Dutch board diversity and firm performance Rick Homan

Dutch board diversity and firm performance Rick Homan1

University of Groningen, Faculty of Economics and Business, Groningen, The Netherlands

Abstract

This study extends previous research on the effects of executive board diversity by examining the relationship between age-, gender- and nationality diversity on firm performance in the Netherlands. Based on a sample of 79 Dutch listed firms studied over the period 2010-2015, this study reports a positive link between board diversity and firm performance. Firm performance is, hereby, estimated using a forward-looking market performance measure (Tobin’s Q) and a backward-looking accounting measure (ROA).

Keywords: Corporate governance, executive board diversity, firm performance, gender, age, nationality

JEL classification: G34, J16, J15, and L25.

“There are not more than five musical notes, yet the combinations of these five give rise to more melodies than can ever be heard.

There are not more than five primary colours, yet in combination they produce more hues than can ever be seen.

There are not more than five cardinal tastes, yet combinations of them yield more flavours than can ever be tasted.”

― Sun Tzu (544bc - 496bc), The Art of War

1. Introduction

In today’s business corporations, employees and top management not only become increasingly diverse in terms of gender, age, and nationality, but also in terms of tenure, experience, educational background and socioeconomic status (Nielsen and Nielsen, 2012; Ruigrok, Peck, and Tacheva, 2007). A reason for this increased importance of diversification can be found in the area of corporate governance, stressing the importance of gender diversity (Adams and Ferreira, 2009). Nowadays, stakeholders such as shareholders, investors and inquisitive readers will almost certainly find the heading ‘corporate governance’ in every annual report. According to the Organisation for Economic Co-operation and Development (OECD, 2015), corporate governance “involves a set of relationships between a company’s management, its board, its shareholders and other stakeholders.(..)”. Political, social and business entities are increasingly calling for demographically diversified boards (Van Veen and Marsman, 2008). This caused developed countries like the United States, Australia and many European countries to take measures in order to influence corporate governance, focusing mainly on increasing diversity of (e.g.) the board (Adams and Kirchmaier, 2016; Adams and Ferreira, 2009). One prime example is the implementation of gender quota laws for boards by governments in order to change the level of diversity (Adams and Kirchmaier, 2016; Terjesen, Aguilera, and Lorenz, 2015). In 2003, Norway became the first country in the world to introduce a gender quota that at least 40% of the company board be composed of women (Ahern and Dittmar, 2012). Recently , Germany imposed a new law that requires women to hold 30% of the top board seats as of January 2016. The Netherlands has a introduced a ‘target quota’ of 30%. This quota has no legal basis, but the Dutch government plays an active role in achieving this aim.

board composition. A longitudinal study by Staples (2007) shows that boards of multinational firms are becoming increasingly diversified and that the social forces spurring this trend are unlikely to disappear anytime soon. There is a growing preference for female or ethnic minority board directors (Netter, Poulsen and Stegemoller, 2009). As director characteristics could affect the effectiveness and their influence in boards, it is likely that differences in nationality and gender will affect firm performance. Therefore, studying diversity can help understand the effects of group composition on board effectiveness and firm performance (Adams, Hermalin, and Weisbach, 2010).

The subject of board diversity has attracted researchers from various disciplines who have tried to link diversity with other aspects of a firm. Some recent examples are board diversity and its impact on corporate governance (Adams and Ferreira, 2009), organizational innovation (Chen, Leung, and Evans, 2015) and corporate social responsibility (Harjoto, Laksmana, and Lee, 2015). Gonzalez and Hagendorff (2016) researched the diversities effect on corporate risk-taking. An increasing amount of studies in the field of corporate governance and finance literature researched the relationship between board diversity on firm performance. One of the first studies in this area was done by Carter, Simkins, and Simpson (2003). They find that the proportion of women and minorities on US boards has a positive relationship with firm value. Giannetti and Zhoa (2016) did a similar study covering US firms in 2001-2011. They find that more diverse boards have greater stock return and volatility.

There is limited research on this subject regarding board diversity in Dutch boards. Using empirical data on 186 listed firms in 2007, Marinova, Plantenga, and Remery (2016) find no relationship between board gender diversity and firm performance. Their study suggests that there is a need for more empirical research due to the limited European evidence on the effect of board diversity on firm performance. Because other studies did find a relationship between board gender diversity and firm performance, it would be a cause for more research, taking into account the limitation of previous studies (see, e.g., Vafaei, Amed, and Mather, 2015; Boubaker, Dang, and Nguyen, 2014; Liu, Wei, and Xie, 2014). Marinova et al. (2016) use a market-based performance indicator (Tobin’s Q) and suggest that future studies should also use an accounting-based measure. In addition, the use of panel data is preferred. Other research by Estélyi and Nisar (2016) argues that there is a need for board diversity study, especially when it comes to nationality diversity among boards.

of 30% women on boards in January 2013. However, the effects of gender diversity in firms are not yet been fully studied and conclusions are mixed (Marinova et al., 2016). More research into these effects is therefore of interest to policy makers in order to predict the results of their policy.

The results represented in this paper indicate a positive relationship between age, gender and nationality diversity and firm performance. However, these results were found not significant and are potentially suffering from reverse causality. Lastly, it could not be proven that age, gender and nationality have a joint significant impact on firm performance.

2. Literature review

This section examines the relevant theories and empirical evidence that explain how board diversity can affect firm performance. Based on this three hypotheses are formulated.

2.1 Corporate governance

The performance of a firm is significantly impacted by corporate governance (Al-Matari, Al-Swidi and Fadzil, 2014). Corporate governance and the role of a company’s board of directors is of utmost importance in economics (Adams et al., 2010). The role of the board of directors is to monitor the activities of a firm (agency perspective) and set forth the corporate strategy that influences firm performance (Adams et al., 2010; Petrovic, 2008). Postma and Van Ees (2000) also note the boards have an interlocking function (resource dependency perspective). The subtasks will vary per firm and are allocated across several directors, depending on the board system. Two board types can be depicted: one-tier (unitary) and two-tier (dual) boards. One-tier systems are more common in Anglo-Saxon countries whereas two-tier systems are more common in continental Europe (Dehaene, De Vuyst, Ooghe, 2001). Two-tier boards have executive directors (management board) that are responsible for the daily operations of the company, and non-executive directors (supervisory) who supervise the executive directors (Jungmann, 2006).

