Professional Documents
Culture Documents
DOI 10.1007/s10551-014-2288-3
A. Palvia
Office of the Comptroller of the Currency, 400 7th Street, SW,
Washington, DC 20219, USA
e-mail: ajay.palvia@occ.treas.gov
E. Vahamaa
Department of Finance and Statistics, Hanken School of
Economics, P.O. Box 287, 65101 Vaasa, Finland
e-mail: emilia.vahamaa@hanken.fi
E. Vahamaa S. Vahamaa (&)
Department of Accounting and Finance, University of Vaasa,
P.O. Box 700, 65101 Vaasa, Finland
e-mail: sami@uva.fi
Introduction
Maleness has become a synonym for insufficient
attentiveness to risk.
(Christopher Caldwell in Time, 2009, Vol. 174, No. 7,
p. 13)
Women and men often act and behave differently.
Gender-based behavioral differences have been widely
documented in the cognitive psychology and behavioral
economics literature, and are perceived to be related to
information processing, diligence, conservatism, overconfidence, and risk tolerance (see e.g., Levin et al. 1988;
Feingold 1994; Powell and Ansic 1997; Byrnes et al. 1999;
Costa et al. 2001; Eckel and Grossman 2002; Nettle 2007;
Schmitt et al. 2008; Croson and Gneezy 2009). In this
paper, we presume that the gender-based behavioral differences between women and men are reflected in the
decisions that top executives and directors make, and
should therefore influence the major strategic and financial
decisions of their firms.
The purpose of this paper is to examine whether bank
capital ratios and default risk are associated with the genders of Chief Executive Officers (CEOs) and board Chairs.1
1
123
A. Palvia et al.
123
led banks were less likely to fail during the financial crisis.
This suggests that conservatism is particularly important
for the survival of smaller banks which may be less able to
absorb external shocks and often face less stringent market
and regulatory oversight.
We utilize instrumental variable regressions and propensity score matching to address endogeneity concerns,
and conduct a number of additional tests to ascertain the
robustness of our results. These tests give further evidence
that female-led banks are associated with higher capital
buffers and lower default rates amidst the crisis. Nevertheless, there are several limitations that should be considered when interpreting our empirical findings. Most
importantly, we are unable to rule out endogeneity caused
by omitted variable biases. We acknowledge that it is
possible female CEOs and board Chairs are symptoms of
some observable or latent factors that affect capital ratios
and the likelihood of bank failure which have not been
controlled for in our empirical analysis. Therefore, our
findings should be viewed as suggestive and need to be
interpreted with caution.
The rest of this paper proceeds in the following manner.
Related Literature section reviews the related literature
on gender-based behavioral differences as well as bank
performance and risk-taking around the recent financial
crisis. Data and Methodology section describes the data
on U.S. commercial banks and presents the methodology
used in the analysis. The empirical findings on the effects
of female CEOs and Chairwomen on capital ratios and
default risk are reported in Results section. Finally,
Conclusions section summarizes the results and provides
concluding remarks.
Related Literature
The implications of gender-based behavioral differences
for financial decisions have received increasing attention in
the literature over the last decades. The experimental
studies by Levin et al. (1988), Johnson and Powell (1994),
Powell and Ansic (1997), Eckel and Grossman ( 2002), and
Fehr-Duda et al. (2006) as well as the empirical studies on
real financial decisions related to household investment
portfolios e.g., by Jianakoplos and Bernasek (1998), Sunden and Surette (1998), Barber and Odean (2001), Dwyer
et al. (2002), Agnew et al. (2003), and Watson and
McNaughton (2007) suggest that women are more conservative and risk averse than men and exhibit less risky
behavior in personal financial decisions.4
4
Croson and Gneezy (2009) provide a review of gender differences
in economic experiments.
123
A. Palvia et al.
123
Definition
Dependent variables
Tier-1 capital
Capital ratio
Failure
A dummy variable which equals one for banks that fail within the next 12 months
Female variables
Female CEO
A dummy variable which equals one for banks that have a female CEO
Female Chair
A dummy variable which equals one for banks that have a female board Chair
A dummy variable which equals one for banks that have either a female CEO or a female board Chair
Control variables
Size
Large bank
Loan growth
A dummy variable which equals one for banks with above median total assets
Logarithm of loan growth
Core deposit
All deposits less deposits in large time-deposit and large-brokered deposit accounts divided by total deposits
Insured deposits
Delinquent loans
Ratio of loans at least 90 days past due or in nonaccrual status to total loans
ROA
Liquidity
Public
Subchapter S
A dummy variable which equals one for closely held banks that are organized under the subchapter-S
MBHC
A dummy variable which equals one for banks that are affiliated with a multibank holding company
CEO duality
A dummy variable which equals one for banks in which the CEO and Chair positions are held by the same individual
Unemployment
PCI
and Cihak (2014) show that the growth rate of the bank is
an important determinant of riskiness. Core deposits,
Insured deposits, Liquidity, and Delinquent loans measure
the stability of the banks funding structure and the quality
of the loan portfolio. These variables reflect funding and
lending risks and are known to be correlated with bank
performance amidst the financial crisis (Dietrich and
Wanzenried 2011; Altunbas et al. 2011; Beltratti and Stulz
2012; Cole and White 2012).
