Board gender diversity and bank risk-taking: Evidence
from the European Union
Kirsten Jager
a,*Supervisor: Dr. V. Purice Co-assessor: Prof. Dr. C.L.M. Hermes
June 13, 2016
a
MSc International Financial Management, Faculty of Economics and Business, University of Groningen, Groningen, The Netherlands
Abstract
This paper extends previous research on the effect of gender diversity on boards of directors by examining the relationship between board gender diversity and the risk-taking of banks. Based on a sample of 95 European banks studied over the period 2002-2014, this study reports a positive relationship between gender diversity and bank risk taking. Risk-taking is, hereby, defined as both portfolio risk and insolvency risk. Furthermore, the results indicate that the relationship between gender diversity and risk-taking increases with country-level shareholder protection and bank-level corporate governance.
Key words: Board diversity, Banks, Gender, Risk-taking, Corporate governance JEL classification: G21, G34, J16
1
1. Introduction
“[I]f Lehman Brothers had been ‘Lehman Sisters’, would the crisis have happened like it did? No.”, argued the former European Union (EU) Commissioner for Competition Neelie Kroes (Forbes, 2009). Her philosophy is based on the idea that women are more risk-averse than men and, consequently, behave differently in the boardroom. However, this argument is still subject to discussion as prior to the financial crisis, female representation on the board of directors of banks had increased steadily while the financial crisis showed that banks were still widely engaged in risky activities. In some countries, the failures in the financial system even caused the entire country teetering perilously on the edge of the abyss. Thus, although Neelie Kroes suggested that increased gender diversity would lead to less risk-taking, it is still not clear whether female directors actually take less excessive risks than their male counterparts.
During the financial crisis, the banking sector has been sharply criticized. As a result of their short-term perspective, many financial institutions were found to take excessive risks in order to make high profits. One of the major deficiencies of the financial sector causing the financial crisis is often argued to be the inadequate corporate governance structures of banks (De Haan and Vlahu, 2016). Given the key role banks play in the financial intermediation in the economy, shortcomings in the corporate governance of banks not only affect individual banks, but also the economy as a whole. In 2010, the Basel Committee on Banking Supervision (2010) has, therefore, responded to these deficiencies by outlining several key areas that required immediate improvements, among which the composition of executive boards. The Basel Committee hereby argued that good corporate governance structures are necessary to guarantee the soundness of the financial system and economic growth, although no specific remarks were made with respect to gender diversity.
2
2020 at least 40 per cent of the non-executive board positions in publicly listed companies should be occupied by women (European Commission, 2016). The European Commission hereby argues that an improved balance between men and women in decision-making positions improves competitiveness and company performance (European Commission, 2012).
Next to the directive adopted by the European Commission, various member states have undertaken new initiatives, including both legislative measures and voluntary initiatives, to promote the presence of women in boardrooms (European Commission, 2016). The most recent example is Germany where the largest publicly listed companies are obliged to ensure that from 2016 at least 30 percent of the supervisory board positions are occupied by women. Furthermore, Germany plans to increase this proportion to 50 percent by 2018 (Deutscher Bundestag, 2015). In case the quota is not met, the respective company is obliged to fill vacant board positions with women until the quota is reached or the vacancy should be left empty. Other examples of EU countries that introduced gender quota enforced by sanctions in the case of non-compliance, include Belgium, France and Italy. Furthermore, the Netherlands has adopted a ‘comply or explain’ mechanism in which firms that do not meet the threshold of 30 percent have to explain why this objective could not be met. Lastly, the United Kingdom continues to rely on voluntary targets in which large companies are encouraged to present their board diversity targets (European Commission, 2012). These examples illustrate that instead of being an exception, gender quotas are quickly becoming the norm in many EU member states.
3
regulators (García-Meca et al., 2015). This research aims to contribute to the existing literature by focusing exclusively on financial institutions.
This paper makes several contributions to the literature on gender diversity and risk-taking, and extends the research by both Berger et al. (2014) and Sila et al. (2016) in several ways. First of all, this research focuses on the period 2002-2014, thereby extending the previous studies that focused on the period up to 2010, and capturing the recent changes in board diversity. Furthermore, in contrast to Sila et al. (2016), this paper does not exclude regulated industries, but instead focuses exclusively on the financial sector. Thirdly, this study extends the single-country research by Berger et al. (2014) by using a sample of banks from 19 European countries. Moreover, the moderating effect of country-level investor protection and bank-level corporate governance on the relationship between gender diversity and a bank’s risk taking is examined.
The results as presented in this study indicate a positive the relationship between gender diversity and risk taking, both in terms of portfolio risk and insolvency risk. Although the difference-in-difference (DID) technique yields contradictory results suggesting a negative relationship between gender diversity and portfolio risk, these results may be biased as a result of the small sample. Additionally, the second part of this study demonstrates that the relationship between gender diversity and risk-taking increases with both country-level investor protection and bank-level corporate governance.
4
2. Literature Review and hypotheses development
This section examines the relevant theories and empirical evidence that explain how executive board diversity can affect bank risk-taking. Based on this, the main hypotheses are formulated.
2.1 Corporate governance and the board of directors
Corporate governance can be defined as “a group of mechanisms used by stakeholders to ensure that directors efficiently manage corporate resources” (Andres and Vallelado, 2008, p. 2571). Shareholders can use various instruments to ensure that the management acts in their interest, including the election of the board of directors, management remuneration schemes, and concentrated ownership (De Haan and Vlahu, 2016). Since this paper focuses on the effect of the board of directors on a bank’s risk-taking behavior, the first instrument will be discussed in more detail. The boards of directors in financial institutions differ from boards in nonfinancial institutions. More specifically, boards in the financial sector tend to be larger and more independent than boards in nonfinancial institutions (Andres, Romero-Merino, Santamaría and Vallelado, 2012). Furthermore, the “directors are subject to more scrutiny than directors of publicly traded non-bank corporations” (García-Meca et al., 2015, p. 203). In this way, the responsibilities of banks’ boards of directors are far more widespread than solely acting in the best interest of the shareholders. Instead banks are also accountable to other stakeholders, since banks are mostly funded by depositors and are subject to extensive regulation (Andres and Vallelado, 2008). Hereby, the complexity of the financial sector exacerbates potential agency problems, which is discussed in the subsequent section.
2.2 Relevant theories
5
2.2.1 Principal-agent theory
The basis for principal-agent theory is the separation between ownership and control in many organizations. It predicts that the agents, or the managers, of a firm do not always behave in the best interest of the principals, the owners, of the firm (Jensen and Meckling, 1976). This is based on the assumption that both the agent and the principal aim to maximize their utility. Principal-agent theory is thereby based on the concepts of risk and return. Shareholders prefer to engage in risky projects with positive net present values, because these will generate higher returns. Due to the concept of limited liability, potential losses will be shared between both the shareholders and the bondholders, while profits will be mainly benefited from by the shareholders (Demsetz, Saidenberg and Strahan, 1997). This is also shown by Laeven and Levine (2008) who found that owners with significant cash flow rights favor increased risk-taking, thereby confirming the statement that shareholders tend to prefer higher risk-taking than managers. Furthermore, bondholders have low monitoring incentives as they are often protected by their respective government in the form of a deposit guarantee system. Therefore, one of the tasks of the managers is to closely scrutinize both the risk and the return of the adopted strategy. Principal-agent theory argues that managers prefer relatively low risk-taking as to protect their job, reputation and invested wealth (Adams and Ferreira, 2009). This results in a principal-agent problem in which the shareholder prefers risky projects while the manager prefers to reduce risk-taking.