2.2 Theories on board diversity

Upper echelons theory

The upper echelons theory suggests that top executive directors’ experiences, values, and personalities affect their choices and subsequently organizational performance. This theory is based on the bounded rationality concept, arguing that due to limited information, biased perception and filtering, individuals do not act objectively in decision-making decisions (Hambrick, 2007, Hambrick & Mason, 1984). The board of directors is viewed as an important decision-making body, and its composition has an impact on firm performance since decision-making is jointly determined by individual board members (Carpenter, Geletkanycz, and Sanders, 2014; Finkelstein, Hambrick, and Cannella, 2009; Finkelstein & Hambrick, 1996). Diversity between these individual board members refers to observable (demographic) and less visible (cognitive) characteristics (Mahadeo, Soobaroyen, and Hanuman, 2012; Marimuthu, 2008). Board diversity is generally investigated under the assumption that analysing these characteristics can help to understand the effect of board diversity on performance (Adams et al., 2010). Recent research studying the role of board diversity and top management teams used the upper echelons perspective in combination with several alternative theories to find the answer to the question if board diversity contributes to firm strategy and firm performance since no single theory is able to explain the correlation between board diversity and firm performance. (Nielsen, 2010; Lynall, Golden, and Hillman, 2003).

Agency theory

Resource dependency theory

Among others, Hillman, Cannella, & Patzold (2000) combined the upper echelons theory with the resource dependency theory to reason how board composition influences firm performance. The resource dependence theory asserts that boards provide firms benefits through linkages to external organisations, providing firms access to resources that would not be available otherwise (Pugliese, Minichilli, and Zattoni, 2014). Pfeffer and Salancik (1978) point to four benefits of these linkages: (1) providing information resources and expertise, (2) provision of communication channels between the firm and network that are important to the firm, (3) access to support from outside organisations (monetary or reputational commitments), and (4) legitimacy for firms. In short, the theory reasons that by selecting directors with diverse backgrounds and various characteristics, a firm is better equipped to have access to these four benefits, resulting is higher firm performance (Hillman et al., 2000).

Hillman and Dalziel (2003), argue that the theory is useful when applied to characteristics such as experience, and social- and political connections but is limited when trying to understand the role of other demographic characteristics of the board such as gender or age. Although the resource dependency is used by many scholars because of its importance in explaining the behaviour of firms with regard to its external environment, recent research scrutinises the theory because it cannot be used to directly explain a firm’s performance (Sharif & Yeoh, 2014; Drees and Heugens, 2013; Davis and Cobb, 2010; Hillman, Withers, and Collins, 2009).

Human capital

2.3 Age diversity

One aspect of this study is board age diversity, a phenomenon of which the effect on firm performance is not yet fully understood (Sila, Gonzalez, and Hagendorff, 2016; Kunce, Boehm, and Bruch, 2013). Early research argues that people of different ages possess unique types of knowledge and cognitive abilities (Horn & Cattell, 1967). Young individuals often have more knowledge of recent technologies and a greater mental capacity (i.e. intellectual speed and efficiency in learning) to adapt to new situations (Bugg, Zook, DeLosh, Davalos, & Davis, 2006; Skirbekk, 2004; Horn & Cattell, 1967). However, when people get older they acquire mental skills that are assumed to improve with experience and learning. Examples are in-depth knowledge and better problem-solving skills (Grund and Westergaard-Nielsen, 2008; Skirbekk, 2004). The human capital theory is used to explain that older board members, having developed more knowledge, skills and experiences, are useful to the organisation (Shore et al., 2009). Age diversity can be categorized into two different types of age-related differences: (1) differences in values and (2) knowledge. Values differ between people of different ages. Historical experiences, training and interacting with other individuals, affect the values and the way of thinking about work and life (Parry and Urwin, 2011; Wagner, 2007). As people get older they find themselves in different stages of their lives (e.g. young professional versus older people with grandchildren) with corresponding values. For example, young people tend to ‘work to live’, spending more time with friends and value social interaction (Sullivan, Forret, Carraher, and Mainiero, 2009). While older people value the belief of hard work, and are said to “live to work”.

Age-related knowledge is shared, combined and integrated within groups and potentially improve the quality of decision-making, creativity, efficiency, problem-solving, and finally labour productivity (Carton & Cummings, 2012; Harrison & Klein, 2007; Horwitz & Horwitz, 2007). According to Grund and Westergaard-Nielsen (2008), firms can benefit from synergies when age-based knowledge and competencies are mixed within a firm. However, less effective communication, cooperation and cohesion within groups with varying ages might lead to conflicts (Bell, Villado, Lukasik, Belau, and Briggs, 2011). In turn, potential synergies are not likely to occur and age diversity might even negatively impact performance (Kunce, Boehm, and Bruch, 2013; Klein, Knight, Ziegert, Lim & Saltz, 2011; Grund & Westergaard-Nielsen, 2008).

Empirical evidence

relationship between age diversity and firm performance in the Netherlands. This relationship turned out to be hyperbolic: Age diversity does increase firm performance but up to a certain point. After this point, an increase in age diversity will have a negative impact on firm performance. However, Engelen et al. (2012) focused on whether board age diversity matters during times of crisis. According to Francis, Hasan & Wu (2012) director age is positively related to firm performance during crises, because experience might be a valuable resource in times of crises. Contrasting results are found by Waelchli and Zeller (2013) consistent with the agency theory. On average the board of directors impose their own life cycle on the firms they lead causing a negative effect on firm performance. A similar research by Taljaard, Ward and Muller (2015) find strong associations of board age diversity with decreased financial performance. Firms did not benefit from increased or accumulated knowledge and experience as assumed by the human capital theory.

Hypothesis 1

Hypothesis 1 is constructed based on the study of Ferrero-Ferrero et al. (2015) and the theories introduced in the previous section. It is expected that increased board diversity has a positive impact on firm performance because age comes with more knowledge and experience as suggested by the human capital theory and older board of directors are assumed to have increased linkages (resource dependency). The following hypothesis was formulated:

𝐻1: 𝐵𝑜𝑎𝑟𝑑 𝑎𝑔𝑒 𝑑𝑖𝑣𝑒𝑟𝑠𝑖𝑡𝑦 ℎ𝑎𝑠 𝑎 𝑝𝑜𝑠𝑖𝑡𝑖𝑣𝑒 𝑒𝑓𝑓𝑒𝑐𝑡 𝑜𝑛 𝑎 𝑓𝑖𝑟𝑚′𝑠 𝑓𝑖𝑛𝑎𝑛𝑐𝑖𝑎𝑙 𝑝𝑒𝑟𝑓𝑜𝑟𝑚𝑎𝑛𝑐𝑒

2.4 Gender diversity

Females are also said to value their responsibilities as directors more, corresponding with an increase in effective corporate governance. By appointing more women to the board of a company they serve as a linkage instrument, using the resource dependency theory as a basis, that can provide advantages to firms due to improved linking with their stakeholders (Hillman, Shropshire, and Cannella, 2007). Diversity is in general positive for organizations by providing wider and better connections with stakeholders, in turn lowering market uncertainty and dependency (Miller & del Carmen, 2009). Carter et al. (2003), argue that female board members are more inclined to ask questions that would not be asked by their male counterpart. In addition, they are not intertwined in ‘old boys’ networks (Adams & Ferreira, 2009; Campbell and Minguez-Vera, 2008). That is why evidence suggests that women are better monitors and are more able to alleviate value-decreasing stakeholder conflicts (Post and Byron, 2015; Adams & Ferreira, 2009).