Public, Subchapter S, and MBHC dummies control for
the organizational and ownership structure of the bank.
Previous studies have documented that these variables may
affect firm-level decisions and governance structures (see
e.g., Depken et al. 2010; Holod 2012; Berger and Bouwman 2013; Cole and Mehran 2013). Given that local
macroeconomic developments are strongly correlated with
bank performance and bank failures (Laeven and Levine
2009; Aubuchon and Wheelock 2010; Altunbas et al. 2011;
Schaeck and Cihak 2014), we include Unemployment and
PCI to control for local economic conditions. Finally, we
include Capital ratio as a control variable in the bank
failure regressions because banks with more equity capital
are less likely to fail (see e.g., Cole and White 2012; Berger
and Bouwman 2013).
123
A. Palvia et al.
Descriptive Statistics
Table 2 reports the descriptive statistics for the three
different female variables (Female CEO, Female Chair,
and Female CEO or Chair), for the three alternative
dependent variables (Tier-1 capital, Capital ratio, and
Failure), and for the control variables. As shown in
Table 2, female CEOs and board Chairs are relatively
uncommon in the U.S. commercial banks. Only about
5.4 % of the banks (1248 firm-year observations) included
in our sample have a female CEO, and about 5.7 % of the
banks (1319 firm-year observations) have a female as the
Chairperson of the board. Nevertheless, it should be noted
that these low percentages of female CEOs and board
Chairs in commercial banks are consistent with the previously documented substantial underrepresentation of
women among top executives and directors in non-financial firms (see e.g., Krishnan and Park 2005; Francoeur
et al. 2008; Jurkus et al. 2011; Faccio et al. 2013; Huang
and Kisgen 2013). Regarding our capital measures,
Table 2 shows that U.S. banks are, on average, wellcapitalized with a mean (median) Tier-1 capital ratio of
about 10.1 (9.2) %. Correspondingly, the mean (median)
equity to total assets ratio is 10.6 (9.7) %. However, the
financial strength of banks varies considerably with the
25th75th percentile range for Tier-1 capital ratios being
almost 300 basis points from 8.2 to 11.1 %. Our four-year
sample contains 270 observations of banks failing within
the next year, representing approximately 1.2 % of the
bank-year observations.12
Table 2 further shows that the banks included in the
sample are very heterogeneous in terms of size, loan
growth, profitability, and the proportion of delinquent
loans. Given that our sample includes large publicly traded
systemically important financial institutions as well as
very small private commercial banks, it is not surprising to
observe that bank size exhibits considerable variation with
the logarithm of total assets ranging from 6.91 ($8.10
million) to 21.28 ($1.91 trillion). The mean (median) return
on assets for the banks in our sample is only 0.6 (0.8) %,
reflecting the severe impact of the financial crisis on bank
profitability. The proportion of delinquent loans to total
loans varies from zero to 32.7 %, with a mean (median) of
1.6 (0.8) %. In about 35 % of our sample banks, the same
individual holds the CEO and board chair positions (CEOChair duality). Finally, as shown in Table 2, approximately
21 % of the banks are publicly traded, about 35 % are
subchapter-S banks, and that almost 20 % of the banks are
affiliated with a multibank holding company.
12
The gender and financial data cover years 20072010 and we use
bank failures during years 2008-2011 in our tests.
123
Pairwise correlation coefficients (not tabulated) demonstrate that the three female dummy variables are positively correlated with Tier-1 capital and Capital ratio and
negatively correlated with Bank failure, suggesting that
female-led banks are associated with lower levels of risktaking.13 Furthermore, the female variables are negatively
correlated with Size, Dual, and Public, indicating that
female CEOs and Chairs are more common in smaller,
private banks, and that female CEOs are less likely to chair
the board. The female dummies are also positively correlated with Liquidity and Insured deposits, both of which
can be interpreted as measures of conservativeness of the
bank.
Results
Female CEOs, Chairwomen, and Bank Equity Capital
We begin our empirical analysis by conducting t-tests for
differences in the mean levels of equity capital between
female-led and male-led banks.14 Given that female CEOs
and Chairs are more common in smaller banks and,
moreover, that smaller banks generally hold more conservative levels of equity capital and are engaged in less risky
operations (see e.g., Berger and Bouwman 2013; Bhagat
et al. 2013), we conduct the t-tests separately for the
complete sample of 22,978 bank-year observations and for
subsamples of small and large banks.15 Consistent with our
research hypothesis, the univariate analysis indicates that
banks with female CEOs and/or board Chairs hold higher
levels of equity capital. In the full sample, banks with
female CEOs have on average 0.6 percentage-points higher
Tier-1 capital and equity capital ratios than banks with
male CEOs, and in the subsample of small banks the difference is even higher, being about 0.8 percentage-points.