The principal, or in this case the shareholder, aims to ensure that the agent acts in the best interest of the principal through monitoring and bonding. Nevertheless, the principal cannot perfectly monitor the agent since the agent always has superior information and this calls for additional mechanisms to prevent managers from diverting profits for their own benefit (Jensen and Meckling, 1976). The principal-agent theory is thus based on information asymmetries that exist between the principal and the agent. Although these asymmetries exist in every sector, the problems are aggravated in the financial sector due to the complexity of the actions of the financial intermediaries (Andres and Vallelado, 2008). These complexities make it difficult for the principal to monitor the agent. Here, the board of directors comes in as a third-party to monitor the management.
6
the board also has an advising role in which it provides advice to the managers with respect to strategic decisions. Although several studies have argued that the board of directors does not have a large influence on the firm’s corporate strategy, recently researchers have argued that executives indeed have an influence on the firm’s strategy, and thereby its risk-taking (Tarus and Aime, 2014). From the principal-agent perspective, gender diversity on the board of directors can improve the organization’s corporate governance mechanisms because it improves the board’s monitoring abilities of the management (Adams and Ferreira, 2009). The reason behind this is that women are argued to have higher expectations with respect to their role as a board member, which results in better monitoring (Fondas and Sassalos, 2000). However, when studying the effect of gender diversity in the boardroom, one should not only focus on individual characteristics, but should also take into account the effect of the diversity in the group as a whole. In this way, Berger et al. (2014) state that “[t]he individual effort provided by a group member is likely to be influenced by group characteristics that determine the degree of mutual monitoring” (p. 49). Here it is argued that a more diverse board is often accompanied by increased board independence, and this improves the ability of the board to monitor the managers (Adams, Haan, Terjesen and Ees, 2015). Additionally, more diverse groups are better able to incorporate various viewpoints and identify a range of solutions (Watson, Kumar and Michaelsen, 1993). So, in order for a board to be effective, it should consist of diverse people who have different, but complementary, backgrounds and personalities. One example of how this diversity can be achieved is through gender diversity (Van den Berghe and Levrau, 2004). Nevertheless, it should be noted that too much heterogeneity on the board of directors can actually harm the decision-making process due to conflicts and misunderstandings (García-Meca et al., 2015).
7
2.2.2 Resource dependence theory
Although agency theory is widely applied in the literature on board gender diversity, resource dependence theory provides an additional perspective on the role of the board of directors. Resource dependence theory argues that organizations are dependent on resources, such as tangible and financial resources, which can be found in their environment (Pfeffer and Salancik, 1978). The organization, therefore, aims to align its resources with the environment. The board of directors can be seen as one of the most important mechanisms for connecting the organization to its environment. In contrast to agency theory, which focuses mainly on the monitoring role of the board of directors, resource dependence theory argues that the board plays a crucial role in the provision of resources and protection of these resources through connections with the external environment (Hillman, Cannella and Paetzold, 2000). In this way, the board of directors can manage interdependencies with the external environment, the corresponding transaction costs, and it can reduce external uncertainty.
By selecting appropriate directors who possess the necessary skills and linkages to important stakeholders, an organization can reduce its dependency on the environment (Hillman, Shropshire and Cannella, 2007). Recently, many organizations have been pressured by their environment to increase the number of female directors because it is argued that increased female representation on the board can improve the legitimacy of the organization. This stems from the idea that a better representation of the demographic characteristics of the stakeholders can attract employees, customers and investors (Mateos de Cabo, Gimeno and Nieto, 2012). So, by increasing the gender diversity on the board, the linkages with the environment but also with the employees can be improved.
8
conflicts are becoming increasingly likely to occur (Berger et al., 2014; García-Meca et al., 2015). Both resource dependence theory and upper echelons theory, which is discussed in the next section, argue that board diversity in terms of demographic differences of the directors can improve the decision-making process since heterogeneous boards allow the board to encompass different viewpoints and consider a wider range of alternatives (Adams and Funk, 2012; Mateos de Cabo et al., 2012). In this way, research suggests that heterogeneous groups outperform homogeneous groups (Hambrick and Mason, 1984; Van Ginkel and Van Knippenberg, 2008).
The effect of gender diversity from a resource dependency perspective depends on the strategy of the organization (Hillman et al., 2000). Hillman et al. (2007) argue that organizations that operate in more complex environments are more dependent upon their environment, and can therefore benefit the most from board gender diversity. Since the financial sector is considered highly complex, this theory implies that banks can significantly benefit from a higher level of female representation on their boards. Resource dependence theory is often seen in conjunction with upper echelons theory, which will be discussed in the subsequent section. While resource dependence theory argues that the board of directors is important with respect to the interconnections with the external environment, upper echelons theory provides more insights into how gender diversity on the board can affect a bank’s risk-taking.
2.2.3 Upper echelons theory
Upper echelons theory is largely related to resource dependence theory and states that directors’ experiences, values and personal characteristics affect the way they perceive situations, and in turn, affect the choices they make (Hambrick, 2007). This theory is based on the idea of bounded rationality, which argues that due to limited information, selective perception and filtering, individuals cannot objectively judge situations when making decisions. Instead strategic decisions are subject to an interpretation bias that is defined by the executive’s perception and personal values (Hambrick and Mason, 1984).
9
assumption that “female directors could influence the decision making of the board […] through gender differences in risk propensity” (Jane Lenard, Yu, Anne York and Wu, 2014, p. 790). One of the psychological arguments brought forward is that men score higher on self-efficacy, which means that they have more confidence in their own performance and, as a result, they are more prone to take risks (Forlani, 2013; Rosenthal, Guest and Peccei, 1996). Furthermore, men are also more likely to have a sensation-seeking personality and, therefore, they are more willing to bear the risk for the sake of the sensation (Zuckerman, 2007). Thirdly, female executives are likely to have different experience and knowledge in comparison to their male counterparts, thereby altering the information available for taking decisions. One example is that female executives tend to be more highly educated than male directors, which has a hampering effect on risk-taking (Hillman, Cannella and Harris, 2002). Nevertheless, it is also argued that female directors generally tend to be younger and less experienced than male directors, which is likely to cause increased risk-taking (Berger et al., 2014).
Thus, from this psychological perspective, an increase in the number of female directors is expected to lead to a decrease in risk-taking, however the opposite is expected if female directors are younger and less experienced. Furthermore, one has to keep in mind that studies in the psychological field that found a negative relationship are largely based on the general population, while studies that focus specifically on directors in the financial sector, have failed to prove that women are indeed more risk-averse (e.g. Adams and Funk, 2012; Deaves, Lüders and Luo, 2008).
2.3 Empirical evidence on gender diversity and risk-taking
10
Thirdly, upper echelons theory argues that increased board diversity is likely to reduce risk-taking due to intrinsic differences between men and women. Wilson and Altanlar (2009) examined the relationship between gender diversity and insolvency risk and found indeed a negative relationship. Their results thereby confirm the upper echelons perspective arguing that the negative relationship is driven by the perspective that women are more risk-averse than men. Nevertheless, Berger et al. (2014) report a positive relationship between female representation on the board of directors and risk-taking, which is potentially driven by a lower age and less experience of the female directors. Lastly, Mateos de Cabo et al. (2012) investigated the opposite relationship and concluded that riskier banks are more likely to hire male directors, thereby, limiting board diversity. This also illustrates the possible endogeneity problem, which will be discussed in more detail in the methodology section.