Empirical evidence

In the past, studies focussing on gender diversity and the effect on firm performance turn out to be positive, negative or nonsignificant (Webber & Donahue, 2001). Marinova et al. (2016) researched the effect of gender diversity on firm performance in the Netherlands and Denmark. Their results show no relation between the share/presence of women on boards and firm performance2_{. Rose, Munch-Madsen,} and Funch (2013) researched the impact of female board representation on corporate performance on a sample in Nordic countries as well as Germany. Their study was similar to this study in the sense that they used the same theoretical foundation. No support was found for any performance impact due to changes in female board representation. Similarly, Engelen et al. (2012) based their research on the human capital theory and found no effect of board gender diversity on firm performance during times of crisis. In their research on women directors on boards, Terjesen et al. (2009) note that the agency-, resource dependency- and human capital theories are not mutually exclusive and all have a different impact on board diversity and firm performance.

Adams and Ferreira (2009) find that women are found to be more active in their monitoring role within the boardroom and therefore provide increased firm performance. This corresponds to what is expected by the agency theory and is also empirically supported by Taljaard et al. (2015) and Carter et al. (2003). However, Adams and Ferreira (2009) also find that more gender diverse boards can suffer from over-monitoring when shareholder protection is strong, causing a negative effect on firm performance. Opposite results are found when shareholders are weakly protected. Proposing that gender diversity is positive depending on the level of shareholder’s protection (Ujunwa, Okoyeuzu, and Nwakoby, 2012; Adams and Ferreira, 2009; Okike, 2007).

Hypothesis 2

Although the literature is ambiguous on the point whether gender diversity has an effect on firm performance, this study draws the hypothesis based on the literature by (i.a.) Adams and Ferreira (2009) and Cartel et al. (2003) supported by the foundations of the presented theory. It is assumed that higher percentage of female representation on the board has a positive effect on a firm’s financial performance due to better monitoring. The following hypothesis was formulated:

𝐻₂: 𝐵𝑜𝑎𝑟𝑑 𝑔𝑒𝑛𝑑𝑒𝑟 𝑑𝑖𝑣𝑒𝑟𝑠𝑖𝑡𝑦 ℎ𝑎𝑠 𝑎 𝑝𝑜𝑠𝑖𝑡𝑖𝑣𝑒 𝑒𝑓𝑓𝑒𝑐𝑡 𝑜𝑛 𝑎 𝑓𝑖𝑟𝑚′𝑠 𝑓𝑖𝑛𝑎𝑛𝑐𝑖𝑎𝑙 𝑝𝑒𝑟𝑓𝑜𝑟𝑚𝑎𝑛𝑐𝑒

2.5 Nationality diversity

Estélyi and Nisar (2016) suggest that director nationality could be an important factor in how the interest of various stakeholders are determined in the corporate world. They argue that foreign board members can influence the quality of a board’s decision-making. Increasing globalisation and operations worldwide require firms having knowledge of international markets as well as experience with different regulatory regimes and it is argued to be crucial knowledge to have within a firm (Rose, et al. 2013). The increased exposure of internationalisation and globalisation by foreign nationals on the board reduces the potential for managerial entrancement and subsequently has the potential to increase firm performance. However, a recent study by Masulis, Wang and Xie (2012) found that board members from foreign countries are weak monitors and cause weaker corporate governance and higher agency costs. As previously indicated, poorer monitoring could lead to worse firm performance (Adams and Ferreira, 2009; Carter et al., 2003). Nationality diversity among board members can increase chances of cross-cultural communication problems (Lehman and Dufrene, 2009) and interpersonal conflicts (Martin, 2014). Other studies have used the agency theory to underpin their finding that nationality diversity causes competitive advantages in terms of increased international networks and commitments to shareholders rights (Oxelheim, Gregorič, Randøy, and Thomsen, 2013; Oxelheim and Randøy, 2003).

Empirical evidence

directors is different in countries outside the U.S. The results show that firm performance is positively influenced by foreign directors following cross-border acquisitions and with directors coming from a country with a strong legal protection of shareholders rights. Their findings are in line with that of Adams and Ferreira (2009) and Carter et al. (2003).

In their study, Masulis, et al. (2012) researched the costs and benefits of foreign independent directors. Their results a show significantly poorer performance of foreign directors on firm performance. These findings are confirmed by Frijns, Dodd and Cimerova (2016) arguing that foreign independent directors are detrimental to firm performance. Frijns et al. (2016) argue this is primarily due to cultural differences whereas Masulis et al. (2012) argue that foreign directors are less effective in monitoring activities and therefore causing higher agency costs. When high-quality legal institutions are present in countries, it is expected that foreign directors cause a poorer firm performance (Miletkov, Moskalev, and Wintoki, 2015).

Rose (2007) finds no significant results on the proportion of foreign nationals on the board and market-based performance. A Nordic study by Randøy, Thomsen, and Oxelheim (2006) analysing the 500 largest companies from Denmark, Norway and Sweden found no significant effect of gender, age and nationality on stock market performance and return on assets (ROA). They conclude that if increased diversity is attractive for a firm (e.g. political preference) it can be achieved without an impact to the market value. Put differently, the market does not view an increase board nationality diversity as a method to increase (neither decrease) e.g. monitoring capabilities (agency perspective), or expand linkages (resource dependency) or non-nationals as being more qualified (human capital).

Hypothesis 3

Based on Estélyi and Nisar (2016) and Miletkov et al. (2015), the hypothesis is drawn that nationality diversity caused by the presence of foreign nationals on the board has a positive effect on a firm’s financial performance.

3. Research Method

This section describes the research method that is used in this study. First, the conceptual model is shown and discussed briefly. In the second part, the variables used in this study are discussed and linked to theory. Third, the collection and used data are presented. Fourth, robustness checks are introduced and potential problems with endogeneity are addressed and explained. The last section also presents the used regression equations.

