Board Gender Diversity and Firm Financial Performance: A Quantile Regression Analysis

International Corporate Governance and Regulation

ISBN: 978-1-78756-536-4, eISBN: 978-1-78756-535-7

ISSN: 1569-3732

Publication date: 1 November 2018

Abstract

Numerous empirical studies have been conducted to analyze the impact of board gender diversity (BGD) on firm performance without being able to establish a clear relationship. In this paper, we reassess the relationship between BGD and firm performance by using a quantile regression approach. Our results indicate that BGD matters only across a subset of the firm performance distribution. Moreover, when the possible endogeneity of the relationship between BGD and firm performance is taken into account, there are some conditions under which a positive and significant relationship is observed for the eight lowest quantiles.

Keywords

Citation

Charles, A., Dang, R. and Redor, E. (2018), "Board Gender Diversity and Firm Financial Performance: A Quantile Regression Analysis", International Corporate Governance and Regulation (Advances in Financial Economics, Vol. 20), Emerald Publishing Limited, pp. 15-55. https://doi.org/10.1108/S1569-373220180000020002

Download as .RIS

Publisher

:

Emerald Publishing Limited

Copyright © 2019 Emerald Publishing Limited


Introduction

Around the world and notwithstanding the manner in which avenues for change are made (from legislated quotas in Norway and France to an explosion of advocacy groups that have applied pressure) efforts calling for action to increase board gender diversity (hereinafter referred to as BGD) are approaching a tipping point (Terjesen, Sealy, & Singh, 2009). Previously considered as an ethical issue (that it is wrong for individuals to be excluded from the highest echelons of an organization purely on the grounds of gender), BGD is now perceived as a value driver: “BGD makes good business case” (Robinson & Dechant, 1997). This latter statement is known as the “business case for diversity,” that is to say that BGD increases firm performance by providing a competitive edge to the firm (Robinson & Dechant, 1997). This business case for diversity identifies four key organizational benefits: overcoming the skills shortage and war of talent (Powell, 1999); responding to emerging consumer markets (Daily, Certo, & Dalton, 1999); better employee performance (Kochan et al., 2003); and increasing a firm’s reputation (Brammer, Millington, & Pavelin, 2009). On the other hand, a more diverse board of directors might also induce more conflict and slower decision-making (Lau & Murnighan, 1998; Miller, Burke, & Glick, 1998; Williams & O’Reilly, 1998). These latter arguments are consistent with the literature on social choice (Arrow, 1951) which posits that costs associated with collective decision-making will be higher when the decision-makers are heterogeneous.

Generally speaking, in the literature, many issues remain unresolved from an empirical standpoint, since a large body of the research shows mixed results. Specifically, some studies report significantly positive relationships, others negative ones, while others insignificant ones. In the BGD literature, the current state of art is not very different regarding its impact on firm financial performance. This relationship is hotly debated in the economics, financial and organizational sciences literatures, at both the theoretical and empirical standpoints, with the evidence revealing mixed results (Dang & Vo, 2012). For example, Liu, Wei, and Xie (2014) and Campbell and Mınguez-Vera (2008) provide evidence that BGD is positively associated with firm performance. Conversely, Adams and Ferreira (2009) document a negative relationship between BGD and firm value. Finally, Francoeur, Labelle, and Sinclair-Desgagné (2008) and Rose (2007) find no evidence of a relationship between BGD and firm performance. This is a puzzling issue. Of course, there are many reasons that may explain such a mixed bag of findings, obvious candidates being the different samples or specifications used across studies.

We believe that such empirical inconsistency regarding the relationship between BGD and firm performance may, in part, be due to the dominant estimation method applied. The literature to date has estimated the above relationship using classical linear regression (ordinary least squares – OLS) or instrumental variables (IVs) (see Dang & Vo (2012) or Terjesen et al. (2009) for a review of the literature). While estimating how BGD “on average” impacts firm financial performance is both intuitive and useful, it seems possible that this relationship may differ across the firm’s conditional performance distribution. In other words, we argue that it is possible that the effect of BGD is different across the conditional distribution of firm performance. Furthermore, the existing literature suffer several methodological shortcomings. For example, problems with reverse causality and unobserved firm heterogeneity have typically not been addressed (e.g., Erhardt, Werbel, & Shrader, 2003; Rose, 2007). Typically, profitable companies might be more likely to appoint female directors to corporate boards, that is, high returns might lead to more BGD, rather than the opposite (Adams & Ferreira, 2009). Also, due to unobserved firm heterogeneity, firms with few female directors on corporate boards might have better performance if they had fewer female directors. It follows, therefore, that the effect of BGD on firm performance remains unresolved, despite an extensive body of literature.

This study makes several contributions to the corporate governance and diversity literature. First, this paper tackles the relationship between BGD and firm performance by using the quantile regression (hereinafter referred to as QR) approach developed by Koenker and Bassett (1978). A natural and relatively simple way to explore differences across the distribution of a firm’s performance is through the use of QR. By construction, a QR answers the question of what is the marginal effect of an explanatory variable at an arbitrary point in the conditional distribution of a dependent variable (Koenker, 2005). This ability to look across the distribution allows an examination of differences at different points of the conditional distribution. We, therefore, go beyond classical regression or its offspring. This approach has been previously used to examine, among others, economic issues (e.g., Mata & Machado, 1996), corporate governance (e.g., Kuan, Li, & Liu, 2012), or firm performance (e.g., Margaritis & Psillaki, 2010). Second, from an empirical standpoint, consistent with Adams and Ferreira (2009) and Liu et al. (2014), we take into account the possible endogeneity of the relationship between BGD and firm performance. Specifically, following Chernozhukov and Hansen (2005, 2006), we introduce a QR estimator that deals with the endogeneity issue. The underlying idea is to add a variable to the regression so that regressors and error terms become independent. To the best of our knowledge, this study is the first research using QR with endogeneity. Third, unlike existing studies, this paper uses a longitudinal study (firms belonging to the S&P 100 Index over the period 1995–2010). We believe that this approach is more powerful in controlling for unobservable heterogeneity. This aspect is often neglected or not taken into consideration by the existing literature. As highlighted by Chhaochharia and Grinstein (2007), one concern with an unbalanced panel of firms is that any effect of BGD might be attributed to the characteristics of firms moving in and out of an index such as the S&P or which have been purchased by other firms. Firms might also change their level of BGD because, if they are no longer part of an index, they are monitored less by stakeholders (Hillman, Shropshire, & Cannella, 2007). Consequently, our longitudinal approach should provide more reliable outcomes. Finally, this paper makes a theoretical contribution to the diversity and governance literature by providing a better understanding of how the relationship between BGD and firm performance operates. As previously mentioned, the current literature provides mixed results regarding the above relationship. As this relationship may be “complex and indirect” (Forbes & Milliken, 1999), we argue that QR should help to understand how BGD affects firm performance. Consequently, this study extends the existing literature by providing evidence of the relationship at different points of the conditional distribution.

A study closely related to our work is that by Solakoglu (2013), who also used a QR approach. However, our study differs significantly in three ways. First, consistent with Adams and Ferreira (2009), our QR specification takes into account the issue of endogeneity between BGD and firm performance. Indeed, firm performance may affect the incentive for women to join corporate boards and the motivation of boards to appoint female directors. Consequently, contrary to Solakoglu (2013), in all our specifications we consider the endogeneity problem. Second, the set of corporate control variables used by Solakoglu (2013) is limited to firm age, firm size, and board size. In our study, consistent with Adams and Ferreira (2009) and Carter, Simkins, and Simpson (2003), we include several more corporate variables that have previously been used. Specifically, we encompass corporate governance mechanisms (the size and the independence of the board and CEO/chair duality) and organizational characteristics (firm size and leverage). Finally, our analysis period covers 16 years of study and 1,186 firm-year observations against only a one-year period and 100 observations (Solakoglu, 2013).

Consistent with the current literature, this study examines the effects of BGD on firm financial performance. We do not depart from other research studying diversity at the board level and its impact at the firm level (see Terjesen et al., 2009). Robinson and Dechant (1997) argue that “evidence of diversity’s impact on the bottom line has not been systematically measured and documented for easy retrieval and use.” The purpose of this chapter is also to reconcile previously mixed results by applying the QR approach (Koenker & Bassett, 1978). We believe that this method offers an alternative approach to identifying the characteristics of a relationship that cannot be ascertained through classical linear regression, as previously used in other studies. We argue that the QR method provides a fairly comprehensive picture of how BGD affects firm performance at different conditional quantiles, whereas classical linear regression only estimates the conditional mean of a dependent variable. As a result, we can address the question of whether the significance as well as the magnitude of BGD is different for different levels of firm performance; that is to say, if the relationship is different depending on the level of firm performance. The goal of this paper is, therefore, to re-interpret the relationship between BGD and firm performance through the QR approach.

The structure of our paper is as follows. The literature review is presented in the “Literature Review” section. The “Methods” section describes the data, the variables and the empirical models used in the study. The “Empirical Results” section presents our results. The “Concluding Remarks” section discusses the theoretical and practical implications of the paper as well as the limitations and future research directions.

Literature Review

Theoretical Framework

Does BGD have positive effects on board performance? Arguments derived from several diverse fields, including finance, economics, law, ethics, organizational behavior, and social psychology, have been advanced to answer this question (Carter, D’Souza, Simkins, & Simpson, 2010). The aim of this section is to summarize this literature.

The resource dependency theory developed by Pfeffer (1972) and Pfeffer and Salancik (1978) views a firm as an open system which is dependent upon external organizations and environmental contingencies. Within this framework, the board of directors is perceived as a vehicle to manage external dependency, reduce uncertainty, and lower transaction costs associated with environmental interdependency by linking the organization with its external environment. Resource dependency assumes that corporate directors are chosen in order to maximize access to critical resources. Under this perspective, female directors may provide different benefits to a firm, such as advice and counsel, legitimacy, communication, commitment, and resources because they bring prestige, legitimacy, skills, competences, and knowledge which are different from those of male directors (Hillman et al., 2007).

The presence of female directors can also serve as a signal to women, both within and outside the organization, that a firm is inclined to allow women to develop their careers fully. This will allow the firm to attract and retain top female talent, since they know they will not be discriminated against on the basis of their gender. Moreover, this can encourage both men and women to work harder to show they have the skills and qualifications to become inside directors.

Efforts to address gender discrimination could also have a positive impact on a firm’s ability to attract customers by improving the legitimacy and reputation of the business: firms with a significant brand image, that sell products or services directly to retail customers or to the government or that have large institutional investors (Ferreira, 2010) may be more likely to appoint diverse directors to conform to societal expectations.

Moreover, it has been shown that female directors have experience that differs from that of their male counterparts (Hillman et al., 2007). According to Watson, Kumar, and Michaelsen (1993) and Wiersema and Bantel (1992), this group diversity encourages creativity and brings a greater breadth of perspectives and more innovative solutions to problems. Therefore, the heterogeneity of backgrounds and experiences brought onto a board by female directors could enable more effective problem-solving, enrich discussion, and improve the decisions made by the board.

Finally, since female directors are more likely to be outside directors (Adams & Ferreira, 2009), they may have more freedom than male directors to ask questions (Bilimoria & Wheeler, 2000), and may be less afraid to ask probing and sometimes difficult questions (Bilimoria & Huse, 1997), thus facilitating effective debate on governance issues (Walt & Ingley, 2003).

However, a heterogeneous board might be less efficient since it could be more difficult and time-consuming to reach a consensus (Lau & Murnighan, 1998; Smith, Smith, & Verner, 2006). Similarly, Forbes and Milliken (1999) acknowledge that while diverse groups more easily obtain information and perspectives drawn from outside the dominant group, the pooling and processing of this information may create conflicts between directors.

Therefore, it is not clear whether increasing BGD would have a positive or a negative impact on firm financial performance.

Empirical Evidence regarding Board Gender Diversity and Firm Performance

Table 1 summarizes the empirical evidence regarding the link between BGD and firm performance. In order to produce this summary, we identified relevant articles (in major journals) by querying bibliographic databases (such as EBSCO Business Source Premier and ScienceDirect) and choosing all academic articles that have the terms “women,” “firm performance,” and “board of directors” (or “board”) in the title, abstract, or keywords.

Table 1.

Empirical Evidence of Board Gender Diversity and Firm Performance (Alphabetical Order).