Based on the literature and the empirical evidence, the relationship between gender diversity and risk-taking can either be positive or negative in which the sign of the relationship depends on the dominating channel. Because of this ambiguous relationship, the following two alternative hypotheses will be tested:
H1a: There is a negative relationship between gender diversity on the board of directors
and bank risk-taking.
H1b: There is a positive relationship between gender diversity on the board of directors
and bank risk-taking.
2.4 The moderating effect of corporate governance
2.4.1 Country-level investor protection
11
can serve as a supplementary monitoring instrument and, thereby, reduce potential agency conflicts. Hereby it is argued that the effectiveness of board composition as a monitoring mechanism is argued to depend on the level of investor protection of a country. This section, therefore, investigates the moderating effect of investor protection on the relationship between gender diversity and bank risk-taking.
Recent literature has investigated the possibilities of both a positive and a negative relationship between investor protection and risk-taking. One argument in favor of a negative relationship states that when investor protection increases and shareholders are better protected, the fear of expropriation by managers decreases. The benefits of monitoring by large shareholders thus decreases. This leads to fewer controlling shareholders, which in turn, provides more leeway to managers to reduce their risk-taking (John et al., 2008). Nevertheless, most studies (e.g. Anginer, Demirguc-Kunt, Huizinga and Ma, 2014; John, Litov and Yeung, 2008) report a positive relationship between investor protection and risk-taking and this argument is based on agency theory. In this way, it is argued that enhanced investor protection can complement the board of directors by serving as an additional monitoring mechanism that reduces potential agency conflicts between shareholders and managers (García-Meca et al., 2015). Based on agency theory, managers prefer to take lower risks than preferred by their respective shareholders as to protect their private benefits, however, the degree of this misalignment is argued to depend on the degree of investor protection (John et al., 2008). John et al. (2008) argue that better investor protection leads to a better alignment of interests between managers and shareholders, and thereby, leads to an increase in risk-taking as preferred by the shareholders. Another explanation for a positive relationship between investor protection and risk-taking is based on the greater influence of external stakeholders, such as the government, in weak investor protection countries (John et al., 2008). As explained before, governments prefer banks to take fewer risks because high risk-taking can lead to failures of which the costs have to be borne by the entire economy. In this case, this means that in weak investor protection countries, imposed regulation by the government has a greater influence on the behavior of the board of directors and thereby stimulates incentives to reduce risk-taking.
12
directors to optimize their decision-making as they cannot easily be held accountable for possible negative effects that may arise as a consequence. This reduces the likelihood that directors promote and acknowledge the benefits of gender diversity on the decision-making process (Post and Byron, 2015). On the other hand, in high investor protection countries, shareholders benefit from better protection and can more easily discharge responsibilities or even replace directors who do not consider the unprecedented benefits of board diversity. This fosters the effects of board gender diversity on the organization’s strategy and risk-taking. It is, therefore, expected that the relationship between board gender diversity and risk-taking is stronger in high investor protection countries as this protection stimulates directors to optimize their decision-making. The second hypothesis is thus formulated as follows:
H2: The effect of board gender diversity on risk-taking is higher for banks located in
countries characterized by high investor protection and lower for banks located in countries characterized by low investor protection.
2.4.2 Bank-level corporate governance
13
In their research on the relationship between shareholder rights, both on country- and firm-level, and firms’ risk-taking, John, Litov and Yeung (2008) found that higher firm-level corporate governance was associated with higher risk-taking. This implies that better corporate governance structures lead to decision-making that better incorporates the interests of the shareholders who prefer relatively high risk-taking. These results are confirmed by Anginer et al. (2014) who found that banks with higher corporate governance scores were characterized by a higher insolvency risk. Furthermore, in their research on gender diversity, corporate governance and firm performance, Adams and Ferreira (2009) concluded that gender diversity on the board of directors can benefit companies with weak shareholder rights. They argue that in the case of weak corporate governance, gender diversity can function as a substitute for corporate governance through its monitoring capability. However, they also state that additional monitoring through enhanced gender diversity can be harmful in the presence of strong shareholder rights (Adams and Ferreira, 2009).
Therefore, the degree of bank-level corporate governance can be considered crucial in with respect to the amplitude of the effect of board gender diversity on the bank’s risk taking. Based on Adams and Ferreira (2009), it is hypothesized that gender diversity can act as a substitute for weak bank-level corporate governance structures, which leads to the following hypothesis:
14
3. Methodology
This section starts by identifying the endogeneity issue when examining the relationship between board gender diversity and the risk-taking of banks. This relationship can be endogenous because of potential heterogeneity and reverse causality (Sila et al., 2016). Subsequently, this section describes the variables and the empirical model.
3.1 Endogeneity problems
The widely accepted assumption in the literature is that the characteristics of the board of directors are not exogenous, but instead they are endogenous. When not sufficiently accounting for endogeneity, this can have a significant impact on the results. Roberts and Whited (2013) argued that “[e]ndogeneity leads to biased and inconsistent parameter estimates that make reliable inference virtually impossible” (p. 494). So, one of the major consequences of endogeneity is that the OLS estimator provides biased and inconsistent results. There are two forms of endogeneity in the relationship between gender diversity and a bank’s risk-taking: omitted variables and reverse causality.
3.1.1 Omitted variables
Nearly all corporate decisions are partly based on variables that are heterogeneous and cannot be directly observed (Roberts and Whited, 2013). In this research, there may be unobservable, omitted bank-level variables, such as corporate social responsibility (CSR) incentives, that affect both board gender diversity and risk-taking. In the case of CSR, a bank aims to improve its relationships with its stakeholders. As a result, investors might perceive the bank as less risky and they will act more positively to bank-specific developments. Furthermore, CSR can also reduce the systemic risk as it decreases a bank’s sensitivity to movements in the economy due to an increase in the loyalty of a bank’s customers. So, based on this stakeholder theory, a more socially responsible bank can be associated with reduced risk-taking behavior (Sila et al., 2016). Simultaneously, there can also exist a relationship between CSR and gender diversity as banks that are engaged in CSR may be more inclined to appoint female directors. Since CSR is fairly difficult to measure and there is no widely accepted proxy for this variable, it is often omitted from the empirical regression (Sila et al., 2016).
15
may be better able to manage risks and may have more influence on the appointment of fellow directors. Since studies show that women generally are better monitors, a powerful or highly competent director may prefer to appoint a male director. This shows that managerial competences can both be correlated with board gender diversity and the bank’s risk-taking (Sila et al., 2016). Similarly to CSR, managerial competences are also unobservable.
One way to deal with omitted variables is to use a fixed or random effects estimator. An entity-fixed effects estimator allows the intercept to differ cross-sectionally, while a time-fixed effects estimator allows the intercept to differ over time (Brooks, 2014). However, the fixed-effects estimator can prove to be inadequate in the presence of reverse causality, which will be discussed subsequently.