3.1 Conceptual model

The concept focuses on board diversity constituted by age, gender and nationality. The goal is to test whether diversity of these factors impact the financial performance of a firm. This study uses the board directors as units of observation. Since there is no widely accepted definition of board diversity, and as meaningful composite measures are difficult to construct, this research tests age, nationality and diversity as proxy measures of board diversity one at a time, using regression analysis. The conceptual model used in this research is schematically shown in appendix C. The basic model that is tested:

𝑃𝑒𝑟𝑓𝑜𝑟𝑚𝑎𝑛𝑐𝑒 = 𝛼 + 𝛽1𝑑𝑖𝑣𝑒𝑟𝑠𝑖𝑡𝑦 + 𝛽2𝑓𝑖𝑟𝑚 𝑠𝑖𝑧𝑒 + 𝛽3𝑏𝑜𝑎𝑟𝑑 𝑠𝑖𝑧𝑒 + 𝛽5𝑙𝑒𝑣𝑒𝑟𝑎𝑔𝑒 + 𝜀

Performance is the financial performance of a firm, measured by either Tobin’s Q, ROA and ROE. diversity is a measure of either age, gender or nationality diversity of the board. Firm size is the natural log of the total assets of a firm, valued at the end of the year. Board size takes into account the size of the board. Leverage is added to take into account financial risk. 𝜀 is the error term. The choice of variables and their link with the theory is explained in more details in the next section. The model is used in several statistical estimations further explained in section 3.4.

3.2 Variables

Dependent variable

e.g., Bohren and Strom, 2010; Adams and Ferreira, 2009; Campbell and Minguez-Vera, 2007). However, alternatives like the market-to-book ratio and market-adjusted stock market returns can also be used.

Tobin’s Q is a forward-looking measure that captures the value of a firm as a whole rather than a sum of its parts. Put differently, it is forward-looking because it includes the expected future cash flow of a firm (i.e. the combined market value of firm’s debt and equity). Tobin’s Q is a ratio of the firm’s market value to its book value. Any value greater than ‘1’ suggests that the firm has intangible assets associated with future growth opportunities. According to Montgomery and Wernerfelt (1988), Tobin’s Q is widely used as a proxy for a firm’s ability to generate shareholder value and is a good proxy for a firm´s competitive advantage. A limitation of Tobin’s Q is the limited availability of the market value of debt. Since market data is (in general) not available for private non-listed firms, only listed firms (of which market data is available) can be used.

Independent variables

In order to achieve a reliable and measurable definition of board diversity, three independent variables are used. Because two-tier boards are common in the Netherlands all regressions are run on the executive board. Non-executives generally have a lower impact on the organizational strategy, consequently reducing the effect of board diversity.

The first independent variable is the board age diversity. As mentioned previously, the units of observation are directors so in order to properly analyse the complete board each director’s age is used to calculate the mean age of a board given a specific year (see similar methodology, e.g. McIntyre, Murphy, and Mitchell, 2007). For example, if the board of a given company and year (i.e. 2015) has three directors with the ages 50, 60 and 70, the average age of the board is 60. The database contains information about the average board of directors age.

Second, board gender diversity is used as an independent variable. Board gender diversity is calculated as proportionate female directors on the board (Sanan, 2016; Dutta and Bose, 2006).

The third independent variable is board nationality diversity. This study uses the Blau heterogeneity indicator (Blau, 1977), designed to measure the level of diversity between individuals (Van Ees, Hooghiemstra, Van der Laan, & Veltrop, 2007). The measure is commonly employed in diversity studies (see, e.g., Triana, Miller and Trzebiatowski, 2013; Campbell & Minguez-Vera, 2008). The Blau indicator measures the proportion of a board that belongs to a specific category (such as nationality). The diversity can be calculated as follows (Engelen et al., 2012):

𝐵 = 1 − ∑ (𝑛𝑘 𝑛) ² 𝐾

𝑘=1

In which a board with n members, out whom 𝑛𝑘 are from one specific category k. The outcome B can vary between 0 and 1, in which 0 is not diversified at all, and 1 is fully diversified.

Control variables

Consistent with previous empirical research (Ferrero-Ferrero et al., 2015; Miller and Triana, 2009; Cheng, 2008), several firm-specific variables that could potentially affect firm performance are taken into account as control variable.

and a growing number of hierarchical positions (De Meulenaere, 2016; Pfeffer & Cohen, 1984). In addition, larger firms are also more likely to have international activities and corresponding complexities that ask for diversity (Oxelheim, 2013). Furthermore, firm size is related to both market returns (Fama and French, 1992) and when firm size is measured by asset size it is also related to Tobin’s Q (Faleye, 2007). Taking the above into account, firm size is added as a control variable. Firm size is measured by the value of the natural log of total assets in the regression measured at year-end.

Second, the size of the board is also included in the analysis. Yermack (1996), using the agency theory as the theoretical basis, finds that Tobin’s Q and board size are inversely related. Similar papers such as Ahern & Dittmar (2012) and Nygaard (2011) find that board size negatively influences returns of a firm. Cheng (2008) shows that firms with larger boards have less variability in firm performance. Smaller board tend to take more risks (Pathan, 2009; Cheng, 2008). Some scholars argued that smaller boards of directors, have fewer problems with communication and decision-making, reducing agency cost. Guest (2009) finds that larger boards have a strong negative impact on a firm’s profitability because their effectiveness is lower compared to smaller boards. Contradicting results have been found by Jackling and Johl (2009), arguing that larger boards bring better information and have more knowledge from more directors. This positive association stem from the resource dependency theory arguing that an increase in directors provides for a larger information pool subsequently leading to better firm performance (Carter et al., 2010). Board size is measured by the number of directors on the board. Earlier studies have used this as a control variable due to its impact on firm performance (Sanan, 2016; Campbell & Minguez-Vera, 2008; Cheng, 2008).

Third, this research takes firm-, industry- and time fixed effects into account. This prevents the results to capture differential time-, firm- or industry trends. Behavioural and sociological research deem organizational factors and their fit with the environment as a major factor of success (Hansen, & Wernerfelt, 1989). In addition, it is known that companies operating in the natural resources and mining; manufacturing; and financial activities sectors are underrepresented (Adams and Kirchmaier, 2016). Furthermore Randøy et al. (2006) notes that board diversity is influenced by industry effects. Hence, an industry dummy in combination with a year dummy is added in the OLS regressions to rule out these and similar industry and time-bound effects (Ferrero-Ferrero et al., 2015). In similar fashion, the use of panel data makes it possible to use firm- and year fixed effects in the fixed effects model regressions (see, e.g., Van Ees, Postma and Sterken, 2003).

financial distress (Loderer and Waelchli, 2010). This study uses the ratio of debt to total assets to measure leverage. High leverage firms are associated with larger boards (Wang, Tsai & Lin, 2013).