The observed differences in capital ratios are statistically
significant at the 1 % level. The t-tests between banks with
female and male board Chairs further demonstrate that
female-led banks are associated with statistically significantly higher capital ratios. Specifically, we find that
female chaired banks have about 0.5 (0.4) percentagepoints higher Tier-1 capital ratios (capital ratios). Overall,
the observed differences in capital ratios are economically
13
Mean
SD
P1
P25
P50
P75
P99
Female variables
Female CEO
22,978
0.054
0.227
Female Chair
22,978
0.057
0.233
22978
0.094
0.292
Dependent variables
Tier-1 capital ratio
22,978
0.101
0.041
0.035
0.082
0.092
0.111
0.235
Capital ratio
22,973
0.106
0.043
0.037
0.084
0.097
0.117
0.241
Bank failure
22,978
0.012
0.108
Size
22,978
12.013
1.309
9.546
11.176
11.886
12.671
16.279
Loan growth
Delinquent loans
22,976
22,981
0.055
0.016
0.203
0.023
-0.292
0.000
-0.027
0.003
0.036
0.008
0.108
0.020
0.691
0.114
Return on assets
22,978
0.006
0.017
-0.047
0.003
0.008
0.013
0.030
Core deposits
22,978
0.826
0.116
0.472
0.766
0.841
0.906
1.000
Insured deposits
22,978
0.809
0.139
0.324
0.745
0.838
0.908
0.999
Liquidity
22,978
0.066
0.070
0.007
0.025
0.042
0.080
0.351
CEO duality
22,976
0.353
0.478
Public
22,978
0.212
0.409
Subchapter-S
22,978
0.351
0.477
MBHC
22,978
0.197
0.398
Unemployment
22,978
7.090
2.421
3.000
5.000
7.100
8.700
12.300
22,978
38.337
4.276
30.578
35.231
38.035
40.750
50.959
Control variables
The table reports summary statistics for the sample of U.S. commercial banks. Female CEO is a dummy variable which equals one for banks that
have a female CEO, Female Chair equals one if the banks Chairperson of the Board of Directors is a female, and Female CEO or Chair is
assigned to one if either the CEO or the board Chair of the bank is a female. Tier-1 capital is measured as Tier-1 capital scaled by total assets less
disallowed intangibles, Capital ratio is the ratio of the banks total equity capital to total assets, and Bank failure is a binary variable which
equals one for banks that fail within one year. Size is the logarithm of total assets, Loan growth is the logarithm of loan growth, Core deposit is
the core deposit ratio measured as all deposits less deposits in large time-deposit and large-brokered deposit accounts scaled by total deposits,
Insured deposits is the ratio of insured deposits to total deposits, Delinquent loans is the ratio of loans at least 90 days past due or in nonaccrual
status to total loans, ROA denotes return on assets calculated as the net income, Liquidity is measured as the ratio of cash balances to total assets,
Public is a dummy variable for publicly traded banks, Subchapter-S is assigned to one if a bank is organized under the subchapter-S, MBHC is a
dummy variable for the banks that are affiliated with a multibank holding company, CEO duality is a dummy variable which equals one for banks
in which the CEO and Chair positions are held by the same individual, Unemployment is the state unemployment rate, and Per capita income is
the state per-capita income PCI
123
123
-0.002** (0.001)
0.003*** (0.001)
0.001*** (0.000)
0.000 (0.000)
MBHC
CEO duality
Unemployment
PCI
49.87
14.60 %
22,971
Yes
0.000 (0.000)
0.001*** (0.000)
0.003*** (0.001)
0.447*** (0.153)
-0.177*** (0.041)
-0.021*** (0.006)
-0.039*** (0.012)
0.004 (0.007)
47.36
14.60 %
22,971
Yes
0.000 (0.000)
0.001*** (0.000)
0.003*** (0.001)
-0.002** (0.001)
-0.009*** (0.001)
0.004** (0.001)
0.056*** (0.013)
0.447*** (0.153)
-0.177*** (0.041)
-0.021*** (0.006)
-0.039*** (0.012)
0.004 (0.007)
-0.008*** (0.001)
-0.007*** (0.001)
0.002 (0.002)
0.004* (0.002)
0.224*** (0.018)
Model (3)
Tier-1 capital
50.22
14.60 %
22,971
Yes
0.000 (0.000)
0.001*** (0.000)
0.004*** (0.001)
-0.002** (0.001)
-0.009*** (0.001)
0.004** (0.001)
0.056*** (0.013)
0.447*** (0.153)
-0.177*** (0.041)
-0.021*** (0.006)
-0.039*** (0.012)
0.004 (0.007)
-0.008*** (0.001)
-0.007*** (0.001)
0.004** (0.002)
0.224*** (0.018)
Model (4)
Tier-1 capital
40.01
11.50 %
22,966
Yes
0.000 (0.000)
0.001*** (0.000)
0.003*** (0.001)
0.005*** (0.001)
-0.009*** (0.001)
0.006*** (0.002)
0.053*** (0.013)
0.462*** (0.166)
-0.177*** (0.045)
-0.019*** (0.006)
-0.037*** (0.014)
0.002 (0.007)
-0.009*** (0.001)
-0.005*** (0.001)
0.005** (0.002)
0.201*** (0.019)
Model (5)
Capital ratio
39.76
11.50 %
22,966
Yes
0.000 (0.000)
0.001*** (0.000)
0.003*** (0.001)
0.005*** (0.001)
-0.009*** (0.001)
0.006*** (0.002)
0.054*** (0.013)
0.463*** (0.166)
-0.176*** (0.045)
-0.019*** (0.006)
-0.037*** (0.014)
0.002 (0.007)
-0.009*** (0.001)
-0.005*** (0.001)
0.003 (0.002)
0.200*** (0.019)
Model (6)
Capital ratio
37.94
11.50 %
22,966
Yes
0.000 (0.000)
0.001*** (0.000)
0.003*** (0.001)
0.005*** (0.001)
-0.009*** (0.001)
0.006*** (0.002)
0.054*** (0.013)
0.462*** (0.166)
-0.177*** (0.045)
-0.019*** (0.006)
-0.037*** (0.014)
0.002 (0.007)
-0.009*** (0.001)
-0.005*** (0.001)
0.002 (0.002)
0.004** (0.002)
0.200*** (0.019)
Model (7)