3.1.2 Reverse causality
A second problem when studying the effect of gender diversity on risk-taking is that it is difficult to determine the direction of the relationship (Adams et al., 2015). In this way, gender diversity may not only affect the risk-taking of the bank, but risk-taking may, in turn, also affect decisions regarding the appointment of new directors. Empirical evidence is provided by Mateos de Cabo et al. (2012) who found that riskier banks are more likely to appoint male directors. This is also referred to as the reverse causality or the simultaneity problem (Roberts and Whited, 2013).
16
the DID technique is presented in figure A1. Matching can be done by both calculating a propensity score and by nearest neighbor matching.
3.2 Variables
This section elaborates on the dependent variables, the independent variables and the control variables.
3.2.1 Dependent variable
The dependent variable comprises the level of the bank’s risk-taking. Similarly to the research by Berger et al. (2014), risk is defined as the bank’s portfolio risk. Portfolio risk is measured as the bank’s risk-weighted assets to its total assets and this measure is widely used by regulators since both Basel II and Basel III incorporate this measure to determine a bank’s capital requirements (De Haan, Oosterloo and Schoenmaker, 2015). Risk-weighted assets are generally calculated by using a standardized approach in which all assets, such as government bonds and residential mortgages, are appointed a certain risk weight of which the specific risk weights are outlined by the Basel II framework. The idea behind risk-weighted assets is that when a bank holds assets that are relatively more risky, more capital is needed to absorb potential losses (De Haan et al., 2015). Portfolio risk is thus in an important indicator of a bank’s risk.
17
3.2.2 Independent variables
The first independent variable in this research is gender diversity (GENDER) on the board of directors. This paper relies on the common measure for gender diversity in the literature, which is the percentage of female board members (Adams and Ferreira, 2009; Berger et al., 2014; García-Meca et al., 2015).
In the second section, the moderating effect of country-level investor protection (IP) is tested. Investor protection is defined as the country-level legal protection of investors against the expropriation by managers and is based on the anti-self-dealing index by Djankov, La Porta, Lopez-de-Silanes and Shleifer (2008). This anti-self-dealing index contains data for 72 countries and is based on the anti-directors rights index presented in an earlier work by La Porta, Lopez-de-Silanes, Shleifer and Vishny (2000). The former anti-directors rights index has been criticized and, therefore, the authors have introduced a new index that addresses these concerns: the anti-self-dealing index (Djankov et al., 2008). In this way, the anti-self-dealing index is based on a questionnaire that measures the obstacles shareholders face when entering a transaction. The index can take a value between zero and one, in which a higher score represents an environment characterized by higher investor protection. An advantage of this index is that it is based on theoretically grounded dimensions of investor protection. Nevertheless, the main disadvantage is that the index is based on data from 2003 and does not incorporate changes in investor protection over time.
Although country-level investor protection is expected to present a moderating effect, a larger effect is expected by bank-level corporate governance (CORPGOV). The corporate governance score is derived from the Datastream Asset4 database and is defined as the overall corporate governance score. This measure is based on four pillars: the function of the board of directors, the board structure, the board’s compensation policy, and the company’s vision and strategy. Hereby, the corporate governance score measures the bank’s systems and operations that ensure an alignment of interest between the board of directors and its shareholders. Furthermore, it measures the bank’s ability to create value for its shareholders. The corporate governance score can take a value between zero and one, in which a higher score represents a better corporate governance framework (Thomson Reuters, 2016).
3.2.3 Control variables
18
to reduce the chance that the results are driven by omitted variables. These control variables are categorized into bank-level and country-level control variables.
With respect to the bank-level control variables, the first control variable is bank size (BANKSIZE), which is measured by the natural logarithm of the total assets. A positive relationship between bank size and a bank’s risk-taking is expected because larger banks are generally expected to have a higher risk-absorbing capacity (Anderson and Fraser, 2000). Moreover, some banks are considered too-big-to-fail by their respective governments, which means that the respective government will step in to prevent the bank from failing, and this can, in turn, induce risk-taking (Berger et al., 2014). Nevertheless, a negative relationship between size and risk is also possible as Bruno and Shin (2014) argue that smaller organizations are more prone to take risks. Secondly, the bank’s capital adequacy ratio (CAR), which is measured by Tier 1 plus Tier 2 capital divided by the risk-weighted assets, enters as a control variable. Morrison and White (2005) argue that an increase in the bank’s capital ratio reduces moral hazard and fosters monitoring, which means that a negative relationship between the capital adequacy ratio and risk is expected. Furthermore, Keeley (1990) demonstrates a positive relationship between a bank’s charter value and risk-taking. The proxy for the bank’s charter value (CV) is the ratio of customer deposits to total assets. The fourth control variable comprises the ratio of customer loans to total assets (CLTA) because a higher ratio of loans is expected to increase portfolio risk (Berger et al., 2014). Additionally, the ratio of off-balance sheet items to total assets (OBSITEMS) is used to account for differences in the banks’ balance sheet composition. Off-balance sheet items can serve as a way to move risky items from the bank’s balance sheet, which means that a higher ratio of off-balance sheet items corresponds to higher risk-taking (De Haan et al., 2015). Fifthly, board size (BOARDSIZE) is also expected to influence risk-taking as Pathan (2009) and Cheng (2008) find that smaller boards lead to increased risk-taking. The last bank-level control variable is asset growth (ASSETGROWTH), because banks may adopt a different attitude towards risk-taking during times of accelerated asset growth (Berger et al., 2014).