3.3 Data collection

A six-year sample, from 2010 to 2015, is selected based on all listed companies on Dutch stock exchanges (AEX, AMX, and AScX). This is not the first study to make use of panel data, so it is in line with similar studies (Carter et al., 2010; Rose, 2007). The use of listed firms is necessary because Tobin’s Q is calculated by means of the market capitalisation which is only available for listed firms (and is frequently updated because of the listing on one of the Dutch exchanges). In addition, listed firms are obliged to publish extensive annual statements that contain more detailed information than unlisted firms. Because panel data is used, it is possible to measure the potential effects of diversity over several years (also strategic decision-making often takes some time to take effect).

Two sources are used to retrieve the relevant data. First of all, all board and director characteristics are gathered using the BoardEx database. For each company, data is gathered about the director’s characteristics: age, gender and nationality. Moreover, data is retrieved about the size of the executive board. In the case of missing data, the cells are left blank. BoardEx was unable to retrieve the nationality of 166 entries which have been manually completed by consulting public resources (i.e. annual report, Bloomberg and management scope, see appendix B).

3.4 Analysis and regressions

All equations used in this study are based on the conceptual model but are adjusted to test for either the independent variable 'age', 'gender' or nationality', corresponding with the three hypotheses. Both OLS regression models and fixed effects models are applied. This section will first the used robustness checks and potential endogeneity. Followed by a description of the OLS models (1-3), fixed effects models (4-6) and fixed effects models with lags (7-9).

Robustness check

As a robustness check, all models are estimated by using ROE as an alternative proxy for firm performance. Previous studies (Lu & White, 2014; Joecks, Pull, and Vetter, 2012) used return on equity as an alternative measure of historic performance and can be used alongside with ROA. The use of a third proxy for firm performance decreases the likelihood that results are based on coincidence due to the choice of a specific proxy (Joecks et al., 2012).

The return on equity is calculated by dividing

net income before taxes by shareholder’s equity.

Endogeneity

Prior research (see, e.g., Marinova et al., 2016; Carter et al., 2010; Jackling & Johl, 2009) suggest that well-performing firms attract more diverse board members. This could suggest that board characteristics and firm performance are jointly endogenous (Hermalin & Weisbach, 2003; Field & Keys, 2003). Adams and Ferreira (2009) suggest that problem with endogeneity occur because of variables that are omitted and affect both firm performance as well as the selection of diverse directors. Put differently, firm performance can impact a firm’s consideration to change its board composition (Bhagat and Black, 2001). The endogeneity problem makes it hard to study the effects of board diversity (Ahern and Dittmar, 2012). To cope with the problem, Adams and Ferreira (2009) used firm fixed effects and prove that the used has a significant impact on the results of their analysis. The use of fixed effects estimations is important because they are able to mitigate omitted variables and help to address unobserved changes of time (Carter et al., 2010). In addition, Liu et al. (2014) and Carter et al. (2010) suggest the use of lagged variables to minimize potential problems with reverse causality. Jackling and Johl (2009) explain: “Overall, prior results suggest that boards respond to poor performance by raising their level of board activity, which in turn is associated with improved operating performance in the following years, thus suggesting a lag effect.” One-year lags are used since theory does not predict what length of time is needed for the effects to occur (Cartel et al., 2010; Farrell & Hersch, 2005).

Ordinary Least Squares models

regressed against both Tobin’s Q, ROA and ROE as proxies of firm performance. The moderating effect of firm size, leverage and board size are also taken into account. The following regression functions is used to answer hypothesis 1 to 3, and can be displayed as follows:

(1) 𝑃𝑒𝑟𝑓𝑜𝑟𝑚𝑎𝑛𝑐𝑒𝑗𝑡= 𝛼𝑗𝑡+ 𝛽1𝑎𝑔𝑒𝑗𝑡+ 𝛽2𝑓𝑖𝑟𝑚 𝑠𝑖𝑧𝑒𝑗𝑡+ 𝛽3𝑠𝑖𝑧𝑒𝑗𝑡 +𝛽4𝑙𝑒𝑣𝑒𝑟𝑎𝑔𝑒𝑗𝑡+ 𝛾𝑗+ 𝜆𝑡+ 𝜀_𝑗𝑡 (2) 𝑃𝑒𝑟𝑓𝑜𝑟𝑚𝑎𝑛𝑐𝑒𝑗𝑡= 𝛼𝑗𝑡+ 𝛽1𝑔𝑒𝑛𝑑𝑒𝑟𝑗𝑡+ 𝛽2𝑓𝑖𝑟𝑚 𝑠𝑖𝑧𝑒𝑗𝑡+ 𝛽3𝑠𝑖𝑧𝑒𝑗𝑡 +𝛽4𝑙𝑒𝑣𝑒𝑟𝑎𝑔𝑒𝑗𝑡+ 𝛾𝑗+ 𝜆_𝑡+ 𝜀_𝑗𝑡 (3) 𝑃𝑒𝑟𝑓𝑜𝑟𝑚𝑎𝑛𝑐𝑒𝑗𝑡= 𝛼𝑗𝑡+ 𝛽1𝑛𝑎𝑡𝑖𝑜𝑛𝑎𝑙𝑖𝑡𝑦𝑗𝑡+ 𝛽2𝑓𝑖𝑟𝑚 𝑠𝑖𝑧𝑒𝑗𝑡+ 𝛽3𝑠𝑖𝑧𝑒𝑗𝑡 +𝛽4𝑙𝑒𝑣𝑒𝑟𝑎𝑔𝑒𝑗𝑡+ 𝛾𝑗+ 𝜆𝑡+ 𝜀𝑗𝑡

Subscript j in the above equation reflects the specific industry being analysed. This study applies industry fixed effects, 𝛾_𝑗, and time fixed effects, 𝜆_𝑡, to control for differences in unobservable variables across industries and time (Adams and Ferreira, 2009). 𝜀_𝑗𝑡 represents the corresponding error terms.