Capital ratio
40.03
11.50 %
22,966
Yes
0.000 (0.000)
0.001*** (0.000)
0.003*** (0.001)
0.005*** (0.001)
-0.009*** (0.001)
0.006*** (0.002)
0.054*** (0.013)
0.462*** (0.167)
-0.176*** (0.045)
-0.019*** (0.006)
-0.037*** (0.014)
0.002 (0.007)
-0.009*** (0.001)
-0.005*** (0.001)
0.004** (0.002)
0.200*** (0.019)
Model (8)
Capital ratio
***, **, * Significance at the 0.01, 0.05, and 0.10 levels, respectively
The table reports the estimates of eight alternative versions of Eq. (1). The dependent variable is Tier-1 capital in Models 1-4 and Capital ratio in Models 5-8. Tier-1 capital is measured as Tier-1 capital scaled by total
assets less disallowed intangibles and Capital ratio is the ratio of the banks total equity capital to total assets. The female variables in the regressions are defined as follows: Female CEO is a dummy variable which
equals one for banks that have a female CEO, Female Chair equals one if the banks Chairperson of the Board of Directors is a female, and Female CEO or Chair is assigned to one if either the CEO or the board Chair
of the bank is a female. The control variables are defined as follows: Size is the logarithm of total assets, Large bank is a dummy variable for banks with above median total assets, Loan growth is the logarithm of loan
growth, Core deposit is the core deposit ratio measured as all deposits less deposits in large time-deposit and large-brokered deposit accounts scaled by total deposits, Insured deposits is the ratio of insured deposits to
total deposits, Delinquent loans is the ratio of loans at least 90 days past due or in nonaccrual status to total loans, ROA denotes return on assets calculated as the net income, Liquidity is measured as the ratio of cash
balances to total assets, Public is a dummy variable for publicly traded banks, Subchapter-S is assigned to one if a bank is organized under the subchapter-S, MBHC is a dummy variable for the banks that are affiliated
with a multibank holding company, CEO duality is a dummy variable which equals one for banks in which the CEO and Chair positions are held by the same individual, Unemployment is the state unemployment rate,
and Per capita income is the state per-capita income PCI. Robust standard errors corrected for clustering at the bank level are reported in parentheses
49.87
0.004** (0.001)
-0.009*** (0.001)
Public
Subchapter-S
F-statistics
0.056*** (0.013)
Liquidity
14.60 %
0.447*** (0.153)
Return on assets
Adjusted R2
-0.177*** (0.041)
Delinquent loans
22,971
-0.021*** (0.006)
Insured deposits
Yes
-0.002** (0.001)
-0.039*** (0.012)
Core deposits
No. of obs.
0.004** (0.001)
-0.009*** (0.001)
0.004 (0.007)
Loan growth
0.057*** (0.013)
-0.008*** (0.001)
-0.008*** (0.001)
-0.007***(0.001)
Large bank
-0.007*** (0.001)
0.003 (0.002)
0.224*** (0.018)
Model (2)
Tier-1 capital
Size
Control variables
Female Chair
Female CEO
0.004** (0.002)
0.225*** (0.018)
Constant
Female variables
Model (1)
Tier-1 capital
Variable
A. Palvia et al.
the amount of core and insured deposits, and the proportion of non-performing loans.
Models 58 in Table 3 are alternative versions of Eq. (1)
with Capital ratio as the dependent variable. The coefficient estimates for the three female dummies in Models
58 are all positive, and thereby indicate that female-led
banks hold higher levels of equity capital. Again, the
estimated coefficients are statistically significant for
Female CEO and Female CEO or Chair in Models 5, 7,
and 8, and statistically insignificant for Female Chair in
Models 6 and 7. Consistent with Models 14, the coefficients for the control variables suggest that larger banks
with lower amounts of core and insured deposits and more
delinquent loans are more likely to have lower capital
ratios.
Overall, the univariate tests as well as the panel
regressions presented in Table 3 indicate that banks with
female CEOs and/or Chairwomen are associated with
higher levels of equity capital. These findings provide
support for our hypothesis, and suggest that the genderbased differences in conservatism and risk tolerance may
have important implications for corporate decisions and
outcomes.
Instrumental Variable Regressions
Our research hypothesis implies that bank capital ratios are
affected by the gender of the banks CEO and board Chair.
We acknowledge that empirical tests of this hypothesis
may suffer from endogeneity problems and reverse causality. Although we have attempted to control for various
bank-specific characteristics as well as state-level macroeconomic developments, it is possible that we have omitted
correlated variables or some unobservable bank characteristics that simultaneously affect both the level of equity
capital and the appointment of female CEOs and board
Chairs. Furthermore, our findings may be influenced by a
self-selection bias if the gender-based differences in conservatism and risk tolerance induce women to self-select
into less risky banks that have more conservative capital
ratios.
