Adj R²
78.14
F-Stat
23.63
78.66
20.59
79.98
23.11
88.34
26.25
59.71
11.69
66.70
11.94
22.29
2.48
84.56
43.21
68.57
14.48
86.65
35.47
98.56
137.4
71.61
14.00
85.34
44.40
81.17
30.49
84.52
26.66
80.33
30.29
71.51
12.92
59.25
8.95
Page 18 of 21
P.K. Ozili & E.R. Outa
C
CFt-1
CFt-2
ΔCFR
ΔCFRt-1
EBCF
SIZE
SIZEt-1
ΔGDP
Table 4: Sensitivity Test
Forward-looking Discretion
Pre-and Post
(Arellano-Bond GMM)
Financial Crisis
(1)
(2)
(3)
0.037***
(11.81)
0.754**
0.759**
(2.43)
(15.59)
0.033
(0.99)
0.00002
(0.53)
0.001
0.002**
(0.47)
(2.55)
-0.040
-0.056
-0.266***
(-0.16)
(-1.16)
(-3.14)
-0.001***
(-7.16)
0.001
0.006
(0.06)
(0.80)
0.0007
0.0007***
0.0001
(0.49)
(2.80)
(1.25)
LISTED
LISTED*EBCF
CRISIS
without South
Africa
(5)
0.069***
(4.21)
0.0002
(0.61)
0.0001
(1.06)
-0.213***
(-6.47)
-0.002***
(-8.04)
-0.053***
(-3.24)
-0.004***
(-3.20)
0.0001
(1.26)
0.008***
(4.25)
-0.006
(-0.06)
0.00009
(1.29)
0.0009
(0.67)
0.077
(0.84)
CRISIS*EBCF
Sarjan (J) test
Hansen p-value
No of instrument
AR(1)
AR(2)
Listed vs
Unlisted
(4)
0.043***
(12.07)
27.81
0.63
44
0.000
0.378
25.24
0.66
44
0.000
0.448
Adjusted R²
13.39
15.78
74.34
F-statistic
54.26
63.13
20.08
Observation
1638
1365
1990
1990
1673
Column 1-4 is estimated with Arellano-Bond GMM estimation and includes robust standard errors clustered by
bank and year (Petersen, 2009). The Hansen J statistic test the adequacy of GMM instruments. AR(1) and AR(2)
test for the presence of first order and second order serial correlation. Column 5 and 6 is estimated using pooled
OLS because of the presence of time invariant variables. Column 7 is estimated with fixed effect OLS and
excludes bank samples from South Africa. T-statistics are reported in parentheses with ***, **, and * indicating
1%, 5%, and 10% significance level, respectively. CFt-1 = one-year lagged commission and fee income to total
asset ratio for bank i at year t-1. CFt-2 = two-year lagged commission and fee income to total asset ratio for bank
i at year t-2. SIZEt-1 = one-year lagged natural logarithm of total asset. ΔCFRt-1 = lagged change in the absolute
value of net commission and fee income for bank i at year t-1. SIZEt-1 = natural logarithm of total asset for firm i
at year t-1. LISTED = dummy variable that take the value 1 if the African bank is listed and zero otherwise.
CRISIS = dummy variable that take the value 1 during the period 2009, 2010, 2011, 2012 and 2013; and zero
otherwise.
Page 19 of 21
P.K. Ozili & E.R. Outa
Table 5: Definition of Variables
Variable
CF
SIZE
ΔCFR
EBCF
ΔGDP
LEGAL
INVPRO
Description
Net commission and fee income divided by total asset
Natural logarithm of total asset
Change in net commission and fee income outstanding
Earnings before net commission and fee income
(Profit before tax minus net commission and fee income)
divided by total asset
Real gross domestic product growth rate
Rule of law index measures the quality of the legal system
across countries.
Investor protection variable that measure the extent of
protection of minority shareholder rights.
LERNER Cross-county banking competitiveness
Source
Bankscope
Bankscope
Bankscope
World Economic forum
archived in Worldbank
database
Kaufmann, World
Governance indicator
La Porta from Doing
Busienss Project archived
in Worldbank Database
Bankscope archived in
Worldbank Database
Page 20 of 21
P.K. Ozili & E.R. Outa
Appendix
Panel A: Full Sample Correlation Matrix (with P-values in Parentheses)
Variables
CF
CF
1.000
ΔCFR
ΔCFR
0.039
0.144
1.000
EBCF
-0.376***
0.000
-0.018
0.484
1.000
-----
SIZE
-0.173***
0.000
-0.037
0.164
0.135***
0.000
ΔGDP
-0.004
0.886
0.058**
0.030
INVPRO
0.207***
0.000
-0.012
0.656
-0.044
0.105
0.123*** -0.192***
0.000
0.000
LEGAL
-0.062**
0.023
-0.029
0.282
0.035
0.191
0.154*** - 0.164*** 0.425***
0.000
0.000
0.000
LERNER
-0.074***
0.006
0.064**
0.019
0.071***
0.008
EBCF
SIZE
ΔGDP
INVPRO
LEGAL
LERNER
1.000
0.085*** -0.179***
0.002
0.000
-0.026
0.333
1.000
0.034
0.207
1.000
----1.000
-----
0.072*** -0.143***
0.008
0.000
1.000
Page 21 of 21