Fixed effects models

Because board age-, gender and nationality diversity are continuous, independent variables it is possible to apply a fixed effects estimator. Three fixed effects equations are used to test for a possible relationship between the independent variables and the dependent variables. The equations tested can be displayed as follows. (4) 𝑃𝑒𝑟𝑓𝑜𝑟𝑚𝑎𝑛𝑐𝑒_𝑖𝑡 = 𝛼_𝑖𝑡+ 𝛽₁𝑎𝑔𝑒_𝑖𝑡+ 𝛽₂𝑓𝑖𝑟𝑚 𝑠𝑖𝑧𝑒_𝑖𝑡+ 𝛽₃𝑠𝑖𝑧𝑒_𝑖𝑡 +𝛽₄𝑙𝑒𝑣𝑒𝑟𝑎𝑔𝑒_𝑖𝑡+ 𝜇_𝑖+ 𝜆_𝑡+ 𝜀_𝑖𝑡 (5) 𝑃𝑒𝑟𝑓𝑜𝑟𝑚𝑎𝑛𝑐𝑒𝑖𝑡 = 𝛼𝑖𝑡+ 𝛽1𝑔𝑒𝑛𝑑𝑒𝑟𝑖𝑡+ 𝛽2𝑓𝑖𝑟𝑚 𝑠𝑖𝑧𝑒𝑖𝑡 + 𝛽3𝑠𝑖𝑧𝑒𝑖𝑡 +𝛽4𝑙𝑒𝑣𝑒𝑟𝑎𝑔𝑒𝑖𝑡+ 𝜇𝑖+ 𝜆𝑡+ 𝜀𝑖𝑡 (6) 𝑃𝑒𝑟𝑓𝑜𝑟𝑚𝑎𝑛𝑐𝑒𝑖𝑡 = 𝛼𝑖𝑡+ 𝛽1𝑛𝑎𝑡𝑖𝑜𝑛𝑎𝑙𝑖𝑡𝑦𝑖𝑡+ 𝛽2𝑓𝑖𝑟𝑚 𝑠𝑖𝑧𝑒𝑖𝑡+ 𝛽3𝑠𝑖𝑧𝑒𝑖𝑡 +𝛽4𝑙𝑒𝑣𝑒𝑟𝑎𝑔𝑒𝑖𝑡+ 𝜇_𝑖+ 𝜆_𝑡+ 𝜀_𝑖𝑡

Fixed effects models including lags

To partially cope with reverse causality three other fixed effects models are tested with the inclusion of independent variables that are lagged by one-year (𝑡 − 1).

(7) 𝑃𝑒𝑟𝑓𝑜𝑟𝑚𝑎𝑛𝑐𝑒𝑖𝑡 = 𝛼𝑖𝑡+ 𝛽1𝑎𝑔𝑒𝑖𝑡−1+ 𝛽2𝑓𝑖𝑟𝑚 𝑠𝑖𝑧𝑒𝑖𝑡−1+ 𝛽3𝑠𝑖𝑧𝑒𝑖𝑡−1 +𝛽4𝑙𝑒𝑣𝑒𝑟𝑎𝑔𝑒𝑖𝑡−1+ 𝜇_𝑖+ 𝜆_𝑡+ 𝜀_𝑖𝑡 (8) 𝑃𝑒𝑟𝑓𝑜𝑟𝑚𝑎𝑛𝑐𝑒_𝑖𝑡 = 𝛼_𝑖𝑡+ 𝛽₁𝑔𝑒𝑛𝑑𝑒𝑟_𝑖𝑡−1+ 𝛽₂𝑓𝑖𝑟𝑚 𝑠𝑖𝑧𝑒_𝑖𝑡−1+ 𝛽₃𝑠𝑖𝑧𝑒_𝑖𝑡−1+ 𝛽₄𝑙𝑒𝑣𝑒𝑟𝑎𝑔𝑒_𝑖𝑡−1+ 𝜇_𝑖+ 𝜆_𝑡+ 𝜀_𝑖𝑡 (9) 𝑃𝑒𝑟𝑓𝑜𝑟𝑚𝑎𝑛𝑐𝑒𝑖𝑡 = 𝛼𝑖𝑡+ 𝛽1𝑛𝑎𝑡𝑖𝑜𝑛𝑎𝑙𝑖𝑡𝑦𝑖𝑡−1+ 𝛽2𝑓𝑖𝑟𝑚 𝑠𝑖𝑧𝑒𝑖𝑡−1+ 𝛽3𝑠𝑖𝑧𝑒𝑖𝑡−1 +𝛽₄𝑙𝑒𝑣𝑒𝑟𝑎𝑔𝑒_𝑖𝑡−1+ 𝜇_𝑖+ 𝜆_𝑡+ 𝜀_𝑖𝑡

Results of the above-mentioned regressions are reported in table 6. Fixed effects are taken into account, 𝜇_𝑖 depicts firm fixed effects. 𝜆_𝑡 is used to control for time fixed effects. 𝜀_𝑖𝑡 represents the corresponding error terms.

Multiple regression: F-test

Since the variables of interest are not correlated with each other it is possible to regress them in one regression. Despite looking at the individual significant of factors, an F-test is performed to test whether the group of independent variables (age, gender and nationality) has an effect on the dependent variables and if they are jointly significant. The unrestricted model is based on the previous OLS equation regressions but now included all independent variables. This equation model is described as follows:

(10) 𝑃𝑒𝑟𝑓𝑜𝑟𝑚𝑎𝑛𝑐𝑒𝑗𝑡= 𝛼𝑗𝑡+ 𝛽1𝑎𝑔𝑒𝑗𝑡+ 𝛽2𝑔𝑒𝑛𝑑𝑒𝑟𝑗𝑡+ 𝛽3𝑛𝑎𝑡𝑖𝑜𝑛𝑎𝑙𝑖𝑡𝑦𝑗𝑡+ 𝛽4𝑓𝑖𝑟𝑚 𝑠𝑖𝑧𝑒𝑗𝑡+ 𝛽₄𝑠𝑖𝑧𝑒_𝑗𝑡 +𝛽₅𝑙𝑒𝑣𝑒𝑟𝑎𝑔𝑒_𝑗𝑡+ 𝛾_𝑗+ 𝜆_𝑡+ 𝜀_𝑗𝑡

To help to mitigate omitted variables and address observed changes over time the fixed effects model is applied to the multiple regression too. The fixed effects model is drafted as follows:

(11) 𝑃𝑒𝑟𝑓𝑜𝑟𝑚𝑎𝑛𝑐𝑒_𝑖𝑡 = 𝛼_𝑖𝑡+ 𝛽₁𝑎𝑔𝑒_𝑖𝑡+ 𝛽₂𝑔𝑒𝑛𝑑𝑒𝑟_𝑖𝑡+ 𝛽₃𝑛𝑎𝑡𝑖𝑜𝑛𝑎𝑙𝑖𝑡𝑦_𝑖𝑡+ 𝛽₄𝑓𝑖𝑟𝑚 𝑠𝑖𝑧𝑒_𝑖𝑡+ 𝛽5𝑠𝑖𝑧𝑒𝑖𝑡 +𝛽6𝑙𝑒𝑣𝑒𝑟𝑎𝑔𝑒𝑖𝑡+ 𝜇𝑖+ 𝜆𝑡+ 𝜀𝑖𝑡