19
3.3 Empirical model
3.3.1 Gender diversity and banks’ risk-taking
The aim of this paper is to identify the relationship between board gender diversity and banks’ risk-taking. The gathered data is a mixture of both time series and cross-sectional data and this panel data structure allows one to take into account characteristics of each specific bank that might drive the results. The dependent variable in the model is the risk (RISK), which is both defined as the risk-weighted assets over the total assets and the inversed Z-score for each bank
i at time t. The dependent variable is the gender diversity on the board (GENDER). The
bank-level control variables include the natural logarithm of the total assets (BANKSIZE), the capital adequacy ratio (CAR), the bank’s charter value (CV), the customer loans over total assets (CLTA), the off-balance sheet items over total assets (OBSITEMS), board size (BOARDSIZE) and total asset growth (ASSETGROWTH). Furthermore, real GDP growth (GDPGROWTH) and country dummies (COUNTRYFE) are included in the regression to account for country differences. The basic OLS regression model includes time fixed effects and can be formulated as follows:
𝑅𝐼𝑆𝐾𝑖𝑡= 𝛼 + 𝛽1𝐺𝐸𝑁𝐷𝐸𝑅𝑖𝑡+ 𝛽2𝐵𝐴𝑁𝐾𝑆𝐼𝑍𝐸𝑖𝑡+ 𝛽3𝐶𝐴𝑅𝑖𝑡+ 𝛽4𝐶𝑉𝑖𝑡+
𝛽5𝐶𝐿𝑇𝐴𝑖𝑡+ 𝛽6𝑂𝐵𝑆𝐼𝑇𝐸𝑀𝑆𝑖𝑡+ 𝛽7𝐵𝑂𝐴𝑅𝐷𝑆𝐼𝑍𝐸𝑖𝑡+ 𝛽8𝐴𝑆𝑆𝐸𝑇𝐺𝑅𝑂𝑊𝑇𝐻𝑖𝑡
+𝛽9𝐺𝐷𝑃𝐺𝑅𝑂𝑊𝑇𝐻𝑡+ 𝐶𝑂𝑈𝑁𝑇𝑅𝑌𝐹𝐸 + 𝜆𝑡+ 𝜀𝑖𝑡 (1)
20
techniques, this allows one to compare the results of this research with previous papers on this topic (Andres and Vallelado, 2008). The following empirical model, including cross-section fixed effects, is constructed:
𝑅𝐼𝑆𝐾𝑖𝑡= 𝛼 + 𝛽1𝐺𝐸𝑁𝐷𝐸𝑅𝑖𝑡+ 𝛽2𝐵𝐴𝑁𝐾𝑆𝐼𝑍𝐸𝑖𝑡+ 𝛽3𝐶𝐴𝑅𝑖𝑡+ 𝛽4𝐶𝑉𝑖𝑡+
𝛽5𝐶𝐿𝑇𝐴𝑖𝑡+ 𝛽6𝑂𝐵𝑆𝐼𝑇𝐸𝑀𝑆𝑖𝑡+ 𝛽7𝐵𝑂𝐴𝑅𝐷𝑆𝐼𝑍𝐸𝑖𝑡+ 𝛽8𝐴𝑆𝑆𝐸𝑇𝐺𝑅𝑂𝑊𝑇𝐻𝑖𝑡
+𝛽9𝐺𝐷𝑃𝐺𝑅𝑂𝑊𝑇𝐻𝑡+ 𝜇𝑖+ 𝜆𝑡+ 𝜀𝑖𝑡 (2)
Nevertheless, even the fixed effects model can fail to adequately deal with the endogeneity issue. Some of the critiques to the fixed effects estimation in the case of bank’s risk-taking presented by Pathan (2009) include that the fixed effects model is not able to capture changes in time-invariant variables, it requires significant variations in the independent variables and it may exacerbate the multicollinearity problem if least square dummy variables are used. Therefore, Berger et al. (2014) and Sila et al. (2016) propose a matching method combined with a difference-in-difference (DID) estimator. This method can account for the endogeneity issue and is estimated by the following equation:
𝑅𝐼𝑆𝐾𝑖𝑡= 𝛼 + 𝛽1𝐵𝑂𝐴𝑅𝐷𝐶𝐻𝐴𝑁𝐺𝐸 + 𝛽2𝑃𝑂𝑆𝑇𝑖𝑡+ 𝛽3𝑃𝑂𝑆𝑇𝑖𝑡∗ 𝐵𝑂𝐴𝑅𝐷𝐶𝐻𝐴𝑁𝐺𝐸𝑖𝑡
+ 𝛽4𝑋𝑖𝑡+ 𝜇𝑖+ 𝜀𝑖𝑡 (3)
Or in differences:
∆𝑅𝐼𝑆𝐾𝑡= 𝛼 + 𝛽1𝐵𝑂𝐴𝑅𝐷𝐶𝐻𝐴𝑁𝐺𝐸 + ∆𝛽4𝑋𝑡+ ∆𝜀𝑡 (4)
Here the BOARDCHANGE variable is a dummy variable that defines the treatment and the control group and takes the value of one if a female is appointed to a board position and zero otherwise. The POST period variable is a dummy variable that takes the value of one after the board change. Furthermore, Xit is a vector of the control variables and μi indicate the
cross-section fixed effects. This method requires board size to be constant.
21
in the bank’s risk-taking is calculated over a 5-year window. The reason to reduce the 7-year window to a 5-year period is because of missing values of which the majority of the missing values was found either in t-3 or in t+3. In order to prevent distortions from these missing values, the observed period is reduced to a 5-year period. Thus, the pre-board change period comprises the two years prior to the board change, following by the actual board change at time
t. This risk is compared to the risk-taking in the post-board change period, which comprises the
two years following the board change.
3.3.2 Corporate governance, gender diversity and risk-taking
With respect to country-level investor protection, it is hypothesized that there exists a stronger effect between gender diversity and risk-taking in banks located in strong investor protection countries and vice versa. In order test for this effect and by following García-Meca et al. (2015), both the investor protection variable (IP) and the interaction effect between gender diversity and investor protection are added to the equation. Since the investor protection variable is fixed over time, it is not possible to estimate the model using cross-section fixed effects. The model is, therefore, tested using OLS with time-fixed effects and country dummies. The following OLS regression model is tested:
𝑅𝐼𝑆𝐾𝑖𝑡= 𝛼 + 𝛽1𝐺𝐸𝑁𝐷𝐸𝑅𝑖𝑡+ 𝛽2𝐼𝑃𝑖+ 𝛽3𝐺𝐸𝑁𝐷𝐸𝑅𝑖𝑡∗ 𝐼𝑃𝑖+ 𝛽4𝐵𝐴𝑁𝐾𝑆𝐼𝑍𝐸𝑖𝑡
+𝛽5𝐶𝐴𝑅𝑖𝑡+ 𝛽6𝐶𝑉𝑖𝑡+ 𝛽7𝐶𝐿𝑇𝐴𝑖𝑡+ 𝛽8𝑂𝐵𝑆𝐼𝑇𝐸𝑀𝑆𝑖𝑡+ 𝛽9𝐵𝑂𝐴𝑅𝐷𝑆𝐼𝑍𝐸𝑖𝑡
+𝛽10𝐴𝑆𝑆𝐸𝑇𝐺𝑅𝑂𝑊𝑇𝐻𝑖𝑡+ 𝛽11𝐺𝐷𝑃𝐺𝑅𝑂𝑊𝑇𝐻𝑡+ 𝐶𝑂𝑈𝑁𝑇𝑅𝑌𝐹𝐸 + 𝜆𝑡+ 𝜀𝑖𝑡 (5)
With respect to bank-level corporate governance, it is hypothesized that there exists a stronger effect between gender diversity and risk taking in banks with weak corporate governance levels and vice versa. Similarly to the research by Adams and Ferreira (2009), both the corporate governance variable (CORPGOV) and the interaction effect between gender diversity and corporate governance are added to the equation. This model is solely tested by using fixed effects to account for the omitted variables bias:
𝑅𝐼𝑆𝐾𝑖𝑡= 𝛼 + 𝛽1𝐺𝐸𝑁𝐷𝐸𝑅𝑖𝑡+ 𝛽2𝐶𝑂𝑅𝑃𝐺𝑂𝑉𝑖𝑡+ 𝛽3𝐺𝐸𝑁𝐷𝐸𝑅𝑖𝑡∗ 𝐶𝑂𝑅𝑃𝐺𝑂𝑉𝑖𝑡 + 𝛽4𝐵𝐴𝑁𝐾𝑆𝐼𝑍𝐸𝑖𝑡
+𝛽5𝐶𝐴𝑅𝑖𝑡+ 𝛽6𝐶𝑉𝑖𝑡+ 𝛽7𝐶𝐿𝑇𝐴𝑖𝑡+ 𝛽8𝑂𝐵𝑆𝐼𝑇𝐸𝑀𝑆𝑖𝑡 +𝛽9𝐵𝑂𝐴𝑅𝐷𝑆𝐼𝑍𝐸𝑖𝑡+ 𝛽10𝐴𝑆𝑆𝐸𝑇𝐺𝑅𝑂𝑊𝑇𝐻𝑖𝑡
22
4. Data
4.1 Data and sample selection
23
per cent of the sample. Nevertheless, since the regressions include country dummies, this should not lead to major deviations.