In order to mitigate endogeneity concerns, we utilize
two-stage instrumental variable regressions to ascertain
whether bank capital ratios are affected by the gender of
the banks CEO and board Chair.16 We use two alternative
instruments in the first-stage regressions. First, following
Huang and Kisgen (2013), we use the level of gender status
equality in the state a bank is located in as an instrument
16
17
123
123
Yes
22,157
0.80 %
No. of obs.
Pseudo R2
Yes
22,971
1.00 %
No. of obs.
Pseudo R2
22,971
Yes
Yes
0.376** (0.191)
22,157
Yes
Yes
0.212** (0.097)
22,966
Yes
Yes
0.404* (0.211)
22,152
Yes
Yes
0.204** (0.097)
1.10 %
22,971
Yes
Yes
0.349*** (0.110)
1.00 %
22,157
Yes
Yes
0.000 (0.000)
22,971
Yes
Yes
0.041 (0.049)
22,157
Yes
Yes
1.325 (3.175)
22,966
Yes
Yes
0.038 (0.052)
22,152
Yes
Yes
1.274(3.062)
2.60 %
22,971
Yes
Yes
0.394*** (0.113)
2.60 %
22,157
Yes
Yes
0.001** (0.000)
22,971
Yes
Yes
0.128** (0.058)
22,157
Yes
Yes
0.200* (0.103)
22,966
Yes
Yes
0.138** (0.063)
22,152
Yes
Yes
0.192* (0.103)
***, **, * Significance at the 0.01, 0.05, and 0.10 levels, respectively
The table reports the estimates of the two-stage instrumental variable regressions. In the first-stage regressions, we use two alternative instruments for the female variables. In Panel A, the instrumental variable is
the level of gender status equality in the state a bank is located in. In Panel B, the instrumental variable is the commonness of female-led banks in the state a bank is located in. In the second-stage regressions,
Tier-1 capital and Capital ratio are regressed on the fitted values of the female variables from the first-stage regressions and all the control variables used in Eq. (1). Tier-1 capital is measured as Tier-1 capital
scaled by total assets less disallowed intangibles and Capital ratio is the ratio of the banks total equity capital to total assets. The female variables in the regressions are defined as follows: Female CEO is a
dummy variable which equals one for banks that have a female CEO, Female Chair equals one if the banks Chairperson of the Board of Directors is a female, and Female CEO or Chair is assigned to one if
either the CEO or the board Chair of the bank is a female. The control variables are defined as follows: Size is the logarithm of total assets, Large bank is a dummy variable for banks with above median total
assets, Loan growth is the logarithm of loan growth, Core deposit is the core deposit ratio measured as all deposits less deposits in large time-deposit and large-brokered deposit accounts scaled by total deposits,
Insured deposits is the ratio of insured deposits to total deposits, Delinquent loans is the ratio of loans at least 90 days past due or in nonaccrual status to total loans, ROA denotes return on assets calculated as the
net income, Liquidity is measured as the ratio of cash balances to total assets, Public is a dummy variable for publicly traded banks, Subchapter-S is assigned to one if a bank is organized under the subchapter-S,
MBHC is a dummy variable for the banks that are affiliated with a multibank holding company, CEO duality is a dummy variable which equals one for banks in which the CEO and Chair positions are held by
the same individual, Unemployment is the state unemployment rate, and Per capita income is the state per-capita income PCI. Robust standard errors corrected for clustering at the bank level are reported in
parentheses
Yes
0.269** (0.111)
Control variables
Instrumented Chair
Instrumented CEO
Female variables
Instrument
Yes
0.001*** (0.000)
Control variables
Instrumented Chair
Instrumented CEO
Female variables
Instrument
Variable
A. Palvia et al.
The results of the t-tests are not tabulated for brevity, but are
naturally available upon request.
22
In our sample, the failure rates of large and small banks are 1.6 and
0.7 %, respectively.
23
Specifically, we estimate failure prediction regressions with bank
size interactions of the following form: Failurej,t?1 = a ?
bankj,t ? b2Femalej,t 9 Large
bankj,t ?
b1Femalej,t 9 Small
b3Capitalj,t ? b4Sizej,t ? b5Loan growthj,t ? b6Core depositsj,t ?
depositsj,t ? b8Delinquent
loansj,t ? b9ROAj,t ?
b7Insured
b10Liquidityj,t ? b11Publicj,t ? b12Subchapter Sj,t ? b13MBHCj,t ?
? b1719
(Year
b14Dualj,t ? b15Unemploymentj,t ? b16PCIj,t
dummies)j,t ? ej,t.
It can be argued that CEOs and board Chairs may have a stronger
influence on corporate decisions-making in smaller, privately-owned
firms. Moreover, as noted by Holod (2012), the CEOs of smaller,
private banks are more likely to hold large ownership stakes, and may
therefore have very different incentives than the CEOs of large
publicly traded banks.
123
A. Palvia et al.
123
interacted with small bank and large bank indicator variables. The estimates of these regressions (not tabulated)
provide further evidence to suggest that female-led banks
are associated with higher capital ratios. The estimates
indicate that the positive relationship between female
CEOs and capital ratios is driven by smaller banks, while
female board Chairs have a positive effect on capital ratios
only in larger banks.