Author(s) (year) N i Country Period Performance Measure Gender Diversity Measure Main Results
Adams and Ferreira (2009) 9,999 (2,500) United States 1998–2003 Tobin’s Q and ROA Proportion of WOCB Negative link
Ahern and Dittmar (2012) 1,230 (248) Norway 2001–2009 Tobin’s Q Proportion of WOCB Negative link
Bøhren and Strøm (2010) 1,290 (229) Norway 1989–2002 Tobin’s Q Proportion of WOCB Negative link
Campbell and Minguez-Vera (2010) 47a (29) Spain 1989–2001 CAARb Dummy variable Positive link
Proportion of WOCB
Campbell and Mınguez-Vera (2008) 408 (68) Spain 1995–2000 Tobin’s Q Dummy variable Positive link (except for dummy variable which yields no significant link)
Proportion of WOCB
Blau and Shannon index
Carter et al. (2010) 2,563 (641) United States 1998–2002 Tobin’s Q and ROA Proportion of WOCB No link
Number of WOCB
Carter et al. (2003) 638 (638) United States 1997 Tobin’s Q and ROA Dummy variable Positive link
Proportion of WOCB
Dale-Olsen, Schøne, and Verner (2013) 73,398c Norway 2003–2007 ROA Proportion of WOCB No link
Erhardt et al. (2003) 112 (112) United States 1998 ROA and ROI Proportion of WOCB Positive link
Mahadeo, Soobaroyen, and Hanuman (2012) 39 (39) Mauritius 2007 ROA Proportion of WOCB Positive link
Miller and del Carmen Triana (2009) 326 (326) United States 2003 ROS and ROI Blau index No link
Rose (2007) 100 (100) Denmark 1998–2001 Tobin’s Q Proportion of WOCB No link
Shrader and Blackburn (1997) 200 (200) United States 1992 ROA, ROS, ROI and ROE Proportion of WOCB Negative link
Smith et al. (2006) 12,085 (2,500) Denmark 1993–2001 Gross profit/net sales, contribution margin/net sales, operating income/net assets, and net income after tax/net assets Proportion of WOCB Positive link (depending on education of women and firm mesure of performance)
Solakoglu (2013) 100 (100) Turkish 2005–2006 ROA and ROI Net income after tax/net assets The link depends on the conditional distribution of firm performance

Notes: aAnnouncements of female appointments; bCumulative average abnormal returns; cUnknown; N i refers to the sample size with the number of firm-year observations (number of firms). The performance measure represents the dependent variable, while the gender diversity measure represents the independent variable.

The mixed results shown in Table 1 are likely to be explained by several factors. First, some of the differences may be due to the different time period (Campbell & Minguez-Vera, 2010). Some studies examine a single year of study (e.g., Carter et al., 2003; Miller & del Carmen Triana, 2009), while others look at several years of data (e.g., Adams & Ferreira, 2009; Smith et al., 2006). Second, the country studied (e.g., the United States, Norway, Spain, and Mauritius) are relatively different in terms of institutional (Grosvold & Brammer, 2011) and political systems (Terjesen & Singh, 2008). Third, from an empirical standpoint, the aforementioned studies in Table 1 use different measures of firm performance (e.g., Tobin’s Q, return on assets (ROA), or return on investment), female representation on the board of directors (e.g., dummy variable, percentage of women on corporate boards (WOCBs), or the Blau index) and different estimation methods (e.g., OLS, fixed effects (FE), or hierarchical regression) (Campbell & Mınguez-Vera, 2008; Carter et al., 2010). These different explanations are likely to contribute to the mixed results shown in Table 1.

Methods

Data and Sample Procedure

All the proxy statements for each firm belonging to the S&P 100 Index were obtained from the Securities and Exchange Commission website. 1 This enabled us to determine the exact composition of the 100 boards of directors between 1995 and 2010. We used the S&P 100 Index (which is a subset of the S&P 500) on the assumption that firms in this Index are among the largest companies in terms of aggregate market capitalization. Indeed, its constituents represent about 57% of the market capitalization of the S&P 500 and almost 45% of the market capitalization of the US equity markets. As a result, the companies in the S&P 100 Index tend to be the largest and most well-established companies. As many studies report correlations between firm size (measured either by market capitalization or revenue) and the number of WOCBs (e.g., Hillman et al., 2007; Peterson & Philpot, 2007), these large firms enabled us to test the impact of BGD on firm financial performance. We therefore assumed that if there were any relationship between BGD and firm financial performance, it would be more likely to be observed in these firms. The S&P 100 Index has been used previously in the literature: for example, Cornett, Marcus, and Tehranian (2008) and George and Longstaff (1993).

Furthermore, consistent with Farrell and Hersch (2005) and Hermalin and Weisbach (1988), we excluded financial firms 2 (SIC codes 6000–6999) and the utilities industry 3 (SIC codes 4900–4999), since the boards of directors in these companies may be subject to regulatory supervision affecting their governance system. Consequently, the final data set consisted of an unbalanced set of 78 firms and 1,186 firm-year observations.

Measures of Variables

Dependent Variable

Firm performance was measured through Tobin’s Q. Specifically, we used an approximation of Tobin’s Q based on the methodology of Lindenberg and Ross (1981). This approach has the advantage of using basic financial and accounting information, thereby avoiding any problems of availability related to the estimation of more theoretically correct methods such as the one developed by Lindenberg and Ross (1981). Consistent with Chung and Pruitt (1994), Tobin’s Q is defined in our study as follows:

(1) Approximate q = ( MVE + PS + DEBT ) / TA
where MVE is the product of a firm’s share price and the number of common stock shares outstanding, PS is the liquidating value of the firm’s outstanding preferred stock, DEBT is the value of the firm’s short-term liabilities net of its short-term assets, plus the book value of the firm’s long-term debt, and TA is the book value of the total assets of the firm.

Chung and Pruitt (1994) find that at least 96.6% of the total variability in the more common Lindenberg and Ross (1981) method of Tobin’s Q is explained by their approximation. Similarly, Perfect and Wiles (1994) find that the correlation coefficient between Lindenberg and Ross’s procedures and the simple Tobin’s Q (very similar to the procedure used here in this study) is 0.93.

In order to enhance normality by correcting for possible positive skewness, a log transformation of Tobin’s Q was employed. This procedure also allowed us to reduce the impact of outliers (Hirsch & Seaks, 1993). Finally, log transformation often reduces heteroskedasticity, since it compresses the scale by which the variables are measured, thereby reducing a 10-fold difference between two values to a two-fold one. Consequently, we took the natural logarithm of Q as our dependent variable. To the best of our knowledge, no studies examining the relationship between BGD and firm performance have used the log transformation of Tobin’s Q.

Independent Variables

In this study, we measured BGD through four measures that have been previously used in the literature. First, consistent with Carter et al. (2003) and Hillman et al. (2007), we used a dummy variable, called “DWOCB,” that equals 1 if there is at least one female director on a corporate board, and 0 otherwise. Second, in line with Adams and Ferreira (2009) and Ahern and Dittmar (2012), we employed the percentage of WOCB (PWOCB), calculated as the number of female directors divided by the total number of directors. Third, consistent with Campbell and Mınguez-Vera (2008) and Miller and del Carmen Triana (2009), we used Blau (1977) index of heterogeneity (BWOCB), measured as ( 1 p i 2 ) , where p i is the percentage of board members in each category i (in this case, male and female) and n is the total number of board members. Blau’s index for BGD can range from 0 (when there is no female director on a corporate board) to 0.50 (which occurs when there is an equal number of female and male directors). Finally, consistent with Liu et al. (2014) and Nekhili and Gatfaoui (2013), we employed the number of female directors (NWOCB). We considered this variable to the extent that Nekhili and Gatfaoui (2013) argued that the percentage of WOCB does not say anything about the number of board seats held by women. Indeed, two firms can exhibit the same percentage of WOCB while having different numbers of WOCB. Therefore, we considered this fourth measure of BGD to ensure the robustness of our results. To the best of our knowledge, no studies have simultaneously used these four measures of BGD.

Control Variables

In line with the previous corporate board (Bebchuk & Cohen, 2005; Bhagat & Bolton, 2008; among others) and BGD (Adams & Ferreira, 2009; Carter et al., 2010; among others) literature, we included several corporate control variables that have previously been studied. Specifically, we included two firm characteristic variables: firm size and leverage. Indeed, the size of a firm is often used as a control variable when analyzing the financial performance of a firm. For example, Fama and French (1992) show that market returns are related to firm size. Subsequently, many studies have shown that asset size is associated with Tobin’s Q (e.g., Faleye, 2007; Yermack, 1996). In studies related to BGD, firm size is often considered in the specifications (Adams & Ferreira, 2009; Liu et al., 2014). Specifically, we measured firm size as the natural logarithm of total assets (Campbell & Mınguez-Vera, 2008; Carter et al., 2010). Proponents of agency theory (Jensen & Meckling, 1976) posit that financial leverage may be an efficient governance mechanism in reducing the conflict of interest between shareholders and managers. As leverage can reduce the agency cost of free cash flow and prevent managers adopting decisions that may destroy value, leverage can, therefore, increase firm value (Jensen, 1986). As a result, leverage is used as a control variable. Campbell and Mınguez-Vera (2008) and Liu et al. (2014), among others, have taken the same approach. Specifically, firm leverage is calculated as total debt divided by total assets (Bhagat & Bolton, 2008; Campbell & Mınguez-Vera, 2008).

Consistent with Carter et al. (2010), we also included a second set of control variables associated with aspects of corporate governance. These variables have been shown to be related to firm performance. First, both the theoretical and empirical literature suggest that we should include board size in the financial performance equation. Indeed, on one side, the majority of US empirical studies show an inverse association between board size and firm value (e.g., Eisenberg, Sundgren, & Wells, 1998; Yermack, 1996), thereby pushing Hermalin and Weisbach (2003) to assert that this relation is one of the most prominent empirical regularities in the literature. On the other hand, Dalton, Daily, Ellstrand, and Johnson (1998) and Pearce and Zahra (1992) find a positive relationship between board size and firm performance. They argue that a larger board will bring intellectual knowledge which, in turn, will increase the quality of strategic decisions that ultimately have an impact on firm performance. The positive relationship between board size and firm performance flows from resource dependency theory, while the negative view is grounded in agency theory. Consistent with Campbell and Mınguez-Vera (2008), we added board size as a control variable. Board size was measured as the total number of directors on the board. Second, from an agency perspective, it is claimed that having a greater proportion of outside (or independent) directors is likely to mean that they act independently in the interests of shareholders when there is a potential conflict between managers and shareholders. However, it must be acknowledged that the evidence on the relation between board composition and firm performance is mixed. For example, Brickley, Coles, and Terry (1994) and Byrd and Hickman (1992) find that independent boards create value, while Hermalin and Weisbach (1991) and Bhagat and Black (2002) find no relation between board composition and Tobin’s Q. Finally, Agrawal and Knoeber (1996) and Yermack (1996) report a negative relationship. These mixed results notwithstanding, the agency theory approach was adopted in this study for the examination of the impact of board composition on firm performance. This line of reasoning has been applied in the past to examine the relationship between BGD and firm performance: for example, Adams and Ferreira (2009) and Carter et al. (2010). As a result, we took board independence into account in our specifications, measured as the fraction of outside (independent) directors on the board. Consistent with Hermalin and Weisbach (1991), among others, we defined “outside” (independent) directors as those who have any affiliation with the firm other than their role as director. Third, the previous literature acknowledges that the type of board leadership and the power of the chief executive officer (CEO) may have an influence on firm financial performance. For instance, Adams, Almeida, and Ferreira (2005) argue that the ability of a CEO to influence decisions is likely to have an impact on firm performance. The assumption underlying this CEO ability is linked to his or her level of power 4 (Finkelstein, 1992). Although the results are mixed – for instance, Goyal and Park (2002) find that the sensitivity of CEO turnover is lower when the CEO and chair duties are assumed by the same person, while Brickley, Coles, and Jarrell (1997) show that a combined CEO-Chair leads to lower market value and cash flows – we added CEO and board chair duality as a control variable, in line with Carter et al. (2010) and Carter et al. (2003). We used a dummy variable equal to 1 if the chair is also the CEO of the firm, and 0 otherwise.

An overview of the variables we used in this study and their definitions is provided in Table 2.

Table 2.

Definition of Variables.