To partially cope with reverse causality all independent variables are lagged by one-ear (t-1), leading to the following regression equation:

(12) 𝑃𝑒𝑟𝑓𝑜𝑟𝑚𝑎𝑛𝑐𝑒_𝑖𝑡 = 𝛼_𝑖𝑡+ 𝛽₁𝑎𝑔𝑒_𝑖𝑡−1+ 𝛽₂𝑔𝑒𝑛𝑑𝑒𝑟_𝑖𝑡−1+ 𝛽₃𝑛𝑎𝑡𝑖𝑜𝑛𝑎𝑙𝑖𝑡𝑦_𝑖𝑡−1+ 𝛽₄𝑓𝑖𝑟𝑚 𝑠𝑖𝑧𝑒_𝑖𝑡−1+ 𝛽₅𝑠𝑖𝑧𝑒_𝑖𝑡−1 +𝛽₆𝑙𝑒𝑣𝑒𝑟𝑎𝑔𝑒_𝑖𝑡−1+ 𝜇_𝑖+ 𝜆_𝑡+ 𝜀_𝑖𝑡

4. Empirical results

Subsections 4.1 and 4.2 present the descriptive - and correlation results. Subsection 4.3 presents the results of the ordinary least squares regression between board age diversity, board gender diversity and board nationality diversity on firm performance. The results of the fixed effects models are presented in subsection 4.4 and 4.5. Lastly, a joint significance test is performed and presented in table 4.6.

4.1 Descriptive statistics

Table 2 provides the descriptive statistics of the companies used in this study. The sample used to test the hypotheses is reviewed. The mean of Tobin’s Q (0.839) suggests that the market value of the listed companies during the sampling period is lower than the book value. However, the variation between the minimum value (0.010) and the maximum value (9.160) is significant. Both ROA and ROE reveal a positive mean and show that the firms in the sample have been financially successful on average, according to the accounting measures (mean of 3.400 and 4.213 respectively). The average age of an executive director was 52 while the oldest director was 81 and the youngest 38.

The average number of directors on an executive board amounts 3, with a minimum of 1 and a maximum of 7 directors. These findings are similar to e.g. Adams et al. (2010) and within the threshold of no more than eight board members as recommended by Jensen (1993). The executive board shows a mean of 0.221 with regard to nationality and is lower than previous research that found values of 0.31 (Kilic, 2015) and 0.35 (Nielsen and Nielsen, 2012). The mean of board gender diversity (0.044) and corresponds with earlier findings of Lückerath-Rovers (2013), although a slightly lower percentage of 4.02% was found.This minor difference could be explained because Lückerath-Rovers (2013) uses a time period 2005-2007 and more women have been appointed to the board since that time. A reason for this increase is the introduction of the voluntary quota by the Dutch government. Previous research has

Variables Firm years Mean St.Dev. Min. Max.

Tobin’s Q 327 0.839 0.978 0.010 9.160

Return on assets (ROA) (%) 390 3.400 8.859 -38.93 43.97

Return on equity (ROE) (%) 382 4.213 58.04 -936.4 434

Board age diversity 380 52.25 5.420 38 81

Board gender diversity 429 0.044 0.117 0 0.500

Board nationality diversity 322 0.221 0.256 0 0.800

Natual log of firm total assets 393 15.00 2.218 7.460 21.23

Executive board size 385 2.842 1.245 1 7

Leverage ratio 338 0.518 0.162 0.099 1.061

Table 2 illustrated the descritpive statistics for all variables. The sample data is collected via the Orbis- and

Descriptive statistics

shown an increase in the percentage of women on the board as result of such a quota (Ahern and Dittmar, 2012).

In terms of total assets, the mean amounts 15.00 which is similar to 45.1 million euro. On average the leverage ratio amounts 0.518 with a minimum of 0.099 and a maximum of 1.061. The former value indicates almost no leverage is used, while the latter indicates there is more debt used than equity.

4.2 Correlation

The variables have to fulfil the classic assumptions in order for ordinary least squares estimators to be the best available (Brooks, 2008). To cope with non-normalities, total assets has been transformed by their natural logarithm. This greatly improved their normality in terms of skewness and kurtosis. Since the total database contains sufficient observations any remaining non-normality will not influence the validity of the research. Moreover, the sample size allows for the use of robust standard errors that allow for the presence of heteroscedasticity (Brooks, 2008). To test the assumptions of linearity between the dependent and independent variables, a scatterplot of the residuals are made and these show a linear relationship. The assumption of statistical independence is dealt with in the next section of this paragraph. All variables are tested for multicollinearity and the results are shown in table 3 on the next page. In the case variables fully correlate with another the value is either -1 or 1 (Brooks, 2008). The closer the calculated value is to zero, the fewer variables correlate with another.

Previous research found that larger firms tend to be more diversified in terms of gender (Adams and Ferreira, 2003). The descriptive statistics do not confirm this finding since no correlation was found between firm size and board gender diversity (-0.00). Firm size does correspond in terms of a higher nationality diversity within the board (0.32). This result corresponds with earlier empirical research of Oxelheim (2013) that found that larger firms on average employ more foreigners on their board. The negative relationship between leverage and the performance indicators (Tobin’s Q, ROE and ROA) can be attributed to the capital structure theory3 _{(based on the agency costs of managerial discretion)} arguing that firms with high leverage cannot take advantage of growth opportunities and is consistent with earlier empirical research (see, e.g., Yazdanfar and Öhman, 2015; Lang, Ofek and Stulz, 1996). Fama and French (2002), argue that excessive debt causes higher agency costs which corresponds to a negative relationship between debt and firm profitability.

Other independent variables show only low (<0.29) or no signs of correlation. According to Liu et al. (2014), a correlation in variables only results in a multicollinearity problem when two variables correlate at 0.7 (absolute value). A variance inflation test (VIF) is performed on the variables to test for

correlation. The independent variables all score less than 1.83, yielding almost no multicollinearity between the predictors. As a general rule of thumb, any value lower than 10 is regarded as a sign of no severe or serious multicollinearity (O’brien, 2007). This result confirms the findings shown in table 3.