In our main analysis, we use Tier-1 capital and Capital
ratio to assess the capital position of the bank. To further
ascertain the robustness of our empirical findings, we
estimate logit panel regressions with a binary variable
Well-capitalized bank as the dependent variable. This risktaking proxy equals one for banks that are well-capitalized
under the Federal Deposit Insurance Corporation (FDIC)
Improvement Act definition for prompt corrective action
by bank regulators.26 The results of these additional
regressions (not tabulated) indicate that female-led banks
are more likely to be well-capitalized under the FDIC
definition than male-led banks.
Fourth, in our main regressions, we use contemporaneous
data on the dependent and independent variables. We
examine the robustness of our results by regressing the risktaking variables on one-year lagged female dummies and
control variables. The lagging of the independent variables
should also further mitigate endogeneity concerns. However,
because data on the female variables are available only from
2007 onwards, these additional regressions are based on a
shorter three-year sample. The estimation results of the
specifications with lagged independent variables (not tabulated) are qualitatively similar to the results reported in
Tables 3 and 5, and thereby suggest that our findings are not
sensitive to lagging of the variables.
Fifth, we acknowledge that the failure prediction
regressions may also be influenced by a self-selection bias
if women self-select into less risky banks that are less
likely to fail.27 To alleviate these concerns, we estimate
two-stage instrumental variable regressions for bank failures. Consistent with the results reported in Table 5, the
instrumental variable regressions (not tabulated) suggest
that female-led banks were less likely to fail during the
crisis, as the coefficient estimates for the instrumented
Female CEO and the instrumented Female CEO or Chair
26
Under this definition, a bank is well-capitalized if the ratio of Tier1 capital to total assets (Tier-1 leverage ratio) is at least 5 %, the ratio
of Tier-1 capital to risk-weighted assets (Tier-1 risk-based ratio) is at
least 6 %, and the ratio of the sum of Tier-1 and Tier-2 capital to riskweighted assets (total risk-based ratio) is at least 10 %.
27
It should be noted that Eq. (2) is a failure prediction regression in
which future bank failures are predicted with variables that are
currently observable. In this type of prediction setup, there cannot be
simultaneity issues and reverse causality would essentially require
that a failure at a future point in time causes the gender of the banks
CEO and board Chair at the present time.
Model (5)
Yes
22,966
59.00 %
664.17
No. of obs.
Pseudo R2
v2 statistics
663.51
58.98 %
22,966
Yes
665.96
59.00 %
22,966
Yes
661.39
58.99 %
22,966
Yes
Yes
665.30
59.06 %
22,966
Yes
662.61
59.04 %
22,966
Yes
Yes
664.42
59.12 %
22,966
Yes
Yes
662.45
59.10 %
22,966
Yes
Yes
***, **, * Significance at the 0.01, 0.05, and 0.10 levels, respectively
The table reports the estimates of eight alternative versions of Eq. (2). Failure is a binary variable which equals one for banks that fail within 1 year. The female variables in the regressions are defined as follows:
Female CEO is a dummy variable which equals one for banks that have a female CEO, Female Chair equals one if the banks Chairperson of the Board of Directors is a female, and Female CEO or Chair is assigned to
one if either the CEO or the board Chair of the bank is a female. The following control variables are used in the regressions (not tabulated): Size is the logarithm of total assets, Large bank is a dummy variable for banks
with above median total assets, Tier-1 capital is measured as Tier-1 capital scaled by total assets less disallowed intangibles, Loan growth is the logarithm of loan growth, Core deposit is the core deposit ratio measured
as all deposits less deposits in large time-deposit and large-brokered deposit accounts scaled by total deposits, Insured deposits is the ratio of insured deposits to total deposits, Delinquent loans is the ratio of loans at
least 90 days past due or in nonaccrual status to total loans, ROA denotes return on assets calculated as the net income, Liquidity is measured as the ratio of cash balances to total assets, Public is a dummy variable for
publicly traded banks, Subchapter-S is assigned to one if a bank is organized under the subchapter-S, MBHC is a dummy variable for the banks that are affiliated with a multibank holding company, CEO duality is a
dummy variable which equals one for banks in which the CEO and Chair positions are held by the same individual, Unemployment is the state unemployment rate, and Per capita income is the state per-capita income
PCI. The reported estimates are based on logistic panel regressions. Robust standard errors corrected for clustering at the bank level are reported in parentheses
Yes
Control variables
-1.675*** (0.545)
0.095 (0.287)
0.287 (0.315)
Model (8)
0.259 (0.308)
-1.354** (0.673)
-0.144 (0.416)
-2.009*** (0.682)
Model (7)
-1.364** (0.669)
Model (6)
Yes
-0.167 (0.263)
Model (4)
-0.076 (0.409)
Yes
0.086 (0.288)
-0.332 (0.377)
Model (3)
-2.015*** (0.672)
Yes
0.023 (0.288)
Model (2)
-0.315 (0.376)
Model (1)
Female Chair
Female CEO
Female variables
Variable
123
A. Palvia et al.
Table 6 Matched-sample regressions
Variable
Model (1)
Tier-1 capital
Model (2)
Tier-1 capital
Model (3)
Tier-1 capital
Model (4)
Capital ratio
Model (5)
Capital ratio
Model (6)
Capital ratio
0.005** (0.002)
Female Chair
0.005** (0.002)
0.002 (0.002)
0.001 (0.002)
0.004** (0.002)
0.004** (0.002)
Control variables
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
No. of obs.