Variable Definition
Tobin’s Q Approximate q = (MVE + PS + DEBT)/TA; where MVE is the product of a firm’s share price and the number of common stock shares outstanding, PS is the liquidating value of the firm’s outstanding preferred stock, DEBT is the value of the firm’s short-term liabilities net of its short-term assets, plus the book value of the firm’s long-term debt, and TA is the book value of the total assets of the firm (Chung & Pruitt, 1994)
DWOCB Dummy variable that equals to “1” if there is at least one female director on corporate boards, and “0” otherwise (Hillman et al., 2007)
PWOCB Percentage of WOCB (Campbell & Mınguez-Vera, 2008)
BWOCB Blau’s index of diversity (Blau, 1977)
NWOCB Total number of female directors on the board (Carter et al., 2010)
Firm size Natural logarithm of total assets (Campbell & Mınguez-Vera, 2008)
Leverage Total debt divided by total assets (Bhagat & Bolton, 2008)
Board size The number of directors on the board (Coles, Daniel, & Naveen, 2008)
Board independence Ratio of number of independent directors to board size (Coles et al., 2008)
Duality Dummy variable that equals to “1” if the chair is also the CEO of the firm, and “0” otherwise (Carter et al., 2003)

Notes: This table defines the variables used in this study. The source of data is, on the one hand, the Securities and Exchange Commission website for the corporate governance data (gender, board size, board independence, and duality of CEO and board chair) and, on the other hand, Thomson ONE Banker for financial data (firm size and leverage).

Empirical and Estimation Models

Empirical Model

Consistent with Adams and Ferreira (2009) and Liu et al. (2014), our fundamental model took the following form:

(2) ( Q ) i , t = α + β 1 ( WOCB ) i , t + β 2 ( Firm _ Char ) i , t + β 2 ( Board _ Char ) i , t + Ψ + η + ε i , t
where i denotes firms in the sample (i = 1, 2, […], 78) and t refers to time period (t = 1995, 1996, […], 2010); Q is the financial performance of a firm measured by the approximation of Tobin’s Q (Chung & Pruitt, 1994); WOCB represents our four alternatives for BGD: a dummy variable that takes the value of 1 if there is at least one female director on the board (DWOCB); the percentage of WOCB (PWOCB); Blau’s index of diversity (BWOCB); and the number of WOCB (NWOCB); Firm_Char and Board_Char are the sets of control variables related to firm characteristics and board characteristics. Finally, the expressions Ψ and η refer to time effects and unobservable heterogeneity, respectively.

The coefficient of primary interest is β 1 and H 0 : β 1 = 0 (the null hypothesis) against the alternative hypothesis which is H a : β 1# 0. Rejection of the null hypothesis implies that BGD has an impact on firm value (either positive or negative).

Methodology

Longitudinal, or panel, analysis 5 was used here to control for any omitted/unobservable variables that could induce causal inference in observational studies (Hsiao, 2003; Lee, 2002). The problem of “omitted variable bias” or “unobserved heterogeneity” is a serious issue in empirical research, especially in the board composition (see Adams, Hermalin, & Weisbach, 2010) and BGD (Adams & Ferreira, 2009) literature. This is because empirical findings may demonstrate different outcomes because of the failure to control for endogeneity and/or the tested relationship is heterogeneous across firms. Traditional longitudinal/panel data analysis with FE could account for endogeneity under certain assumptions (Wooldridge, 2013). Adams and Ferreira (2009) employ firm FE in their analysis and show that firm FE have a significant impact on the relationship between BGD and financial performance. We believe that fixed-effects analysis is important to the extent that it helps to mitigate omitted variables and address unobserved changes over time. All regressions were estimated with robust standard errors in order to correct for residual heteroskedasticity issues (Gujarati & Porter, 2009).

Another concern when analyzing the relationship between BGD and firm performance is simultaneity/reverse causality (Adams & Ferreira, 2009). In practice, BGD may affect firm performance but it is also possible that profitable firms may appoint female directors. In other words, Adams and Ferreira (2009) argue that firm performance can affect both the incentive for and the motivation behind BGD. We addressed the issue of reverse causality by employing the generalized method of moments (GMM), also known as the GMM estimator, following Arellano and Bover (1995). This methodology accounts for endogeneity, and the system GMM estimator uses as instruments the lagged values of the dependent variable both in levels and in differences. Furthermore, it uses the lagged values of other regressors that may suffer from endogeneity. Finally, the system GMM estimator controls for unobserved heterogeneity and for the persistence of the dependent variable. Specifically, we used Arellano and Bond’s (1991) difference GMM and Blundell and Bond (1998) system GMM.

According to Wintoki, Linck, and Netter (2012), the Arellano-Bond dynamic panel data estimator takes a variety of factors, such as endogeneity from unobservable heterogeneity, simultaneity, or a dynamic relation, that may have an impact on the relationship between board structure and past firm performance. Arellano and Bond (1991) used a GMM framework to develop their instruments. They transformed the model into first-differences, which eliminated the individual effects from the model. They showed that the lagged dependent variable values (levels) represent legitimate instruments for the first-differences variable. They also assumed that the residuals were free from second-order serial correlation. The Arellano-Bond estimator was previously used in BGD literature by Adams and Ferreira (2009) and Liu et al. (2014).

However, Arellano and Bover (1995) argue that lagged values provide little information regarding a first-differenced variable, especially when they are serially correlated. As a result, Blundell and Bond (1998) suggest an alternative approach using the GMM “system” estimator. Specifically, in addition to the first-differences used by Arellano and Bond (1991), Blundell and Bond (1998) use lagged first-differences as instruments in a non-transformed (levels) equation. Not used in the BGD literature, the Blundell-Bond estimator has been used in corporate governance literature (Ammann, Oesch, & Schmid, 2011; Wintoki et al., 2012; among others). Therefore, we used the Blundell and Bond’s (1998) approach to confirm (or dispel) the Arellano and Bond (1991) differences.

Quantile Regression: An Overview

QR estimates the conditional quantiles of a response variable in a linear model. It provides a more complete picture of the possible relationships between a response variable and the explanatory variables at different points of the conditional distribution (Koenker & Bassett, 1978; Koenker & Hallock, 2001). In addition, QR does not require strict assumptions, as does classical linear regression as to normality, homoskedasticity, and the absence of outliers (Johnston & DiNardo, 2007).

We summarize the linear QR model below. Specifically, let (y i , x i ), where subscript i denotes a sample of observations from some population and x i is a vector of explanatory variables that match y i . If we assume that the θ th quantile of the conditional distribution y i is linear in x i , we can write the conditional QR model (Buchinsky, 1998) as follows:

(3) y i = x i β θ + u θ i
where β θ is an unknown k × 1 vector of regression parameters associated with the θ th quantile, x i is a k × 1 vector of independent variables, y i is the dependent variable, and µ θi is the unknown error term.

The θ th conditional quantile of y given x is:

(4) Quant θ ( y i / x i ) inf { y i = F i ( y / x ) 0 } = x i β θ
and
(5) Quant θ ( μ θ i / x i ) = 0

The θ th QR (0 < θ < 1) of y is the solution to the minimization of the sum of absolute deviation residuals:

(6) min β θ { i : y x i β θ | y i x i β θ | + i : y x i β ( 1 θ ) | y i x i β θ | } = min β θ ρ θ ( μ θ i )

As to our research question, the QR specification with panel data takes the following form:

(7) Quant θ ( y i t / x i t ) = α + β θ WOCB i t + β θ x i t + μ i t
where y it is the dependent variable at quantile θ. WOCB it represents BGD measured through the percentage of WOCB and the x it vector encompasses board characteristics and firm characteristics.

Chernozhukov and Hansen (2005, 2006) argue that in observational studies (such as those on board composition; see Adams et al., 2010), the variable of interest is endogenous, therefore the use of conventional QR (presented previously) is inconsistent with and even inappropriate for capturing the causal effects of these variables on QR outcomes. As a result, consistent with the aforementioned authors, we dealt with endogeneity by using a control variable approach. In short, the underlying idea was to add a variable to the regression, such that once we condition on this variable, regressors and error terms became independent. This so-called control variable is usually unobservable and needs to be estimated at the first stage. This method allowed us to evaluate the impact of endogenous variables or treatments on the entire distribution of outcomes. Numerous methods have been proposed to deal with endogeneity (e.g., Chernozhukov & Hansen, 2005, 2006). Consistent with Kwak (2010), under Stata©, we implemented the instrumental variable quantile regression (IVQR) method of Chernozhukov and Hansen (2004, 2008), since that it is computationally efficient for a small number of endogenous variables. Specifically, we used the Stata© command “ivqreg,” which performs QR using the robust standard error formula from Chernozhukov and Hansen (2006) for an exactly identified IV case and the formula in Chernozhukov and Hansen (2008) for an over-identified IV case to evaluate the heterogeneous marginal effects of the endogenous treatment variables. 6

Finally, we used the bootstrap method to calculate the standard errors for the regression coefficients. Indeed, Buchinsky (1995) recommends this method for small samples as it produces more robust results. Specifically, 1,000 bootstrap replications were implemented to guarantee a small variability in the covariance matrix.

Descriptive Statistics and Correlation Analysis

Descriptive Statistics

Table 3 presents descriptive statistics for the variables used in this study, as well as the corresponding variance inflation factors (VIF 7 ).

Table 3.

Summary Statistics.

Variable Mean Std Min Percentile 25 Median Percentile 75 Max
Panel A: Performance measure
Tobin’s Q 2.51 2.47 0.22 1.13 1.74 2.99 26.79
Panel B: Female directors
DWOCB 0.94 0.00 1.00 1.00 1.00 1.00
PWOCB 0.15 0.09 0.00 0.09 0.14 0.20 0.60
BWOCB 0.24 0.11 0.00 0.17 0.25 0.32 0.50
NWOCB 1.86 1.08 0.00 1 2 2 6
Panel C: Firm characteristics
Firm size 4.38 0.50 2.48 4.09 4.40 4.65 5.90
Leverage 22.28 15.31 0.00 11.43 21.00 30.40 131.32
Panel D: Corporate governance characteristics
Board size 12.05 2.56 5.00 11.00 12.00 13.00 23
Board independence 0.46 0.29 0.00 0.22 0.50 0.71 0.95
Duality 0.75 0.44 0.00 0.00 1.00 1.00 1.00

Notes: This table presents the descriptive statistics used in this study by showing mean, standard deviation (Std), minimum (Min), percentile 25, median, percentile 75, and maximum (Max). The variables are defined in Table 2.

Panel A of Table 3 reports summary statistics for the approximation of Tobin’s Q. The mean average of Tobin’s Q is 2.51 (median = 1.74) over the period study. By way of comparison, we noticed that our Tobin’s Q was higher than those mentioned by Adams and Ferreira (2009) and Carter et al. (2010), which were, respectively 2.09 and 1.19 (mean average of Tobin’s Q). This result presumably reflects, first, the study period (1995–2010 against 1996–2003 and 1998–2002, respectively) and, second, the type of companies selected (S&P 100 against S&P 500, MidCaps, and S&P SmallCaps, and S&P 500, respectively).

Panel B of Table 3 presents the summary statistics for our various measures of BGD. Over the period 1995–2010, we noted that almost 94% of the companies in our sample have at least one woman on their board. Our result contrasts with that of Hillman et al. (2007), who reported that only 56% of the 1,000 largest US companies (in terms of sales) between 1990 and 2003 had one or more female directors on the boards. Furthermore, we find that 15.2% of board seats are held by female directors. This is higher than the figures previously reported. For instance, Adams and Ferreira (2009) report a value of 9.3%, while Farrell and Hersch (2005) find that women held 6.9% of the Fortune 500 board seats over the period 1990–1999. In addition, the average Blau’s index of diversity is equal to 0.24 (STD = 0.11). In comparison, Miller and del Carmen Triana (2009) report an average Blau gender diversity score of 0.21 (STD = 0.10) for companies belonging to Fortune 500 in 2003. Finally, the average number of female directors on corporate boards in our study is 1.86 (STD = 1.08), while Carter et al. (2010) show that the average number of female directors on boards is equal to 1.30 (STD = 0.93). Overall, our BGD measures are higher than those reported by the current literature, probably because of the sample (S&P 100) and the time period studied.

Panels B and C present the summary statistics for the control variables used in this study. For instance, average firm size (measured by the logarithm of sales) is 4.38, while Adams and Ferreira (2009) report a value of 7.26. In addition, the average leverage of our companies is 22.28%. Bhagat and Bolton (2008) find a mean leverage of 42.69%. In terms of corporate governance, we notice that board size and board independence are equal to 12.05 and 46%, respectively. Adams and Ferreira (2009) report values of 9.38 and 63%, respectively.

Correlation Analysis

Table 4 provides a correlation matrix of our key variables. As a rule of thumb, a correlation of 0.70 may indicate a multicollinearity issue (e.g., Mela & Kopalle, 2002). Table 4 indicates that the highest correlation coefficient is equal to 0.97 (in bold) between PWOCB (percentage of WOCB) and BWOCB (Blau’s index of diversity). To the extent that we use these two measures of BGD interchangeably in our specifications, the high correlation is not an issue. We make the same comments regarding NWOCB (number of female directors) and PWOCB, and between NWOCB and BWOCB. We can see that no other correlation coefficient has an absolute value higher than 0.70. In addition, we generated the VIF to confirm this result. The highest observed VIF value in our study variables is 1.33, which is well below the conventional cutoff of 10.0 (Chatterjee & Hadi, 2012). Consequently, we conclude that multicollinearity had little impact on our further analyses.