Data on all bank-level variables was collected through Datastream after which missing values were complemented by Bankscope in case the known values were similar. Furthermore, data on board-level variables, including gender diversity and board size was collected through the Datastream Asset4 Environmental, Social and Governance (ESG) Universe. Lastly, data on GDP growth was collected through the Worldbank database.
Table 1
This table represents the distribution of banks across the different countries. The full sample consists of 95 banks located in the European Union. All banks are publicly listed.
Country Banks Percentage
AT 3 3.16 BE 3 3.16 CY 1 1.05 CZ 1 1.05 DE 11 11.58 DK 3 3.16 ES 7 7.37 FI 1 1.05 FR 4 4.21 GB 22 23.16 GR 6 6.32 HU 1 1.05 IE 2 2.11 IT 13 13.68 MT 1 1.05 NL 2 2.11 PL 8 8.42 PT 2 2.11 SE 4 4.21 All 95 100
4.2 Descriptive statistics
24
size is 17.213, which is relatively high due to a domination of one-tier board structures in this sample. The descriptive statistics for the full sample per year can be found in table A31.
Table 2
This table reports the mean, median, maximum, minimum, standard deviation and the number of observations for the two dependent variables RWA/TA and ZSCORE, the independent variables GENDER and IP, and the control variables for the full sample. The sample comprises 305 observations of 95 banks between 2002 and 2014. A description of the variables is presented in Table A1.
Variable Mean Median Maximum Minimum Std. Dev. Obs.
Executive board characteristics
Gender diversity 0.128 0.100 0.600 0.000 0.115 305
Board size 17.213 16.000 35.000 5.000 4.771 305
Firm characteristics
Total assets (log) 8.419 8.295 10.009 7.235 0.651 305
Total assets growth 0.100 0.049 4.315 -0.329 0.322 305
Capital adequacy ratio 0.130 0.127 0.247 -0.057 0.031 305
Customer loans / Total assets 0.119 0.075 0.729 0.003 0.131 305
Charter value 0.445 0.442 0.820 0.055 0.153 305
Off-balance sheet items / Total assets 0.270 0.221 1.884 0.000 0.208 305
ROE -0.165 0.075 0.981 -42.985 2.642 305
Corporate governance score 0.749 0.794 0.954 0.113 0.158 161
Country characteristics
GDP growth -0.169 0.455 5.336 -8.998 2.929 305
Investor protection 0.437 0.382 0.927 0.204 0.213 305
Risk measures
Risk-weighted assets / Total assets 0.513 0.524 0.894 0.124 0.195 305
Inversed Z-score 0.111 0.082 2.066 -1.160 0.195 286
Multicollinearity can be present when studying board characteristics and occurs when board variables are correlated to each other. In order to test for multicollinearity, the correlation coefficients are presented in table 42. The highest correlations can be found between charter value and total assets (-0.566) and between total assets and gender (0.503). The negative relationship between charter value and bank size is explained by De Nicoló (2001). In this case the interpretation is that larger banks are characterized by higher total assets, which can be found in the denominator of charter value. The negative relationship indicates that for larger banks, the customer deposits do not increase by the same rate as their total assets. The second coefficient indicates that larger banks are characterized by a higher female representation on the board of directors. This is in line with the findings by García-Meca et al. (2015) and Sila et
1 The country-level investor protection variable is not included in this table as this is a time-invariant variable. 2 The lagged variable of gender diversity is not included in this table as this will show similar correlations to the
25
al. (2016) who found a positive correlation between firm size and gender diversity. Moreover, Adams and Ferreira (2003) argue that larger firms are characterized by increased gender diversity because these firms are more likely to be put in the spotlight or women are attracted to larger firms and thus more likely to be appointed to a board position.
27
5. Empirical results
In order to investigate whether board gender diversity affects banks’ risk-taking, this research examines whether the risk-taking behavior of banks that have more diverse boards differs from banks that have less diverse boards in terms of gender diversity. Furthermore, the moderating effect of country-level investor protection and bank-level corporate governance is studied.
5.1 Gender diversity and risk-taking
5.1.1 OLS and fixed effects estimations
The first step in identifying the relationship between female representation on executive boards and bank risk-taking is to use an OLS and a fixed effects estimator. Since this relationship is likely to be subject to endogeneity, both methods would produce biased estimates in the presence of omitted variables and/or reverse causality. Nevertheless, these estimators are included because they can provide valuable insights into the causes of endogeneity in the relationship. First of all, if the relationship between gender diversity and risk-taking is driven by factors that are included in the regression, the OLS estimator would produce significant results once these factors are included. However, if unobservable or omitted variables are driving the relationship, these elements should be captured by the fixed effects estimator (Sila et al., 2016).
28
Table 4
This table presents the results of the OLS and the fixed effects estimations on the relationship between gender diversity and risk taking. All models include year dummy variables and the OLS estimators include country dummies. Robust t-statistics are presented in parentheses. *, ** and *** denote the statistical significance at the 10%. 5% and 1% level respectively.
RWA / TA Inversed Z-score
Variable OLS Fixed effects OLS Fixed effects
Gender diversity 0.142** 0.124** 0.296* 0.474**
(2.17) (2.35) (1.82) (2.33)
Total assets (log) -0.074*** -0.085 -0.022 0.534**
(-5.11) (-1.55) (-0.60) (2.55)
Capital adequacy ratio -0.340* -0.633*** 1.324*** 2.615***
(-1.67) (-4.18) (2.66) (4.52)
Charter value 0.240*** -0.020 -0.053 0.820**
(4.74) (-0.24) (-0.42) (2.52)
Customer loans / Total assets -0.114** 0.057 -0.030 0.088
(-2.05) (1.30) (-0.22) (0.52)
Off-balance sheet items / Total assets 0.112*** 0.034 0.017 0.094
(4.27) (1.52) (0.27) (1.11)
Total asset growth -0.009 -0.000 -0.006 -0.024
(-0.54) (-0.01) (-0.16) (-0.54) Board size -0.001 0.002 0.000 -0.001 (-0.34) (1.46) (0.06) (-0.21) GDP growth -0.002* 0.000 0.031*** 0.019** (-0.67) (0.13) (3.97) (2.29) Constant 1.025*** 1.245*** 0.023 -5.153*** (6.95) (2.65) (0.063) (-2.86)
Period-fixed effects Yes Yes Yes Yes
Cross-section fixed effects No Yes No Yes
R2 0.837 0.946 0.226 0.391
Adjusted R2 0.816 0.930 0.122 0.204
Observations 305 305 286 286
29
which is much larger in magnitude. Here, the results show that a 1 percent increase in gender diversity corresponds to a 0.474 percent increase in risk-taking. These results support hypothesis 1b. Moreover, the results are line with the findings by Berger et al. (2014) who also found evidence of a positive relationship between gender diversity and risk-taking. Agency theory argues that increased gender diversity on the board of directors improves the bank’s monitoring capabilities. More specifically, the regression results suggest that the board enhances its monitoring in favor of the shareholders who prefer higher risk-taking. Furthermore, the regression results as presented in this paper do not control for director-specific characteristics such as experience and age. As shown by Berger et al. (2014), newly appointed female directors are often younger and have less experience. It should, therefore, be acknowledged that these two characteristics may potentially drive the positive relationship between gender diversity and bank risk-taking as shown in the regression results.