2,472
2,612
4,271
2,472
2,612
4,271
Adjusted R2
13.30 %
14.60 %
15.50 %
10.60 %
14.00 %
12.70 %
F-statistics
11.00
7.37
13.55
8.23
6.43
11.00
Variable
Model (1)
Failure
Model (2)
Failure
Model (3)
Failure
Model (4)
Failure
Model (5)
Failure
Model (6)
Failure
0.549 (0.872)
Female Chair
0.354 (0.707)
-0.461 (0.586)
-3.355* (2.014)
1.524 (1.247)
-6.161*** (2.101)
1.626 (1.211)
-3.413*** (1.332)
0.167 (0.678)
Control variables
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
No. of obs.
2,472
2,612
4,271
2,472
2,612
4,271
Pseudo R2
78.85 %
84.58 %
72.55 %
80.12 %
86.33 %
73.73 %
v2 statistics
99.31
469.90
157.95
85.22
313.82
149.70
The table reports the estimates of six alternative versions of Eqs. (1) and (2) using a propensity score matched sample of banks. In Panel A, the dependent variable
Tier-1 capital in Models 13 and Capital ratio in Models 46. Tier-1 capital is measured as Tier-1 capital scaled by total assets less disallowed intangibles and Capital
ratio is the ratio of the banks total equity capital to total assets. In Panel B, the dependent variable is Failure, which is a binary variable which equals one for banks
that fail within 1 year. The female variables in the regressions are defined as follows: Female CEO is a dummy variable which equals one for banks that have a female
CEO, Female Chair equals one if the banks Chairperson of the Board of Directors is a female, and Female CEO or Chair is assigned to one if either the CEO or the
board Chair of the bank is a female. The following control variables are used in the regressions (not tabulated): Size is the logarithm of total assets, Large bank is a
dummy variable for banks with above median total assets, Loan growth is the logarithm of loan growth, Core deposit is the core deposit ratio measured as all deposits
less deposits in large time-deposit and large-brokered deposit accounts scaled by total deposits, Insured deposits is the ratio of insured deposits to total deposits,
Delinquent loans is the ratio of loans at least 90 days past due or in nonaccrual status to total loans, ROA denotes return on assets calculated as the net income,
Liquidity is measured as the ratio of cash balances to total assets, Public is a dummy variable for publicly traded banks, Subchapter-S is assigned to one if a bank is
organized under the subchapter-S, MBHC is a dummy variable for the banks that are affiliated with a multibank holding company, CEO duality is a dummy variable
which equals one for banks in which the CEO and Chair positions are held by the same individual, Unemployment is the state unemployment rate, and Per capita
income is the state per-capita income PCI. The specifications with Tier-1 capital and Capital ratio as the dependent variable are estimated as panel regressions, while
the specifications Failure as the dependent variable are estimated as logistic panel regressions. Robust standard errors corrected for clustering at the bank level are
reported in parentheses
***, **, * Significance at the 0.01, 0.05, and 0.10 levels, respectively
123
Holod (2012) argues that the CEOs of smaller, private banks are
more likely to hold larger ownership stakes of those banks, while the
CEOs of larger, publicly traded banks are more likely to be hired
agents with relatively small ownership stakes.
29
A minority-owned financial institution is defined as a bank that is
at least 51 % owned by minorities, such as African Americans,
Hispanic Americans, or women.
30
Anecdotal evidence suggests that firms in which the CEO and the
Chairperson of the board are relatives (i.e., spouses, siblings or
descendants) tend to be family controlled. Thus, we presume that a
common last name of the CEO and board Chair can be used as a
proxy for controlling family ownership.
The evidence regarding the effects of executive education on risktaking is mixed. While Bertrand and Schoar (2003) find that
executives with MBA degrees follow more aggressive business
strategies, the results of Berger et al. (2014) suggest that executives
with higher levels of education are more conservative. Ahern and
Dittmar (2012) and Huang and Kisgen (2013) document that female
executives and directors are more likely to hold an MBA degree.
123
A. Palvia et al.
Furthermore, it should be noted that our sample is limited to four fiscal years around the financial crisis. We
argue that this period of severe financial turmoil provides
an expedient setting to examine the effects of female CEOs
and Chairwomen on capital ratios and bank failures.
Nevertheless, it is possible that the relation between
executive gender and the degree of bank conservatism is
different in different business cycles. Moreover, given the
short sample period, we are unable to analyze the relationship between bank risk-taking and CEO and Chair
gender through time. It would be interesting to examine,
for instance, if the appointment of a female CEO actually
leads to an increase in the banks capital buffers. Finally,
we acknowledge that our sample of U.S. commercial banks
is severely unbalanced toward male-led banks and femaleled banks comprise only about 5 % of the observations.
This low proportion of banks with female CEOs and
Chairwomen may create a bias in our estimations, which
we have tried to alleviate in our robustness checks through
the use of propensity score matching.