Table 4.

Correlation Matrix.

Variables 1 2 3 4 5 6 7 8 9 10
1. Tobin’s Q 1.00
2. DWOCB −0.04 1.00
3. PWOCB −0.01 0.45*** 1.00
4. BWOCB −0.02 0.57*** 0.97*** 1.00
5. NWOCB −0.08*** 0.44*** 0.92*** 0.90*** 1.00
6. Firm size −0.35*** 0.25*** 0.14*** 0.20*** 0.28*** 1.00
7. Leverage −0.28*** 0.13*** 0.12*** 0.11*** 0.16*** 0.14*** 1.00
8. Board size −0.21*** 0.31*** 0.04 0.09*** 0.38*** 0.42*** 0.23*** 1.00
9. Board independence −0.12*** 0.18*** 0.11*** 0.13*** 0.13*** 0.13*** 0.12*** 0.15*** 1.00
10. Duality −0.09*** 0.05 0.05 0.05 0.05 0.08*** 0.10*** 0.05 −0.14*** 1.00
VIF 1.22 1.15 1.04 1.06 1.20 1.25 1.08 1.33 1.08 1.05

Notes: The table presents the correlation matrix among all the variables employed in this study. The variables are defined in Table 2. *** indicates significance at the 1% level, respectively.

Table 4 also reveals several significant correlations between variables. First, our measures of BGD appear to be significantly correlated (at the 1% level), in particular between NWOCB, PWOCB, and BWOCB, thereby confirming the use of these measures in our specifications to measure the effect of BGD on firm performance. Second, there is a significant and positive correlation between BGD and firm size. This result suggests, in line with Hillman et al. (2007), that larger firms are more likely to appoint female directors to the board.

Trends in BGD

Fig.1 shows the time trends of BGD from 1995 to 2010. Fig.1 shows that, on average, the female representation on boards of directors gradually rose from 11.93% in 1995 to 17.94% in 2010 (PWOCB). The average number of WOCB increased significantly from 1.50 in 1995 to 2.14 in 2015 (NWOCB). The difference in the means were statistically significant at the 1% level. This seems to suggest that women’s representation on corporate boards increased significantly during the study period.

Fig. 1. 
Board Gender Diversity Trends: 1995–2010.

Fig. 1.

Board Gender Diversity Trends: 1995–2010.

Bivariate Relationships

Figs. 2–4 provide an initial assessment of the relationship between WOCB on Tobin’s Q using three different BGD measures. For reliability, the graphs only show the most recent five years for the sample.

Fig. 2. 
Representation of WOCB and Median Tobin’s Q.

Fig. 2.

Representation of WOCB and Median Tobin’s Q.

Fig. 3. 
Number of WOCB and Median Tobin’s Q.

Fig. 3.

Number of WOCB and Median Tobin’s Q.

Fig. 4. 
Blau’s Index of Diversity and Median Tobin’s Q.

Fig. 4.

Blau’s Index of Diversity and Median Tobin’s Q.

Fig. 2 shows the median Tobin’s Q each year according to whether a firm had any female directors. The four graphs within Fig. 2 are separated into quartiles of Tobin’s Q to account for the possibility that the relationship differed depending on the region of the distribution. Fig. 2 raises two interesting points. First, there are no clear trends in the lower two quartiles to indicate that any kind of relationship took place between the representation of WOCB and Tobin’s Q. Second, however, and quite surprisingly, in the quartile there seems to be a clear tendency for firms without WOCB to significantly outperform those with at least one female director on the board.

Fig. 3 seems to show a similar path when examining the number of WOCB. There are no clear patterns in the lower three quartiles, whereas in the highest quartile the firms without any WOCB performed markedly better than those with even one WOCB. Among the best-performing firms, women seem to be absent from the corporate boards.

Fig. 4 uses a different metric for WOCB: Blau’s index of diversity. For ease of understanding, the index has been split into quartiles. 8 In comparison with the previous figures, there is no particular trend, even in the highest quartile of Tobin’s Q.

In conclusion, these comparisons suggest that, if there is any relationship between BGD and firm performance, it may depend, on the one hand, on which part of the conditional distribution one is considering and, on the other, which measure of BGD is used.

Empirical Results

Baseline Estimation

Table 5 presents the results based on the FE method with robust standard errors clustered on the firm. Models 1–4 demonstrate that the coefficients for BGD are not statistically different from 0, which means that there is no evidence of a significant relationship between Tobin’s Q and BGD. This is the case regardless of which of the four BGD measures was used. Consequently, on the basis of the FE method, BGD does not appear to have any effect on Tobin’s Q. At first glance, our results are in line with the work of Francoeur et al. (2008) and Rose (2007), among others, who find no relationship between BGD and firm financial performance.

Table 5.

Fixed Effects of the Relationship Board Gender Diversity and Firm Performance.

Independent Variable Model 1: DWOCB Model 2: PWOCB Model 3: BWOCB Model 4: NWOCB
WOCB measure 0.0934 −0.010 −0.031 −0.010
(0.121) (0.254) (0.343) (0.026)
Firm size 0.528*** −0.512*** −0.511*** −0.506***
(0.106) (0.107) (0.106) (0.106)
Leverage −0.012*** −0.014*** −0.012*** −0.012***
(0.003) (0.003) (0.003) (0.003)
Board size 0.008 0.008 0.009 0.010
(0.011) (0.011) (0.011) (0.011)
Board independence −0.017 −0.019 −0.019 −0.019
(0.084) (0.084) (0.084) (0.084)
Duality of CEO and board chair −0.041 −0.042 −0.042 −0.043
(0.057) (0.056) (0.056) (0.056)
Constant 3.075*** 3.081*** 3.080*** 3.061***
(0.454) (0.461) (0.461) (0.468)
Number of observations 1,179 1,179 1,179 1,179
R² 0.153 0.151 0.151 0.151
Firm fixed effects Yes Yes Yes Yes

Notes: The source of data is, on the one hand, the Securities and Exchange Commission website for the corporate governance data (gender, board size, board independence and duality of CEO and board chair) and, on the other hand, Thomson ONE Banker for financial data (firm size and leverage). The variables are defined in Table 2. Robust standard errors are in parentheses. Asterisks indicate significance at 0.001 (***), 0.01 (**), and 0.05 (*) levels.

As far as the control variables are concerned, the variables associated with board characteristics are not significantly correlated with Tobin’s Q. However, the variables associated with the characteristics of the firm are significant. Specifically, firm size (except for the Model 1) and leverage have a negative impact on the firm performance (at the 0.1% level) regardless of which BGD measures are used. These results suggest that larger companies have lower Tobin’s Q, consistent with Campbell and Mınguez-Vera (2008), and higher leverage also seems to lower Tobin’s Q.

As suggested by Adams and Ferreira (2009), the fixed-effects method takes into partial account the issue of endogeneity in BGD. In order to confirm (or dispel) the previous results, we used fixed-effects instrumental variables (FE-IVs; see Liu et al., 2014). We used a company’s industry as an instrument. Indeed, previous research has shown a company’s industry to be a factor that potentially drives the presence of WOCB. For instance, Brammer, Millington, and Pavelin (2007) document significant cross-sector variation in BGD in industries such as retail, media, and banking, in comparison with any other industry. Hillman et al. (2007) also find that industries that employ a higher proportion of women will have greater BGD. As a result, we used the company’s industry as our instrument. To do this, we use Campbell’s (1996) industry classifications, which allowed us to capture any variation across industries. Specifically, we used industry dummies (basic industry, capital goods industry, textiles/trade industry, consumer durables industry, transportation industry, utilities industry, petroleum industry, food/tobacco industry, services industry, construction industry, and leisure industry, based on the SIC codes).

Table 6 presents the results based on the FE-IV method with robust standard errors. Model 5 shows that the presence of WOCB (DWOCB) does not have a significant effect on Tobin’s Q. This result seems to suggest that the mere presence of WOCB does not affect firm performance, as highlighted by Campbell and Mınguez-Vera (2008). On the other hand, the other three measures of BGD (PWOCB, BWOCB, and NWOCB) are positively and significantly correlated with firm performance (at the 5% level). These positive and significant coefficients tend to indicate that BGD has a significant influence on firm financial performance.

Table 6.

Fixed Effects Instrumentals Variables Models.

Independent Variable Model 5: DWOCB Model 6: PWOCB Model 7: BWOCB Model 8: NWOCB
WOCB measure 1.846 4.498* 6.008* 0.520*
(1.602) (2.121) (2.991) (0.265)
Firm size −0.782** −0.921*** −0.907*** −0.885***
(0.240) (0.200) (0.204) (0.197)
Leverage −0.013*** −0.014*** −0.014*** −0.013***
(0.002) (0.002) (0.002) (0.002)
Board size −0.020 0.034* 0.041* −0.029
(0.026) (0.015) (0.019) (0.021)
Board independence 0.002 −0.030 −0.037 −0.017
(0.071) (0.069) (0.070) (0.071)
Duality of CEO and board chair −0.035 −0.001 −0.005 0.004
(0.046) (0.050) (0.050) (0.053)
Constant 2.907*** 3.508*** 3.553*** 4.229***
(0.301) (0.325) (0.350) (0.638)
Number of observations 1,179 1,179 1,179 1,179
R² 0.034 0.093 0.101 0.105
Firm fixed effects Yes Yes Yes Yes

Notes: The source of data is, on the one hand, the Securities and Exchange Commission website for the corporate governance data (gender, board size, board independence and duality of CEO and board chair) and, on the other hand, Thomson ONE Banker for financial data (firm size and leverage). The variables are defined in Table 2. Robust standard errors are in parentheses. Asterisks indicate significance at 0.001 (***), 0.01 (**), and 0.05 (*) levels.

Concerning the control variables, the results shown in Table 6 reiterate the significant negative effects found in Table 5 for firm size and leverage (at the 0.1% level), and the non-significance of the board’s variables (except for board size in Models 6 and 7, which is positively and significantly correlated with Tobin’s Q at the 5% level).

The accuracy of the inferences from Table 6 depends on the quality of the instruments. Regressions of each BGD variable on SIC codes returned a small R² of 0.034 for whether any women were present on the board, although the F test easily rejected the null of no explanatory power. The R² for the other BGD measures were higher and ranged from 0.093 to 0.105. Sargan-Hanson statistics were also calculated following each of the instrumental variable models. The null hypothesis of the test was that the set of instruments (the SIC category plus the other exogenous variables in the model) is valid (i.e., do not correlate with the error and are not part of the second stage model). The results failed to reject the null hypothesis of valid instruments; the smallest p-value for the chi-square distributed statistic was 0.0923 following the first model. These results support the use of SIC as an instrument.

Dynamic Models

The results in the previous section do not take into account dynamics as a continued but decaying effect of the predictors. Consistent with Liu et al. (2014), in order to address the endogeneity issue, we first applied the Arellano-Bond method by including the lagged value of Tobin’s Q, as shown in equation (2). Table 7 presents results from a dynamic panel model using the Arellano-Bond estimator, which relies on a matrix of earlier lagged variables to instrument the lagged dependent variable. The estimator also allows endogenous right-hand-side variables that are similarly instrumented. Only the first lag of the outcome is included in the models; subsequent lags were not found to be significant.

Table 7.

Arellano-Bond Dynamic Panel Estimations.

Independent Variable Model 9: DWOCB Model 10: PWOCB Model 11: BWOCB Model 12: NWOCB
Lagged Tobin’s Q 0.600*** 0.581*** 0.569*** 0.591***
(0.062) (0.056) (0.056) (0.056)
WOCB measure −0.056 −0.798 −1.022 −0.049
(0.092) (0.537) (0.749) (0.049)
Firm size 0.040 0.131 0.135 0.048
(0.077) (0.125) (0.130) (0.115)
Leverage −0.004 −0.004 0.003 −0.003
(0.003) (0.003) (0.003) (0.003)
Board size 0.013 0.008 0.004 0.015
(0.011) (0.010) (0.011) (0.011)
Board independence 0.036 0.056 0.058 0.076
(0.091) (0.082) (0.083) (0.084)
Duality of CEO and board chair −0.004 0.013 0.017 0.019
(0.042) (0.041) (0.043) (0.043)
Constant 0.067 −0.151 −0.174 −0.004
(0.400) (0.510) (0.531) (0.546)
Number of observations 1,023 1,023 1,023 1,023

Notes: The source of data is, the Securities and Exchange Commission website for the corporate governance data (gender, board size, board independence and duality of CEO and board chair) and Thomson ONE Banker for financial data (firm size and leverage). The regressions are estimated by using a dynamic panel data estimator as proposed by Arellano and Bond (1991). The variables are defined in Table 2. Robust standard errors are in parentheses. Asterisks indicate significance at 0.001 (***), 0.01 (**), and 0.05 (*) levels.