Turning to the control variables, the effect of bank size is inconclusive. Based on the OLS results, bank size is negatively related to portfolio risk, while the fixed effects estimator shows that bank size is positively related to insolvency risk. The capital adequacy ratio is negatively related to risk, which is in line with the argument that an increase in capital reduces the moral hazard. Furthermore, the significant coefficients of charter value indicate a positive relationship with risk taking, which is similar to the results by Keeley (1990). Fourthly, the control variable customer loans to total assets is only significant in the first model and reports a negative relationship. This contradicts the expectations as a higher ratio of customer loans was expected to increase a bank’s risk. With respect to the significant coefficient, off-balance sheet items reports the expected positive sign indicating that an increase in off-balance sheet items corresponds to an increase in risk-taking. The coefficients on total asset growth and board size are insignificant and it is, therefore, not possible to draw conclusions from this. Lastly, GDP growth is positively related to portfolio risk, whereas it presents a negative relationship with respect to insolvency risk. This outcome can be explained as follows: during upturns in the business cycle banks are inclined to take more risk, which is reflected in their portfolio risk, while at the same time their assets increase in value, which is reflected by a decline in insolvency risk (De Haan et al., 2015).
5.1.2 Robustness check
one-30
year lagged gender diversity and board characteristics variables as proposed by Liu et al. (2014). The idea behind this is that directors need time before they are able to influence an organization’s strategy and risk-taking. Additionally, by using lagged variables, the possibility that the bank’s risk-taking affects the appointment of female directors, and thereby the reverse causality problem, is minimized (Liu et al., 2014). Thus, as a robustness check, the previous model is estimated again, but now including a one-year lag of both gender diversity and board size. The results of the robustness check are reported in table A6. This estimation method does not change the sign on the coefficient of gender diversity with respect to portfolio risk and the R2 is only somewhat smaller in comparison to the results of the original model. This suggests that the relationship between gender diversity and portfolio risk is not driven by reverse causality. With respect to the effect on the inversed Z-score, the results of the robustness check are insignificant. It is, therefore, not possible to draw conclusions from this.
5.1.3 Difference-in-difference matching estimations
31
Figure 1
This figure depicts the board characteristics, including the average board size, the number of women and the percentage of women on the board of directors, of the full sample per year.
In the DID estimator, banks from the treatment group are compared to banks from the control group. Both groups comprise board changes that did not alter the total board size. The treatment group is defined as banks that experienced an increase in female representation on the board of directors, while the control group consists of board changes in which gender diversity either remained constant or decreased. The descriptive statistics of both groups are presented in table A7. The banks are matched on the year of the board change, the size of the bank measured as the natural logarithm of total assets and return on equity (ROE). Furthermore, the control variables as presented in table 2 are included in the DID estimator. This research both employs propensity score matching and the nearest neighbor matching technique. In propensity score matching, a propensity score is calculated for every observation, after which observations from the treatment group are matched to observations from the control group. This involves a propensity score restriction, which defines the range in which observations can be matched (Sila et al., 2016). In this study, the propensity score restriction was set at 10 percent due to the relative small sample. Secondly, nearest neighbor matching involves matching an observation from the treatment group to an observation from the control group that is most similar. Here the number of matches per observation should be determined, which involves a trade-off between bias and variance (Roberts and Whited, 2013). While a small number of matches leads to the least biased results, it is also accompanied by the largest variance. Due to the small sample, the number of matches for each treatment observation was set at 1 in this study.
0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 % o f w o m e n Bo ar d m e m ber s Year
Board characteristics over time
32
The results of the DID matching techniques are presented in table 5. The results show a negative coefficient on the relationship between gender diversity and portfolio risk. For the propensity score matching technique, this coefficient is significant at the 10 percent level. This result suggests that a 1 percent increase in the percentage of female directors corresponds to a decrease of 0.034 percent in the bank’s portfolio risk. This result weakly support hypothesis Ia and contradicts the previous findings. This results is in line with the findings by Muller-Kahle and Lewellyn (2011) who reported a negative relationship between gender diversity and subprime lending. In this respect, gender diversity on the board of directors could benefit banks that aim to reduce their portfolio risk. When examining the effect of gender diversity on the bank’s insolvency risk, the results show a positive relationship, however these are not significant in the DID estimations.
In sum, the DID estimator rules out the possibility that the results are driven by the reverse causality problem and is, thus, more likely to produce unbiased coefficients. Normally, the DID estimations would provide the most reliable results as, in this way, it is possible to extract the effect of gender diversity from other patterns that affect both groups in a similar way. Nevertheless, given the small sample of this study, the DID estimator may not be able to present the true relationship and the results should be interpreted with caution.
Table 5
This table shows the results from the average treatment effects (ATE) from the difference-in-difference estimations. The difference is calculated as the difference in risk between the pre- and the post-period. The pre-period is defined as the two years prior to the board change, while the post-period is defined as the 2 years following the board change. Treatment banks are matched with control banks both by propensity score matching and nearest neighbor matching. Matching variables are the control variables as defined in table A1 and additionally, the board change must occur in the same year. One match per observation is allowed. Robust t-statistics are presented in parentheses. *, ** and *** denote the statistical significance at the 10%, 5% and 1% level respectively.
RWA / TA Inversed Z-score
Propensity score
matching Nearest neighbor matching Propensity score matching Nearest neighbor matching
ATE Board change -0.034* -0.027 0.029 0.037
(-1.81) (-1.61) (1.02) (1.18)
Nr. of matches 24 35 27 46
5.2 The moderating effect of corporate governance
5.2.1 Country-level investor protection
33
Table 6
This table shows the results of the OLS estimator regarding the moderating effect of country-level investor
protection on the relationship between gender diversity and risk-taking. All estimations include year dummy variables and country dummies. Robust t-statistics are presented in parentheses. *, ** and *** denote the statistical
significance at the 10%, 5% and 1% level respectively.
Variable RWA / TA Inversed Z-score
Gender diversity 0.142** -0.183 0.296* 0.435
(2.17) (-1.33) (1.82) (1.26)
Investor protection 0.441** 0.331* 1.107* 1.152**
(2.24) (1.66) (2.31) (2.35)
Gender diversity * Investor protection 0.864*** -0.369
(2.69) (-0.46)
Total assets (log) -0.074*** -0.076*** -0.022 -0.021
(-5.11) (5.30) (-0.60) (0.58)
Capital adequacy ratio -0.340* -0.323 1.323*** 1.318***
(-1.67) (-1.61) (2.66) (2.64)
Charter value 0.240*** 0.228*** -0.053 -0.046
(4.74) (4.52) (-0.42) (-0.36)
Customer loans / Total assets -0.114* -0.105* -0.030 -0.036
(-2.05) (1.90) (-0.22) (-0.26)
Off-balance sheet items / Total assets 0.112*** 0.111*** 0.018 0.018
(4.27) (4.28) (0.27) (0.28)
Total asset growth -0.009 -0.008 -0.006 -0.007
(-0.54) (-0.46) (-0.16) (-0.17) Board size -0.001 -0.001 0.000 0.000 (-0.34) (-0.33) (0.06) (0.07) GDP growth -0.002 -0.003 0.031*** 0.032*** (-0.67) (-0.90) (3.97) (3.98) Constant 0.981*** 1.050*** -0.375 -0.404 (5.94) (6.35) (-0.92) (-0.98)
Period fixed effects Yes Yes Yes Yes
Cross-section fixed effects No No No No
R2 0.837 0.841 0.226 0.227
Adjusted R2 0.816 0.821 0.122 0.119
Observations 305 305 286 286
34
benefits of diversity on the decision-making process, which is in line with the arguments of resource dependence theory (Post and Byron, 2015). The third and the fourth column present the results with respect to insolvency risk. Both the coefficient of gender diversity and investor protection are positive, which is in line with the effect on portfolio risk. In the fourth column, the interaction effect is added, however, this coefficient is insignificant. It is, therefore, not possible to draw conclusions from these results.