Conclusions
The purpose of this paper is to examine whether bank
capital ratios and default risk are associated with the gender
of the banks CEO and Chairperson of the board. In particular, using a large panel of U.S. commercial banks, we
empirically examine whether banks with female CEOs and
board Chairs are associated with more conservative levels
of equity capital and lower default risk. Our analysis is
motivated by the well-documented behavioral differences
between women and men. Given that women are generally
more conservative and less inclined to take extreme risks,
we postulate that female CEOs and board Chairs assess
risks more conservatively, and may thereby hold higher
levels of equity capital and reduce the default risk of their
banks during periods of market stress. Since the recent
financial crisis has often been attributed to excessive risktaking by banks, and was characterized by numerous bank
failures and bailouts, we consider this period of severe
financial turmoil to provide an expedient setting to examine
the potential effects of female leadership on capital buffers
and actual bank failures.
The empirical findings reported in this paper demonstrate that bank capital ratios and the likelihood of failure
during the financial crisis are associated with the gender of
the banks CEO and Chairperson of the board. In particular, we document that banks with female CEOs are more
conservative and hold higher levels of equity capital after
controlling for the banks asset risk and other attributes.
The observed differences in capital ratios are economically
significant and indicate that female-led banks hold about
123
56 % more equity capital than male-led banks. Furthermore, we document a negative association between female
CEOs and Chairwomen and bank default risk during the
financial crisis. Although neither CEO nor Chair gender is
related to bank failure in general, we find strong evidence
suggesting that smaller female-led banks were less likely to
fail during the financial crisis. This finding may indicate
that female-induced conservatism is particularly important
for the survival of smaller banks that are less able to absorb
external shocks and may face less stringent market and
regulatory oversight.
Overall, the results documented in this paper provide
support for the view that female executives and directors may
inherently promote more conservative strategies and less risky
financial decisions. In 2010, Ms. Christine Lagarde of the IMF
famously stated that if Lehman Brothers had been Lehman
Sisters, todays economic crisis clearly would look quite different (Lagarde 2010). Our empirical analysis demonstrates
that Ms. Lagardes provocative argument may contain at least
some element of truth to it. We believe that our findings may
have important implications for bank supervisors, regulators,
depositors, and other stakeholders. In general, our results
suggest that the advancement of women in the banking
industry may be consistent with stakeholders interests. The
main supervisory implication is that executive gender may
contain useful complementary information for evaluating the
safety and soundness of banks. From a public policy perspective, the documented benefits of female leadership for
bank stability may be of interest to regulators when setting
future policies for promoting gender equality and the
advancement of women in business.
Acknowledgments We wish to thank two anonymous referees,
David Aristei, Allen N. Berger, Gerald P. Dwyer, Jason Park, Seppo
Pynnonen, Daniel A. Rogers, and seminar participants at Stockholm
University, Hanken School of Economics, Pablo de Olavide University, the Office of the Comptroller of the Currency, the 53rd Southern
Finance Association Meeting, the 49th Eastern Finance Association
Meeting, the 25th Australasian Finance and Banking Conference, and
the 16th International Conference on Macroeconomic Analysis and
International Finance for insightful comments and discussions. E. and
S. Vahamaa gratefully acknowledge the financial support of the
Academy of Finland, the Foundation for Economic Education, the
Marcus Wallenberg Foundation, and the NASDAQ OMX Nordic
Foundation. All views expressed in this paper are those of the authors
alone and do not necessarily reflect those of the Office of the
Comptroller of the Currency or the U.S. Department of the Treasury.
References
Adams, R., Almeida, H., & Ferreira, D. (2005). Powerful CEOs and
their impact on corporate performance. Review of Financial
Studies, 18, 14031432.
Adams, R., & Ferreira, D. (2009). Women in the boardroom and their
impact on governance and performance. Journal of Financial
Economics, 94, 291309.
123
A. Palvia et al.
Krishnan, H., & Park, D. (2005). A few good womenOn top
management teams. Journal of Business Research, 58, 17121720.
Krishnan, G., & Parsons, L. (2008). Getting to the bottom line: An
exploration of gender and earnings quality. Journal of Business
Ethics, 78, 6576.
Laeven, L., & Levine, R. (2009). Bank governance, regulation and
risk taking. Journal of Financial Economics, 93, 259275.
Lagarde, C. (2010). Women, power and the challenge of the financial
crisis. The International Herald Tribune, May 11, 2010.
Levin, I., Snyder, M., & Chapman, D. (1988). The interaction of
experimental and situational factors and gender in a simulated
risky decision-making task. Journal of Psychology, 122, 173
181.
Malmendier, U., & Tate, G. (2005). CEO overconfidence and
corporate investment. Journal of Finance, 60, 26612700.
Malmendier, U., Tate, G., & Yan, J. (2011). Overconfidence and
early-life experiences: The effect of managerial traits on
corporate financial policies. Journal of Finance, 66, 16871733.
Nettle, D. (2007). Empathizing and systemizing: What are they, and
what do they contribute to our understanding of psychological
sex differences? British Journal of Psychology, 98, 237255.
Peni, E. (2014). CEO and Chairperson characteristics and firm
performance. Journal of Management and Governance, 18,
185205.
123