The estimates show that the bulk of the model’s explanatory power comes from the lagged dependent variable, meaning that there are very strong carryover effects of firm performance from one year to the next. The coefficients for the remaining variables were interpreted as the effect on the dependent variable in the current year. Given that the coefficient estimate for the lagged dependent variable is less than 1, the effects of the independent variables will also carry over into subsequent years at a decaying rate. None of the current-year effects were found to be significant. Follow-up tests found significant first-order autocorrelations but non-significant second-order autocorrelations, consistent with the estimator’s assumptions. The models fitted using robust standard errors, so the Sargan test was unavailable.

Table 7 shows that none of our BGD variables is significantly correlated to firm performance (Tobin’s Q) at the 10% level. Regarding the control variables, none is statistically significant.

The Arellano-Bond estimator is known to be biased in cases where the coefficient on the lagged dependent variable approaches 1 or if the panel-level variance is large relative to the firm-level error. Table 8 presents results using the estimator developed by Blundell and Bond (1998), which, by adding additional moment conditions to the GMM estimator, can overcome these limitations. The results, however, are almost identical to those in Table 7.

Table 8.

Blundell-Bond Dynamic Panel Estimations.

Independent Variable Model 13: DWOCB Model 14: PWOCB Model 15: BWOCB Model 16: NWOCB
Lagged Tobin’s Q 0.683*** 0.686*** 0.687*** 0.692***
(0.053) (0.053) (0.052) (0.052)
WOCB measure 0.086 −0.370 −0.217 −0.003
(0.089) (0.412) (0.578) (0.035)
Firm size 0.053 0.096 0.063 0.028
(0.064) (0.084) (0.083) (0.075)
Leverage −0.007 −0.005 −0.004 −0.004
(0.004) (0.004) (0.004) (0.004)
Board size −0.001 0.003 0.002 0.004
(0.012) (0.011) (0.011) (0.010)
Board independence −0.083 −0.031 −0.021 −0.025
(0.09) (0.088) (0.089) (0.085)
Duality of CEO and board chair −0.047 −0.028 −0.026 −0.034
(0.043) (0.040) (0.041) (0.040)
Constant 0.129 −0.002 0.075 0.183
(0.343) (0.352) (0.351) (0.363)
Number of observations 1,101 1,101 1,101 1,101

Notes: The source of data is the Securities and Exchange Commission website for the corporate governance data (gender, board size, board independence and duality of CEO and board chair) and Thomson ONE Banker for financial data (firm size and leverage). The regressions are estimated by using a dynamic panel data estimator as proposed by Blundell and Bond (1998). The variables are defined in Table 2. Robust standard errors are in parentheses. Asterisks indicate significance at 0.001 (***), 0.01 (**), and 0.05 (*) levels.

In other words, the dynamic panel models do not support the assertion that BGD has any effect on firm performance (measured by Tobin’s Q), regardless of the measure of BGD used in our specifications. Our results seem consistent with Liu et al. (2014), who find a positive and significant relationship at the 10% level with return on sales as a measure of performance, and a non-significant relationship with ROA as another measure of performance.

Quantile Regressions

In this section, we present the results of the QR estimations. Empirical investigation was conducted by estimating equation (2) for different values of θ (the 10th, 20th, […] and 90th quantiles). By doing so, we could examine the impact of the explanatory variables at different points of the conditional distribution. As explained previously, we first used the FE QR models to determine if BGD has a differential effect on firm financial performance depending on the portion of the firm distribution one is modeling. Table 9 presents the results using a FE quantile estimator presented at each quantile in the distribution.

Table 9.

Fixed Effects Quantile Regression.

Independent Variable Quantile Regressions
10th Quant 20th Quant 30th Quant 40th Quant 50th Quant 60th Quant 70th Quant 80th Quant 90th Quant
Panel A: DWOCB (dummy variable)
DWOCB 0.134 0.223 0.367 0.579* 0.718 0.925 1.216 0.564 −0.278
(0.119) (0.154) (0.217) (0.256) (0.370) (0.583) (0.972) (1.382) (2.134)
Firm size −0.219* −0.407*** −0.572** −0.838*** −1.027*** −1.03*** −1.333*** −1.697*** −2.038**
(0.097) (0.118) (0.182) (0.227) (0.247) (0.303) (0.397) (0.443) (0.632)
Leverage −0.009* −0.014** −0.017** −0.019** −0.023*** −0.024** −0.025* −0.028** −0.033**
(0.004) (0.005) (0.005) (0.006) (0.007) (0.008) (0.011) (0.010) (0.012)
Board size 0.019 0.024 0.024 0.032 0.026 −0.007 −0.013 −0.022 −0.006
(0.015) (0.019) (0.029) (0.035) (0.039) (0.048) (0.055) (0.057) (0.087)
Board independence 0.091 0.175 0.224 0.232 0.206 0.144 −0.068 −0.615 −1.262
(0.126) (0.162) (0.225) (0.261) (0.300) (0.376) (0.468) (0.523) (0.704)
Duality of CEO and board chair −0.051 −0.094 −0.041 −0.164 −0.253 −0.367 −0.520* −0.567 −0.315
(0.077) (0.094) (0.154) (0.192) (0.210) (0.242) (0.257) (0.308) (0.512)
Constant 1.719*** 2.677*** 3.481*** 4.771*** 6.007*** 6.673*** 8.494*** 11.87*** 15.321***
(0.423) (0.552) (0.843) (1.092) (1.288) (1.709) (2.200) (2.458) (3.144)
Panel B: PWOCB (percentage of WOCB)
PWOCB 0.541 0.717 1.372* 2.003** 2.066* 2.225 2.279 0.159 −1.078
(0.294) (0.396) (0.580) (0.692) (0.864) (1.151) (1.574) (2.039) (3.051)
Firm size −0.275** −0.429** −0.646*** −0.914*** 0.975*** −0.999** −1.296** −1.66** −1.935**
(0.100) (0.135) (0.194) (0.233) (0.260) (0.325) (0.440) (0.529) (0.690)
Leverage −0.009* −0.012* −0.016** −0.022** −0.021** −0.023* −0.022* −0.028* −0.037**
(0.004) (0.005) (0.006) (0.007) (0.008) (0.005) (0.011) (0.012) (0.013)
Board size 0.022 0.029 0.038 0.043 0.036 0.029 0.006 −0.022 −0.002
(0.013) (0.018) (0.028) (0.034) (0.041) (0.009) (0.064) (0.075) (0.107)
Board independence 0.176 0.164 0.106 0.104 0.235 0.014 −0.116 −0.604 −1.373*
(0.121) (0.151) (0.198) (0.240) (0.284) (0.051) (0.445) (0.510) (0.695)
Duality of CEO and board chair −0.042 −0.070 −0.059 −0.123 −0.211 0.311 −0.486 −0.481 −0.350
(0.074) (0.010) (0.156) (0.203) (0.226) (0.367) (0.317) (0.375) (0.600)
Constant 1.928*** 2.755*** 3.768*** 5.185*** 5.76*** 6.467*** 8.629*** 12.169*** 14.975***
(0.431) (0.592) (0.906) (1.131) (1.350) (1.734) (2.388) (2.700) (3.186)
Panel C: BWOCB (Blau’s index)
BWOCB 0.723 0.843 1.843* 2.391* 2.391* 2.766 2.076 −0.299 −2.019
(0.426) (0.589) (0.830) (1.021) (1.021) (1.436) (1.766) (2.245) (3.266)
Firm size −0.269** −0.428*** −0.632*** −0.876*** 2.549* −0.985** −1.261** −1.635** −2.135**
(0.103) (0.128) (0.189) (0.232) (1.241) (0.326) (0.438) (0.528) (0.730)
Leverage −0.009* −0.013* −0.016* −0.022** −0.95*** −0.024* −0.024* −0.028* −0.034**
(0.004) (0.005) (0.007) (0.008) (0.247) (0.009) (0.011) (0.012) (0.013)
Board size 0.024 0.030 0.037 0.043 0.037 0.016 0.013 −0.018 0.022
(0.013) (0.020) (0.029) (0.036) (0.045) (0.053) (0.067 (0.075) (0.104)
Board independence 0.177 0.178 0.136 0.132 0.324 0.289 −0.147 −0.609 −1.380
(0.114) (0.120) (0.181) (0.241) (0.230) (0.363) (0.472) (0.597) (0.785)
Duality of CEO and board chair −0.037 −0.077 −0.049 −0.154 −0.213 −0.253 −0.512 −0.522 −0.347
(0.072) (0.092) (0.152) (0.209) (0.245) (0.273) (0.321) (0.386) (0.589)
Constant 1.883*** 2.784*** 3.481*** 5.148*** 5.694*** 6.509*** 8.735*** 12.128*** 15.586***
(0.422) (0.542) (0.843) (1.101) (1.202) (1.640) (2.212) (2.600) (3.155)
Panel D: NWOCB (Number of WOCB)
NWOCB 0.112 0.062 0.148* 0.184* 0.196* 0.212 0.163 0.021 −0.135
(0.055) (0.049) (0.070) (0.082) (0.098) (0.114) (0.140) (0.172) 0.282)
Firm size −0.262* −0.428*** −0.613** −0.838*** −0.951*** −0.976*** 1.247** −1.647*** −1.924**
(0.106) (0.127) (0.187) (0.213) (0.232) (0.283) (0.384) (0.452) (0.612)
Leverage −0.008* −0.013** −0.016** −0.022*** −0.023** −0.023** −0.024* −0.028* −0.037**
(0.003) (0.004) (0.006) (0.006) (0.007) (0.009 (0.011) (0.011) (0.012)
Board size 0.015 0.021 0.015 0.020 0.012 −0.018 −0.027 −0.028 0.015
(0.013) (0.019) (0.028) (0.032) (0.037) (0.046) (0.058) (0.067) (0.106)
Board independence 0.176 0.196 0.135 0.135 0.264 0.295 −0.120 −0.596 −1.504
(0.121) (0.164) (0.228) (0.280) (0.324) (0.403) (0.523) (0.604) (0.784)
Duality of CEO and board chair 0.161 −0.070 −0.050 −0.155 −0.258 −0.242 −0.491 −0.478 −0.387
(0.125) (0.096) (0.155) (0.198) (0.213) (0.236) (0.272) (0.357) (0.641)
Constant 1.928*** 2.901*** 3.925*** 5.277*** 6.131*** 6.900*** 9.148*** 12.173*** 14.835***
(0.492) (0.589) (0.890) (1.067) (1.160) (1.480) (1.988) (2.173) (2.388)

Notes: The source of data is, on the one hand, the Securities and Exchange Commission website for the corporate governance data (gender, board size, board independence and duality of CEO and board chair) and, on the other hand, Thomson ONE Banker for financial data (firm size and leverage). The variables are defined in Table 2. Standard errors are in parentheses. Asterisks indicate significance at 0.001 (***), 0.01 (**), and 0.05 (*) levels.

As shown in Table 9, while the coefficients of any of our BGD measures are not significantly different from zero in the two lower 10th and 20th quantiles, we find that BGD is positively and significantly correlated with firm performance (at the 5% level) 9 in the three following upper quantiles of the 30th, 40th, and 50th. However, above the median the relationship between BGD and firm performance returns to being non-significant at conventional levels of significance. In other words, the FE QR models suggest that BGD matters only across the subset of the firm performance distribution that covers the 30th quantiles of Tobin’s Q up to the median but not at the high and low quantiles.

Overall, consistent with Solakoglu (2013), our results seem to indicate that BGD has a different effect on firm performance over the different points of the conditional distribution. However, our results contrast with the aforementioned study to the extent that the effect of WOCB is positive and significant (at the 5% level) between the 30th and 50th quantiles. On the other hand, Solakoglu (2013) find that BGD improves performance for average or above-average performing firms.

The results in Table 10 take into account possible endogeneity in the BGD variable through FE (Adams & Ferreira, 2009). In order to confirm the previous results, we use the IVQR of Kwak (2010) based on the method of Chernozhukov and Hansen (2004, 2008). Following Buchinsky (1995) recommendations, we used 1,000 bootstrap replications, and the standard deviation of the bootstrap distribution was used to calculate the respective standard errors. As a result, they are robust to heteroskedasticity and any general dependence between the explanatory variables and the error term.

Table 10.

Fixed Effects Panel with WOCB Instrumented by Industry.