One has to take into consideration that since the investor protection measure does not vary over time, it is not possible to use a fixed effects estimator that takes into account inherent firm-level variations. Caution is, therefore, desired when drawing conclusions from these results. In order to account for the omitted variables bias, the subsequent section examines the moderating effect of firm-level corporate governance on the relationship between gender diversity and risk-taking by applying a fixed effects estimator.
5.2.2 Bank-level corporate governance
35
gender diversity is likely to act as a substitute for weak corporate governance levels as a result of enhanced monitoring. In contrast, the results suggest that the effect of gender diversity on risk-taking is stronger for banks with high corporate governance levels. Similarly to the moderating effect of investor protection, an explanation for the positive interaction effect can be that in banks with high corporate governance levels, the benefits of diversity on the decision-making process are better acknowledged (Post and Byron, 2015).
Table 7
This table presents the results of the fixed effects estimators regarding the moderating effect of bank-level corporate governance on the relationship between gender diversity and risk-taking. All models include year dummy variables. Robust t-statistics are presented in parentheses. *, ** and *** denote the statistical significance at the 10%, 5% and 1% level respectively.
RWA / TA Inversed Z-score
Variable
Gender diversity 0.085 0.087 -0.001 -0.001
(1.42) (1.42) (-0.04) (-0.04)
Corporate governance -0.061 -0.057 0.014 0.004
(-1.32) (-1.12) (0.64) (0.20)
Gender diversity * Corporate governance -0.069 0.244*
(-0.24) (1.72)
Total assets (log) -0.176** -0.171** -0.058* -0.068**
(-2.48) (-2.34) (-1.81) (-2.13)
Capital adequacy ratio 0.049 0.045 -0.119 -0.099
(0.13) (0.12) (-0.67) (-0.57)
Charter value (0.263** 0.262** -0.191*** -0.193***
(2.28) (2.26) (-3.09) (-3.17)
Customer loans / Total assets 0.012 0.018 -0.014 -0.030
(0.16) (0.22) (-0.33) (-0.69)
Off-balance sheet items / Total assets -0.028 -0.026 0.014 0.007
(-0.60) (-0.54) (0.65) (0.32)
Total asset growth 0.028* 0.028* 0.004 0.004
(1.84) (1.81) (0.62) (0.66) Board size 0.001 0.001 0.002* 0.002* (0.41) (0.36) (1.74) (2.00) GDP growth -0.002 -0.002 0.002 0.002 (-0.42) (0.36) (1.08) (0.93) Constant 1.829*** 1.793** 0.651** 0.740** (2.80) (2.66) (2.19) (2.50)
Period fixed effects Yes Yes Yes Yes
Cross-section fixed effects Yes Yes Yes Yes
R2 0.971 0.971 0.870 0.877
Adjusted R2 0.956 0.955 0.794 0.802
36
6. Conclusion
Over the past few years, gender diversity on the board of directors has become a hotly debated topic. The recent pressure by national governments on organizations to increase the number of women in the boardroom has only provided further stimulus to research the effect of gender diversity on an organization’s performance and risk-taking. Nevertheless, relatively few scholars have focused on the banking sector, while the financial crisis has dramatically exposed the negative consequences of excessive risk-taking behavior. This research has, therefore, sought to complement the existing literature by studying the effect of gender diversity in the boardroom on a bank’s risk-taking. Up until today, agency theory, resource dependence theory, upper echelons theory and recent studies have failed to demonstrate a clear relationship between female representation on the board and banks’ risk-taking. The scope of this study comprised a sample of listed European banks, which were examined over the period between 2002 and 2014. Bank risk-taking was both defined as portfolio risk and insolvency risk. The panel data structure allowed for several estimation techniques, including pooled OLS, fixed effects and DID. The results of the different estimators stress the importance of accounting for the potential endogeneity issue.
37
Since the DID estimator accounts for the endogeneity issue of board composition, this estimation technique would normally provide the most reliable results. However, due to the small sample, the DID results should be interpreted with caution.
The subsequent part of this research examined the moderating effect of country-level investor protection and showed that the effect of gender diversity on bank risk-taking increases with the level country-level investor protection. In practice, these results suggest that an increase in the percentage of female directors in a common law country such as the United Kingdom, would have a much larger impact on bank risk-taking than a similar increase in gender diversity in a civil law country. In the light of proposed quota, this conclusion is especially relevant to policymakers as they have to be aware of the potential consequences of the quota. In addition to the effect of country-level investor protection, the moderating effect of bank-level corporate governance was tested. These findings suggest that, similarly to the effect of investor protection, the effect of gender diversity on risk-taking increases with the corporate governance level. An explanation can be found in resource dependence theory, which argues that the advantages of gender diversity on the decision-making process are better acknowledged in environments that are characterized by a better alignment between the managers and the shareholders.
38
7. References
Adams, R.B., Ferreira, D., 2003. Diversity and incentives: evidence from corporate boards. Working paper.
Adams, R.B., Ragunathan, V., 2015. Lehman sisters. Unpublished Working Paper.
Adams, R.B., Ferreira, D., 2009. Women in the boardroom and their impact on governance and performance. Journal of Financial Economics 94(2), 291-309.
Adams, R.B., Funk, P., 2012. Beyond the glass ceiling: does gender matter? Management Science 58(2), 219-235.
Adams, R.B., Haan, J., Terjesen, S., Ees, H., 2015. Board diversity: moving the field forward. Corporate Governance: An International Review 23(2), 77-82.
Adams, R.B., Mehran, H., 2012. Bank board structure and performance: evidence for large bank holding companies. Journal of Financial Intermediation 21(2), 243-267.
Ahern, K.R., Dittmar, A.K., 2012. The changing of the boards: the impact on firm valuation of mandated female board representation. Quarterly Journal of Economics 127(1), 137-197.
Anderson, R.C., Fraser, D.R., 2000. Corporate control, bank risk taking, and the health of the banking industry. Journal of Banking & Finance 24(8), 1383-1398.
Andres, P., Romero-Merino, M.E., Santamaría, M., Vallelado, E., 2012. Board determinants in banking industry: an international perspective. Managerial and Decision Economics 33(3), 147-158.
Andres, P., Vallelado, E., 2008. Corporate governance in banking: the role of the board of directors. Journal of Banking and Finance 32(12), 2570-2580.
Anginer, D., Demirguc-Kunt, A., Huizinga, H., Ma, K., 2014. Corporate governance and bank insolvency risk: international evidence. Policy research working paper.
Basel Committee on Banking Supervision, 2010. Principles for enhancing corporate governance. http://www.bis.org/publ/bcbs176.pdf (accessed April 9, 2016). Basel Committee on Banking Supervision, 2006. Enhancing corporate governance for
banking organisations. http://www.bis.org/publ/bcbs122.pdf (accessed April 9, 2016). Berger, A.N., Kick, T., Schaeck, K., 2014. Executive board composition and bank risk taking.
Journal of Corporate Finance 28, 48-65.