Independent Variable Quantile Regressions
10th Quant 20th Quant 30th Quant 40th Quant 50th Quant 60th Quant 70th Quant 80th Quant 90th Quant
Panel A: DWOCB (dummy variable)
DWOCB 1.116* 1.074 1.706 2.495 2.54 2.909 2.873 −0.882 −6.376
(0.518) (0.626) (1.027) (1.428) (1.579) (1.886) (2.622) (3.730) (5.335)
Firm size −0.257*** −0.382*** −0.583*** −0.834*** −0.939*** −1.036*** −1.232*** −1.697*** −2.413***
(0.048) (0.057) (0.096) (0.121) (0.134) (0.135) (0.194) (0.212) (0.35)
Leverage −0.009*** −0.013*** −0.014*** −0.018*** −0.021*** −0.023*** −0.024*** −0.028*** −0.033***
(0.002) (0.002) (0.003) (0.004) (0.004) (0.004) (0.006) (0.005) (0.007)
Board size 0.027** 0.024** 0.026 0.035 0.016 −0.006 −0.038 −0.018 0.042
(0.009) (0.009) (0.015) (0.020) (0.019) (0.023) (0.036) (0.037) (0.059)
Board independence 0.181* 0.175* 0.231* 0.370** 0.351* 0.300 −0.092 −0.530 −1.186*
(0.074) (0.071) (0.116) (0.143) (0.159) (0.204) (0.325) (0.327) (0.497)
Duality of CEO and board chair −0.002 −0.081 −0.051 −0.163 −0.239* −0.283 −0.496** −0.489* −0.293
(0.056) (0.054) (0.089) (0.129) (0.114 (0.151) (0.182) (0.217) (0.300)
Constant 0.795 1.741** 2.194** 2.843* 3.919** 4.652** 6.856** 13.105*** 22.039***
(0.478) (0.558) (0.834) (1.207) (1.318) (1.602) (2.439) (3.938) (5.827)
Panel B: PWOCB (percentage of WOCB)
PWOCB 0.205 0.579 0.375 0.346 1.721 2.772 6.226 5.498 −0.849
(1.375) (1.654) (2.741) (3.527) (3.903) (4.513) (4.466) (5.159) (6.519)
Firm size −0.208*** −0.353*** −0.49*** −0.755*** −0.886*** −0.942*** −1.314*** −1.532*** −2.009***
(0.057) (0.066) (0.101) (0.108) (0.122) (0.141) (0.175) (0.224) (0.329)
Leverage −0.009*** −0.013*** −0.015*** −0.018*** −0.021*** −0.023*** −0.025*** −0.027*** −0.034***
(0.002) (0.002) (0.003) (0.004) (0.004) (0.004) (0.005) (0.005) (0.006)
Board size 0.019 0.025** 0.032 0.041* 0.024 0.013 −0.008 −0.019 −0.012
(0.010) (0.009) (0.017) (0.020) (0.020) (0.024) (0.030) (0.037) (0.056)
Board independence 0.115 0.177* 0.254* 0.300* 0.293 0.179 −0.041 −0.679* −1.222*
(0.077) (0.079) (0.125) (0.144) (0.159) (0.195) (0.269) (0.334) (0.555)
Duality of CEO and board chair −0.027 −0.087 −0.094 −0.243* −0.305* −0.309* −0.62*** −0.628** −0.252
(0.062) (0.059) (0.087) (0.116) (0.124) (0.146) (0.188) (0.211) (0.322)
Constant 1.719*** 2.475*** 3.222*** 4.771*** 5.609*** 6.196*** 8.138*** 10.339** 15.141***
(0.299) (0.350) (0.606) (0.810) (0.947) (1.002) (1.312) (1.895) (2.354)
Panel C: BWOCB (Blau’s index)
BWOCB 4.544*** 5.146*** 8.404*** 10.318*** 11.906*** 14.961*** 17.317*** 17.736*** 14.433
(0.948) (1.404) (1.539) (1.604) (1.771) (2.119) (2.360) (3.890) (8.372)
Firm size −0.279*** −0.456*** −0.693*** −0.864*** −0.981*** −1.179*** −1.192*** −1.285*** −1.687***
(0.054) (0.074) (0.076) (0.090) (0.111) (0.113) (0.165) (0.213) (0.297)
Leverage −0.009*** −0.013*** −0.014*** −0.018*** −0.022*** −0.023*** −0.021*** −0.024*** −0.034***
(0.002) (0.002) (0.003) (0.003) (0.003) (0.003) (0.005) (0.005) (0.006)
Board size 0.024* 0.032** 0.039* 0.038* 0.021 0.012 −0.038 −0.062* −0.047
(0.010) (0.012) (0.015) (0.017) (0.020) (0.021) (0.030) (0.031) (0.056)
Board independence 0.148 0.159 0.099 0.170 0.201 0.133 −0.434 −0.717* −1.696**
(0.080) (0.088) (0.127) (0.145) (0.153) (0.182) (0.259) (0.282) (0.573)
Duality of CEO and board chair 0.013 −0.048 −0.113 −0.169 −0.199 −0.288 −0.637*** −0.499** −0.483
(0.052) (0.062) (0.086) (0.102) (0.124) (0.149) (0.159) (0.194) (0.314)
Constant 1.259*** 2.173*** 2.989*** 3.831*** 4.663*** 5.631*** 6.909*** 8.178*** 12.116***
(0.222) (0.298) (0.440) (0.532) (0.580) (0.562) (0.827) (1.249) (2.037)
Panel D: NWOCB (Number of WOCB)
NWOCB 0.207* 0.237* 0.385* 0.547*** 0.675*** 0.932*** 1.215*** 1.022** 0.337
(0.084) (0.107) (0.152) (0.164) (0.180) (0.223) (0.235) (0.338) (0.638)
Firm size −0.275*** −0.443*** −0.712*** −0.916*** −1.023*** −1.202*** −1.252*** −1.375*** −1.875***
(0.052) (0.068) (0.100) (0.119) (0.135) (0.142) (0.190) (0.234) (0.342)
Leverage −0.009*** −0.013*** −0.015*** −0.018*** −0.021*** −0.024*** −0.021*** −0.025*** −0.034***
(0.002) (0.002) (0.003) (0.003) (0.003) (0.004) (0.005) (0.005) (0.006)
Board size 0.025** 0.028** 0.026 −0.018*** 0.013 −0.001 −0.050 −0.054 −0.027
(0.009) (0.010) (0.016) (0.003) (0.019) (0.023) (0.030) (0.036) (0.060)
Board independence 0.162* 0.179* 0.177 0.019 0.272 0.175 −0.359 −0.582 −1.394*
(0.069) (0.078) (0.128) (0.017) (0.161) (0.188) (0.300) (0.324) (0.582)
Duality of CEO and board chair −0.005 −0.101 −0.081 0.270 −0.228 −0.289* −0.683*** −0.529** −0.334
(0.056) (0.054) (0.081) (0.148) (0.125) (0.143) (0.177) (0.191) (0.300)
Constant 1.548*** 2.549*** 3.742*** 4.759*** 5.486*** 6.474*** 7.727*** 9.273*** 14.033***
(0.224) (0.258) (0.429) (0.543) (0.612) (0.620) (0.928) (1.380) (2.072)

Notes: The source of data is, on the one hand, the Securities and Exchange Commission website for the corporate governance data (gender, board size, board independence and duality of CEO and board chair) and, on the other hand, Thomson ONE Banker for financial data (firm size and leverage). The variables are defined in Table 2. Standard errors are in parentheses. Asterisks indicate significance at 0.001 (***), 0.01 (**), and 0.05 (*) levels.

Table 10 displays the IVQR results. As a reminder, we used Campbell (1996) industry classifications as our IV. In contrast with the previous results in Table 9, we achieved somewhat different results. First, among our four measures of BGD, two (DWOCB and PWOCB) were not statistically correlated with firm performance at the 5% level. 10 Second, BWOCB and NWOCB yielded some significant results. These two measures are positively and significantly different from zero at the first quantile (θ = 0.10) up to the 8th quantile (θ = 0.80) (at the 0.1% and the 5% levels and below, respectively). Furthermore, we notice that the size of the coefficients of DWOCB and PWOCB increase from the 1st quantile through to the 7th. Third, with regard to the control variables, we notice that the size of a firm is negatively and significantly correlated with firm performance (at the 5% level) in any model of Table 10. This result is consistent with Campbell and Mınguez-Vera (2008) and Liu et al. (2014). Similarly, firm leverage is negatively and significantly different from zero (at the 5% level). This result is line with Liu et al. (2014). These results seem to suggest that firm size and leverage have a negative effect on firm value.

Overall, our results based on IVQR tend to show that BGD has a varying impact on firm performance over the different points of the conditional distribution. As previously pointed out, the mere presence of WOCB has no significant impact on firm value (Campbell & Mınguez-Vera, 2008). The non-significance of PWOCB (percentage of WOCB) at any quantile seems to confirm the assertion of Harrison and Klein (2007) that Blau’s index is an optimal measure of diversity (compared to the proportion of WOCB) as it captures more variations within a group of people and is not skewed in a proportion of any one category (i.e., gender; see Bantel & Jackson, 1989). Furthermore, as noted by Nekhili and Gatfaoui (2013), in a practical manner, the percentage of WOCB does not reveal anything regarding the influence of female directors within corporate boards. Their number (NWOCB) is more significant. Our results seem to confirm this assertion. Finally, the significance and increasing effect of BGD suggest that WOCB contributes significantly to a firm’s performance at any point of the distribution.

Concluding Remarks

The purpose of this paper was to reassess the relationship between BGD and firm performance. Indeed, many studies in the past were conducted to test the impact of BGD on firm performance without been able to establish a clear correlation between the presence of women on boards and firm financial performance. We believe that the previous studies employed estimation methods that were inappropriate to depicting fully how BGD affects performance. First, as shown by Adams and Ferreira (2009), it is crucial to address the endogeneity of gender diversity in performance regressions. The lack of focus on the endogeneity problem causes errors in both the significance and the magnitude of the estimated relationship. However, in most of the previous studies, problems with reverse causality and unobserved firm heterogeneity had typically not been addressed. Second, in previous studies, linear regression or IVs were used to estimate the “average impact” of BGD on firm performance. If the effect of BGD is different across the conditional distribution of firm performance, the average effect might be 0, but this can hide differences between firms with different levels of performance. Therefore, the fact that some studies produced, on average, non-significant results does not mean that there is never a relationship between BGD and firm performance. Similarly, the fact that the average effect of BGD on firm performance is positive (or, alternatively, negative) does not mean that the relationship between BGD and firm performance is positive (or, alternatively, negative) for all levels of firm performance. For this paper, we decided to reassess the relationship between BGD and firm performance by using a methodology that would make it possible to study how BGD influences firm financial performance at different points of the conditional distribution and would take into account the possible endogeneity of the relationship between BGD and firm performance. By looking at the impact of BGD at different points of the conditional distribution, this study provides a better understanding of how the relationship between BGD and firm performance operates.

As expected, we show that the results are sensitive to the methodology used to estimate the relationship. While the results based on the fixed-effects method with robust standard errors clustered on the firm document no evidence of a significant relationship between Tobin’s Q and BGD, those based on the fixed-effects instrumental variables method with robust standard errors document a positive and significant relationship at the 5% level between three of the four WOCB measures used and firm performance. With regard to dynamic models (the Arellano-Bond and the Blundell-Bond dynamic panel models), the relationship between Tobin’s Q and BGD is not confirmed.

To examine the impact of BGD at different points of the conditional distribution, we used fixed-effects QR and showed that BGD matters only across a subset of the firm performance distribution that covers the 30th quantiles of Tobin’s Q up to the median. Moreover, when a fixed-effects panel with WOCB instrumented by industry QR is used, although two of our WOCB measures are not statistically correlated with firm performance at the 5% level, the two other WOCB measures are positively and significantly different from 0 from the 1st quantile up to the 8th.

Overall, our empirical results confirm that it is crucial to address the possible endogeneity in the BGD variable and that the significance and magnitude of BGD are different for different levels of firm performance. The mixed results obtained in previous studies can be explained, at least in part, by the failure to account for these two effects.

Theoretical Implications

The most important contribution of this paper is to demonstrate that the relationship between BGD and firm performance differs across the firm’s conditional performance distribution. This finding is significant because it suggests an explanation for the mixed results of previous studies. Our results indicate that the significance and magnitude of the relationship between BGD and firm performance depend on the performance of the firm. Since the performance of firms differs significantly from one study to another, previous empirical studies have not been able to establish a clear correlation between the presence of women on the board and firm financial performance.

Our results indicate that the relation between gender diversity and firm performance appears to be complex. The inclusion of WOCBs is particularly beneficial for poor and averagely performing firms. Indeed, these firms are the ones that can benefit the most from the presence of women on the board. In particular, diverse boards could bring a greater depth of perspective to the development of corporate strategy and increase the likelihood of creative and innovative problem-solving, which could ultimately raise the performance of these firms. Moreover, gender-diverse boards are generally seen as tougher monitors, since women frequently allocate more effort to monitoring corporate management and are more likely to hold CEOs accountable for poor stock performance (Adams & Ferreira, 2009). A better monitoring by female directors could be beneficial to poor and average performing firms.

In contrast, the most successful companies do not benefit from higher BGD. For high-performing firms, the benefits associated with the presence of female directors could be compensated for by integration and communication problems, which slow down decision-making processes. Moreover, firms that perform well require less monitoring, so the presence of female directors could cause over-monitoring and, therefore, have a negative impact on firm performance. These results are in the direction of those of Adams and Ferreira (2009), who show that BGD has a positive impact on firm performance in firms that otherwise have weak governance, whereas in firms with strong governance an increased presence of women has a negative impact.

Practical Implications

Since the relationship between BGD and firm performance differs across the firm’s conditional performance distribution, board regulations requiring increases in board diversity (e.g., board gender quotas) may be harmful or unnecessary for some firms. One of the main drawbacks associated with board regulations is that they do not recognize the specific needs of each firm and may prevent some of them from designing a board capable of responding to their unique corporate needs. Importantly, regardless of whether they make it possible to increase female participation on corporate boards quickly, these regulations, particularly board gender quotas, have been shown to produce negative consequences (Adams & Ferreira, 2009; Ahern & Dittmar, 2012; Matsa & Miller, 2013) that should not be ignored or overlooked in the name of political correctness and egalitarian idealism. Knowing that BGD could be of great benefit for some firms, but much less important for others, quotas could lead to non-optimal board composition.

However, our results clearly indicate that the vast majority of firms benefit from an increased representation of women on boards. In particular, for poor and averagely performing firms, we document a positive correlation between BGD and firm performance. It seems important, therefore, to continue to encourage these firms to increase female representation on their boards. Overall, our results bring grist to the mill for the proponents of the “business case of diversity” (Robinson & Dechant, 1997), such as Catalyst, that BGD may improve firm performance under certain conditions. The relationship is not as straightforward as Catalyst 11 seems to claim, however. We confirm their general assertion while highlighting nuances in the relationship.

Limitations and Future Research Directions

Our study is not without limitations. First, we were obliged to limit our sample to the 100 largest firms in the US because of the considerable amount of data that needed to be collected to complete the study. Although the constituents of the S&P 100 Index represent a large part of the market capitalization of the US equity markets, our results cannot be generalized to other kinds of organizations. Further studies are needed on small- and medium-sized firms, on entrepreneurial and private equity firms, and on public sector and non-profit organizations, since they are significantly different (e.g., in terms of capital structure or ownership).

Second, we limited our analysis of board diversity to BGD. Other forms of board diversity, such as ethnic, age, background, and experience, may have an impact on firm performance (Walt & Ingley, 2003). Thus, an analysis of how the other forms of board diversity factors influence firm performance at different points of the conditional distribution would offer some interesting lessons.

Third, we analyzed the impact of BGD on firm performance in the US context (English-origin countries; see La Porta, Lopez-de-Silanes, & Shleifer, 1999). According to Grosvold and Brammer (2011) and Terjesen and Singh (2008), the institutional, cultural, and political systems are important when analyzing BGD and its effect on firm performance. Hence, further studies would be useful and are necessary to confirm (or dispel) our results regarding the impact of BGD on firm performance over the different points of the conditional distribution.

Notes

2

Financial institutions, insurance companies, and real estate companies.

3

Electric, gas, and sanitary services.

4

Finkelstein (1992) defines “power” as the “capacity of individual actors to exert their will” (p. 506).

5

According to Hsiao (2003, p. 2), “a longitudinal, or panel, data set is one that follows a given sample of individuals over time, and thus provides multiple observations on each individual in the sample.”

6

See Kwak (2010) for more details regarding the “ivqreg” command.

7

The VIF provides an index that measures how much the variance of an estimated regression coefficient is increased because of collinearity.

8

Note that the figures would be the same if we had used the percentage of WOCB.

9

With the exception of DWOCB, which is not statistically different from 0 at conventional levels of significance.

10

DWOCB only yielded a single significant result at θ = 0.10.

11

Catalyst, Inc. is a non-profit organization that promotes inclusive workplaces for women; http://www.catalyst.org.

References

Adams, Almeida, & Ferreira (2005) Adams, R. B. , Almeida, H. , & Ferreira, D. (2005). Powerful CEOs and their impact on corporate performance. Review of Financial Studies, 18(4), 14031432.

Adams & Ferreira (2009) Adams, R. B. , & Ferreira, D. (2009). Women in the boardroom and their impact on governance and performance. Journal of Financial Economics, 94(2), 291309.

Adams, Hermalin, & Weisbach (2010) Adams, R. B. , Hermalin, B. E. , & Weisbach, M. S. (2010). The role of boards of directors in corporate governance: A conceptual framework and survey. Journal of Economic Literature, 48(1), 58107.

Agrawal & Knoeber (1996) Agrawal, A. , & Knoeber, C. R. (1996). Firm performance and mechanisms to control agency problems between managers and shareholders. Journal of Financial & Quantitative Analysis, 31(3), 377397.

Ahern & Dittmar (2012) Ahern, R. , & Dittmar, A. (2012). The changing of the boards: The impact on firm valuation of mandated female board representation. Quarterly Journal of Economics, 127(1), 137197.

Ammann, Oesch, & Schmid (2011) Ammann, M. , Oesch, D. , & Schmid, M. M. (2011). Corporate governance and firm value: International evidence. Journal of Empirical Finance, 18(1), 3655.

Arellano & Bond (1991) Arellano, M. , & Bond, S. (1991). Some tests of specification for panel data: Monte Carlo evidence and an application to employment equations. Review of Economic Studies, 58(2), 277297.

Arellano & Bover (1995) Arellano, M. , & Bover, O. (1995). Another look at the instrumental variable estimation of error-components models. Journal of Econometrics, 68(1), 2951.

Arrow (1951) Arrow, J. (1951). Social choice and individual values. New York, NY: Wiley.

Bantel & Jackson (1989) Bantel, A. , & Jackson, S. E. (1989). Top management and innovations in banking: Does the composition of the top team make a difference? Strategic Management Journal, 10(S1), 107124.

Bebchuk & Cohen (2005) Bebchuk, L. A. , & Cohen, A. (2005). The costs of entrenched boards. Journal of Financial Economics, 78(2), 409443.

Bhagat & Black (2002) Bhagat, S. , & Black, B. (2002). The non-correlation between board indpendence and long-term firm performance. Journal of Corporation Law, 27(2), 231273.

Bhagat & Bolton (2008) Bhagat, S. , & Bolton, B. (2008). Corporate governance and firm performance. Journal of Corporate Finance, 14(3), 257273.

Bilimoria & Huse (1997) Bilimoria, D. , & Huse, M. (1997). A qualitative comparison of the boardroom experiences of U.S. and Norwegian women corporate directors. International Review of Women and Leadership, 3(2), 6376.

Bilimoria & Wheeler (2000) Bilimoria, D. , & Wheeler, J. V. (2000). Women corporate directors: Current research and future directions. In M. J. Davidson & R. J. Burke (Eds.), Women in management: Current research issues. (Vol. II, pp. 138164). London: SAGE.

Blau (1977) Blau, P. M. (1977). Inequality and heterogeneity: A primitive theory of social structure. New York, NY: Collier Macmillan.

Blundell & Bond (1998) Blundell, R. , & Bond, S. (1998). Initial conditions and moment restrictions in dynamic panel data models. Journal of Econometrics, 87(1), 115143.

Bøhren & Strøm (2010) Bøhren, Ø. , & Strøm, R. Ø. (2010). Governance and politics: Regulating independence and diversity in the board room. Journal of Business Finance & Accounting, 37(9/10), 12811308.

Brammer, Millington, & Pavelin (2007) Brammer, S. , Millington, A. , & Pavelin, S. (2007). Gender and ethnic diversity among UK corporate boards. Corporate Governance: An International Review, 15(2), 393403.

Brammer, Millington, & Pavelin (2009) Brammer, S. , Millington, A. , & Pavelin, S. (2009). Corporate reputation and women on the board. British Journal of Management, 20(1), 1729.

Brickley, Coles, & Jarrell (1997) Brickley, J. A. , Coles, J. L. , & Jarrell, G. (1997). Leadership structure: Separating the CEO and chairman of the board. Journal of Corporate Finance, 3(3), 189220.

Brickley, Coles, & Terry (1994) Brickley, J. A. , Coles, J. L. , & Terry, R. L. (1994). Outside directors and the adoption of poison pills. Journal of Financial Economics, 35(3), 371390.

Buchinsky (1995) Buchinsky, M. (1995). Estimating the asymptotic covariance matrix for quantile regression models: A Monte Carlo study. Journal of Econometrics, 68(2), 303338.

Buchinsky (1998) Buchinsky, M. (1998). Recent advances in quantile regression models. Journal of Human Resources, 33(1), 88126.

Byrd & Hickman (1992) Byrd, J. W. , & Hickman, A. (1992). Do outside directors monitor managers? Evidence from tender offer bids. Journal of Financial Economics, 32(2), 195221.

Campbell (1996) Campbell, J. Y. (1996). Understanding risk and return. Journal of Political Economy, 104(2), 298345.

Campbell & Mınguez-Vera (2008) Campbell, K. , & Mınguez-Vera, A. (2008). Gender diversity in the boardroom and firm financial performance. Journal of Business Ethics, 83(3), 435451.

Campbell & Minguez-Vera (2010) Campbell, K. , & Minguez-Vera, A. (2010). Female board appointments and firm valuation: Short and long-term effects. Journal of Management and Governance, 14(1), 3759.

Carter, D’Souza, Simkins, & Simpson (2010) Carter, D. A. , D’Souza, F. , Simkins, B. J. , & Simpson, W. G. (2010). The gender and ethnic diversity of US boards and board committees and firm financial performance. Corporate Governance: An International Review, 18(5), 396414.

Carter, Simkins, & Simpson (2003) Carter, D. A. , Simkins, B. J. , & Simpson, W. G. (2003). Corporate governance, board diversity, and firm value. Financial Review, 38(1), 3353.

Chatterjee & Hadi (2012) Chatterjee, S. , & Hadi, A. S. (2012). Regression analysis by example. Hoboken, NJ: Wiley-Interscience.

Chernozhukov & Hansen (2004) Chernozhukov, V. , & Hansen, C. (2004). The effects of 401(k) participation on wealth distribution: An instrumental quantile regression analysis. Review of Economics & Statistics, 86(3), 735751.

Chernozhukov & Hansen (2005) Chernozhukov, V. , & Hansen, C. (2005). An IV model of quantile treatment effects. Econometrica, 73(1), 245261.

Chernozhukov & Hansen (2006) Chernozhukov, V. , & Hansen, C. (2006). Instrumental quantile regression inference for structural and treatment effect models. Journal of Econometrics, 132(2), 491525.

Chernozhukov & Hansen (2008) Chernozhukov, V. , & Hansen, C. (2008). Instrumental variable quantile regression : A robust inference approach. Journal of Econometrics, 142(1), 379398.

Chhaochharia & Grinstein (2007) Chhaochharia, V. , & Grinstein, Y. (2007). The changing structure of US corporate boards: 1997–2003. Corporate Governance: An International Review, 15(6), 12151223.

Chung & Pruitt (1994) Chung, H. , & Pruitt, S. W. (1994). A simple approximation of Tobin’s q. FInancial Management, 23(3), 7074.

Coles, Daniel, & Naveen (2008) Coles, J. L. , Daniel, N. D. , & Naveen, L. (2008). Boards: Does one size fit all? Journal of Financial Economics, 87(2), 329356.

Cornett, Marcus, & Tehranian (2008) Cornett, M. M. , Marcus, A. J. , & Tehranian, H. (2008). Corporate governance and pay-for-performance: The impact of earnings management. Journal of Financial Economics, 87(2), 357373.

Daily, Certo, & Dalton (1999) Daily, C. M. , Certo, S. T. , & Dalton, D. R. (1999). A decade of corporate women: Some progress in the boardroom, none in the executive suite. Strategic Management Journal, 20(1), 9399.

Dale-Olsen, Schøne, & Verner (2013) Dale-Olsen, H. , Schøne, P. , & Verner, M. (2013). Diversity among Norwegian boards of directors: Does a quota for women improve firm performance? Feminist Economics, 19(4), 110135.

Dalton, Daily, Ellstrand, & Johnson (1998) Dalton, D. R. , Daily, C. M. , Ellstrand, A. E. , & Johnson, J. L. (1998). Meta-analytic reviews of board composition, leadership structure and financial performance. Strategic Management Journal, 19(3), 269290.

Dang & Vo (2012) Dang, R. , & Vo, L.-C. (2012). Women on corporate boards of directors: Theories, facts and analysis. In D. Nguyen & S. Boubaker (Eds.), Board directors and corporate social responsibility (pp. 321). Basingstoke: Palgrave Macmillan.

Eisenberg, Sundgren, & Wells (1998) Eisenberg, T. , Sundgren, S. , & Wells, M. T. (1998). Larger board size and decreasing firm value in small firms. Journal of Financial Economics, 48(1), 3554.

Erhardt, Werbel, & Shrader (2003) Erhardt, N. L. , Werbel, J. D. , & Shrader, C. B. (2003). Board of director diversity and firm financial performance. Corporate Governance: An International Review, 11(2), 102111.

Faleye (2007) Faleye, O. (2007). Classified boards, firm value, and managerial entrenchment. Journal of Financial Economics, 83(2), 501529.

Fama & French (1992) Fama, E. F. , & French, R. (1992). The cross-section of expected stock returns. Journal of Finance, 47(2), 427465.

Farrell & Hersch (2005) Farrell, A. , & Hersch, P. L. (2005). Additions to corporate boards: The effect of gender. Journal of Corporate Finance, 11(1/2), 85106.

Ferreira (2010) Ferreira, D. (2010). Board diversity. In H. Baker & R. T. Anderson (Eds.), Corporate governance: A synthesis of theory, research, and practice (pp. 225242). Hoboken, NJ: Wiley.

Finkelstein (1992) Finkelstein, S. (1992). Power in top management teams: Dimensions, measurement, and validation. Academy of Management Journal, 35(3), 505538.

Forbes & Milliken (1999) Forbes, D. P. , & Milliken, F. J. (1999). Cognition and corporate governance: Understanding boards of directors as strategic decision-making groups. Academy of Management Review, 24(3), 489505.

Francoeur, Labelle, & Sinclair-Desgagné (2008) Francoeur, C. , Labelle, R. , & Sinclair-Desgagné, B. (2008). Gender diversity in corporate governance and top management. Journal of Business Ethics, 81(1), 8395.

George & Longstaff (1993) George, T. J. , & Longstaff, F. A. (1993). Bid-ask spreads and trading activity in the S&P 100 index options market. Journal of Financial and Quantitative Analysis, 28(3), 381397.

Goyal & Park (2002) Goyal, V. K. , & Park, C. W. (2002). Board leadership structure and CEO turnover. Journal of Corporate Finance, 8(1), 4966.

Grosvold & Brammer (2011) Grosvold, J. , & Brammer, S. (2011). National institutional systems as antecedents of female board representation: An empirical study. Corporate Governance: An International Review, 19(2), 116135.

Gujarati & Porter (2009) Gujarati, D. N. , & Porter, D. C. (2009). Basic econometrics. Boston, MA: McGraw-Hill/Irwin.

Harrison & Klein (2007) Harrison, D. A. , & Klein, J. (2007). What’s the difference ? Diversity constructs as separation, variety or disparity in organizations. Academy of Management Review, 32(4), 11991228.

Hermalin & Weisbach (1988) Hermalin, B. E. , & Weisbach, M. S. (1988). The determinants of board composition. RAND Journal of Economics, 19(4), 589606.

Hermalin & Weisbach (1991) Hermalin, B. E. , & Weisbach, M. S. (1991). The effects of board composition and direct incentives on firm performance. Financial Management, 20(4), 101112.

Hermalin & Weisbach (2003) Hermalin, B. E. , & Weisbach, M. S. (2003). Boards of directors as an endonenously determined institution: A survey of the economic literature. Economic Policy Review, 9(1), 720.

Hillman, Shropshire, & Cannella (2007) Hillman, A. J. , Shropshire, C. , & Cannella, A. A. J. (2007). Organizational predictors of women on corporate boards. Academy of Management Journal, 50(4), 941952.

Hirsch & Seaks (1993) Hirsch, B. T. , & Seaks, T. G. (1993). Functional form in regression models of Tobin’s q. Review of Economics and Statistics, 75(2), 381385.

Hsiao (2003) Hsiao, C. (2003). Analysis of panel data. Cambridge: Cambridge University Press.

Jensen (1986) Jensen, M. C. (1986). Agency costs of free cash flow, corporate finance, and takeovers. American Economic Review, 76(2), 323329.

Jensen & Meckling (1976) Jensen, M. C. , & Meckling, W. H. (1976). Theory of the firm: Managerial behavior, agency costs and ownership structure. Journal of Financial Economics, 3(4), 305360.

Johnston & DiNardo (2007) Johnston, J. , & DiNardo, J. E. (2007). Econometric methods. (4th ed.). New York, NY: The Mc-Graw-Hill Companies.

Kochan et al. (2003) Kochan, T. , Bezrukova, K. , Ely, R. , Jackson, S. , Joshi, A. , Jehn, K. , […] Thomas, D. (2003). The effects of diversity on business performance: Report of the diversity research network. Human Resource Management, 42(1), 321.

Koenker (2005) Koenker, R. (2005). Quantile regression. Cambridge: Cambridge University Press.

Koenker & Bassett (1978) Koenker, R. , & Bassett, G. Jr . (1978). Regression quantiles. Econometrica, 46(1), 3350.

Koenker & Hallock (2001) Koenker, R. , & Hallock, F. (2001). Quantile regression. Journal of Economic Perspectives, 15(4), 143156.

Kuan, Li, & Liu (2012) Kuan, T.-H. , Li, C.-S. , & Liu, C.-C. (2012). Corporate governance and cash holdings: A quantile regression approach. International Review of Economics and Finance, 24, 303314.

Kwak (2010) Kwak, D. W. (2010). Implementation of instrumental variable quantile regression (IVQR) methods. Working Paper, Michigan State University.

La Porta, Lopez-de-Silanes, & Shleifer (1999) La Porta, R. , Lopez-de-Silanes, F. , & Shleifer, A. (1999). Corporate ownership around the world. Journal of Finance, 54(2), 471517.

Lau & Murnighan (1998) Lau, D. C. , & Murnighan, J. (1998). Demographic diversity and faultlines: The compositional dynamics of organizational groups. Academy of Management Review, 23(2), 325340.

Lee (2002) Lee, M.-J. (2002). Panel data econometrics: Methods-of-moments and limited dependent variables. San Diego, CA: Academic Press.

Lindenberg & Ross (1981) Lindenberg, E. B. , & Ross, S. A. (1981). Tobin’s q ratio and industrial organization. Journal of Business, 54(1), 132.

Liu, Wei, & Xie (2014) Liu, Y. , Wei, Z. , & Xie, F. (2014). Do women directors improve firm performance in China? Journal of Corporate Finance, 28, 169184.

Mahadeo, Soobaroyen, & Hanuman (2012) Mahadeo, J. , Soobaroyen, T. , & Hanuman, V. (2012). Board composition and financial performance: Uncovering the effects of diversity in an emerging economy. Journal of Business Ethics, 105(3), 375388.

Margaritis & Psillaki (2010) Margaritis, D. , & Psillaki, M. (2010). Capitalstructure, equity ownership and firm performance. Journal of Banking and Finance, 34(3), 621632.

Mata & Machado (1996) Mata, J. , & Machado, J. A. F. (1996). Firm start-up size: A conditional quantile approach. European Economic Review, 40(6), 13051323.

Matsa & Miller (2013) Matsa, D. A. , & Miller, A. R. (2013). A female style in corporate leadership? Evidence from quotas. American Economic Journal: Applied Economics, 5(3), 136169.

Mela & Kopalle (2002) Mela, C. F. , & Kopalle, P. (2002). The impact of collinearity on regression analysis: The asymmetric effect of negative and positive correlations. Applied Economics, 34(6), 667677.

Miller, Burke, & Glick (1998) Miller, C. C. , Burke, L. M. , & Glick, W. H. (1998). Cognitive diversity among upper-echelon executives: Implications for strategic decision processes. Strategic Management Journal, 19(1), 3958.

Miller & del Carmen Triana (2009) Miller, T. , & del Carmen Triana, M. (2009). Demographic diversity in the boardroom: Mediators of the board diversity-firm performance relationship. Journal of Management Studies, 46(5), 755786.

Nekhili & Gatfaoui (2013) Nekhili, M. , & Gatfaoui, H. (2013). Are demographic attributes and firm characteristics drivers of gender diversity? Investigating women’s positions on French boards of directors. Journal of Business Ethics, 118(2), 227249.

Pearce & Zahra (1992) Pearce, J. A. , & Zahra, S. A. (1992). Board composition from a strategic contingency perspective. Journal of Management Studies, 29(4), 411438.

Perfect & Wiles (1994) Perfect, S. B. , & Wiles . (1994). Alternative constructions of Tobin’s q: An empirial comparison. Journal of Empirical Finance, 1(3–4), 313341.

Peterson & Philpot (2007) Peterson, C. A. , & Philpot, J. (2007). Women’s roles on U.S. Fortune 500 boards: Director expertise and committee memberships. Journal of Business Ethics, 72(2), 177196.

Pfeffer (1972) Pfeffer, J. (1972). Size and composition of corporate boards: The organization and its environment. Administrative Science Quarterly, 17(2), 218228.

Pfeffer & Salancik (1978) Pfeffer, J. , & Salancik, G. R. (1978). The external control of organizations: A resource dependence perspective. New York, NY: Harper & Row.

Powell (1999) Powell, G. N. (1999). Introduction: Examining the gender and work. In G. N. Powell (Ed.), Handbook of gender & work (pp. ixxx). Thousand Oaks, CA: Sage Publications.

Robinson & Dechant (1997) Robinson, G. , & Dechant . (1997). Building a business case for diversity. Academy of Management Executive, 11(3), 2131.

Rose (2007) Rose, C. (2007). Does female board representation influence firm performance? The Danish evidence. Corporate Governance: An International Review, 15(2), 404413.

Shrader & Blackburn (1997) Shrader, C. B. , & Blackburn, V. B. (1997). Women in management and firm financial performance: An exploratory study. Journal of Managerial Issues, 9(3), 355372.

Smith, Smith, & Verner (2006) Smith, N. , Smith, V. , & Verner, M. (2006). Do women in top management affect firm performance? A panel study of 2,500 Danish firms. International Journal of Productivity & Performance Management, 55(7), 569593.

Solakoglu (2013) Solakoglu, N. (2013). The role of gender diversity on firm performance: A regression quantile approach. Applied Economics Letters, 20(17), 15621566.

Terjesen, Sealy, & Singh (2009) Terjesen, S. , Sealy, R. , & Singh, V. (2009). Women directors on corporate boards: A review and research agenda. Corporate Governance: An International Review, 17(3), 320337.

Terjesen & Singh (2008) Terjesen, S. , & Singh, V. (2008). Female presence on corporate boards: A multi-country study of environmental context. Journal of Business Ethics, 83(1), 5563.

Walt & Ingley (2003) Walt, N. V. D. , & Ingley, C. (2003). Board dynamics and the influence of professional background, gender and ethnic diversity of directors. Corporate Governance: An International Review, 11(3), 218234.

Watson, Kumar, & Michaelsen (1993) Watson, W. E. , Kumar, K. , & Michaelsen, L. (1993). Cultural diversity’s impact on interaction process and performance: Comparing homogeneous and diverse task groups. Strategic Management Journal, 36(3), 590602.

Wiersema & Bantel (1992) Wiersema, M. F. , & Bantel, A. (1992). Top management team demography and corporate strategic change. Academy of Management Journal, 35(1), 91121.

Williams & O’Reilly (1998) Williams, Y. , & O’Reilly, C. A. (1998). Demography and diversity in organizations: A review of 40 years of research. In B. M. Staw & S. M. Sutton (Eds.), Research in organizational behavior (Vol. 20, pp. 77140). Stamford, CT: JAI Press.

Wintoki, Linck, & Netter (2012) Wintoki, M. B. , Linck, J. S. , & Netter, J. M. (2012). Endogeneity and the dynamics of internal corporate governance. Journal of Financial Economics, 105(3), 581606.

Wooldridge (2013) Wooldridge, J. M. (2013). Introductory econometrics: A modern approach. Mason, IA: South-Western, Cengage Learning.

Yermack (1996) Yermack, D. (1996). Higher market valuation of companies with a small board of directors. Journal of Financial Economics, 40(2), 185211.