Variables (1) (2) (3) (4) (5) (6) (7) (8) (9)

(1) Tobin’s Q 1.00

(2) Return on assets (ROA) (%) 0.57 1.00

(3) Return on equity (ROE) (%) 0.15 0.47 1.00

(4) Board age diversity -0.04 -0.02 -0.02 1.00

(5) Board gender diversity -0.10 -0.09 0.01 -0.20 1.00

(6) Board nationality diversity 0.08 0.09 0.00 0.23 0.10 1.00

(7) Natual Log of firm total assets -0.27 -0.03 0.07 0.07 -0.00 0.32 1.00

(8) Leverage ratio -0.41 -0.17 -0.21 -0.12 0.15 0.07 0.19 1.00

(9) Executive board size -0.08 -0.02 0.06 -0.15 0.15 0.32 0.26 0.08 1.00

Table 3 Correlation Matrix

4.3 Ordinary least squares

The first step in identifying the relationship between board diversity and firm performance is the use of an OLS, and a fixed effects model (presented in section 4.3). This relationship is likely to be subject to endogeneity, and thus both would produce biased estimates in the presence of omitted variables and possible reverse causality. Therefore, a second fixed effects model (section 4.5) is applied to partially take reverse causality into account. If the relationship between board diversity and firm performance is determined by factors that are included in the regression, the OLS estimator would produce a significant result once these factors are included. In contrast, if unobservable or omitted variables are determining the relationship, these factors should be captured by the fixed effects estimator (Sila, et al. 2016).

Table 5 reports the standard ordinary least squares regression results that correspond with equation (1), (2), and (3) presented in section 3.4. The equation is run three times to show the effects of the independent variable on the three proxies of firm performance (Tobin’s Q, ROE and ROA). The coefficient of primary interest is 𝛽₁𝑑𝑖𝑣𝑒𝑟𝑠𝑖𝑡𝑦 (measured by age) and 𝐻₀: 𝛽₁= 0 and 𝐻_𝑎: 𝛽₁≠ 0. The equation tests for the hypothesis that board age diversity positively impacts firm performance.

Using Tobin’s Q, as a proxy for performance reveals that the coefficient for the average age of the executive board is only marginally different from zero. The explanatory variable is found not significant, which means there is no evidence of a significant relation between firm performance and board age diversity. The accounting measures ROA and ROE do have a positive coefficient different from zero but no significant results are found. Results are similar to Ferrero-Ferrero (2015) who finds that age diversity positively impact performance. The findings are not in line with Waelchli and Zeller (2013) who find a negative relation between board age and firm performance. The contrasting results could be because Waelchli and Zeller (2013) use a sample of unlisted firms whereas this study only focuses on listed firms.

Since there is only partial evidence of executive board gender diversity having a negative effect on firm performance, the second null hypothesis cannot be rejected. Results are not in line with previous research (see, e.g., Lückerath-Rovers, 2013; Carter et al., 2010; Campbell and Mínguez-Vera, 2008; Erhard et al, 2003; Carter et al., 2003) who all report a positive relationship between gender diversified boards and firm performance. Adams and Ferreira (2009) suggest that higher gender diversity can lead to over-monitoring and increased agency costs but this contradicts the significant result when using Tobin’s Q. Low, Roberts, and Whiting (2016) provide evidence that female director appointments and mandating gender quotas can reduce firm performance with strong cultural resistance. This could indicate economic theories are less relevant but cultural theories are more applicable to explain this result.

Third, the estimations using regression (3) are testing hypothesis 3, whether there is a positive interaction between board nationality diversity and firm performance. Results show that the coefficient of board nationality diversity on Tobin’s Q is 0.050 different from zero and not significant. In addition, ROA and ROE show a negative coefficient implicating exactly the opposite of hypothesis 3. Former evidence reports a positive relationship between board nationality diversity and firm performance (Carter et al., 2003; Oxelheim and Randøy, 2003). However, a more recent study of Carter et al. (2010) reports no significant effect.

(1) (2) (3) (4) (5) (6) (7) (8) (9)

VARIABLES AGEB TQ AGEB ROE AGEB ROA GBX TQ GBX ROE GBX ROA NBX TQ NBX ROE NBX ROA

Board age diversity -0.004 0.714 0.181

[0.011] [0.490] [0.123]

Board gender diversity -0.552** -1.154 -5.665

[0.235] [23.945] [4.326]

Board nationality diversity 0.050 -15.286 -1.814

[0.176] [14.031] [2.733]

Log total assets -0.068*** 2.858* 0.183 -0.068*** 2.914* 0.195 -0.072*** 1.525* 0.149

[0.025] [1.658] [0.264] [0.024] [1.664] [0.261] [0.025] [0.866] [0.307] Leverage -1.207*** -71.196 -5.864 -1.154*** -73.721 -5.830 -1.155*** 9.474 -2.524 [0.207] [51.738] [4.598] [0.218] [52.963] [4.613] [0.256] [21.657] [5.300] Board size 0.085*** 4.921 0.349 0.090*** 4.607 0.344 0.081** 2.938 0.418 [0.031] [3.418] [0.342] [0.031] [3.134] [0.347] [0.034] [3.060] [0.471] Constant 1.945*** -35.904 -3.340 1.749*** -3.153 4.702 1.822*** -21.524 3.305 [0.612] [32.731] [7.862] [0.336] [16.341] [4.185] [0.354] [22.495] [4.737] Observations 271 301 307 275 305 311 246 251 257 R-squared 0.402 0.123 0.162 0.410 0.121 0.161 0.385 0.104 0.105

Year dummies YES YES YES YES YES YES YES YES YES

Industry dummies YES YES YES YES YES YES YES YES YES

Robust standard errors in brackets *** p<0.01, ** p<0.05, * p<0.1

Table 4 shows the ordinary least squares regression with Tobin's Q, return on equity (ROE) and return on assets (ROA) as dependent variables. Age, gender and nationality board diversity are the independent variables. Total assets, leverage of the firm and executive board size are used as control variables. Variable definitions can be found in appendix A, table 1.

4.4 Fixed effects

Table 5 reports the fixed effects models based on equations (4), (5) and (6). Equation (4) only report significant variables for firm size and leverage. This suggests that larger firms, and firms having more leverage, have a lower Tobin’s Q.

As shown in column (5), the negative relationship between board gender diversity and Tobin’s Q is no longer significant and consistent with the result obtained by Carter et al. (2010). This indicates that the significant negative relationship uncovered by the statics OLS equation (5) may be driven by omitted variable biases. Note that one of the accounting measures’ coefficient is negative while the market measure reports a positive coefficient this could be attributed to earlier research by Miller and Traina (2009) arguing that the market views more gender diversity as a positive signal.

Equation (6) is used to test hypothesis 3. Since the coefficients are positive, this might indicate that boards that are more diversified in terms of nationality show a higher performance. However, none of the equations shows significance in the explanatory variable board nationality diversity. Therefore the null hypothesis cannot be rejected. Moreover, the used fixed effects model does produce better estimations but it does not take into account other potential sources of endogeneity, which are likely to occur when researching board diversity and performance (Wintoki, Linck, and Netter, 2012). Findings are in line with research of Randøy et al. (2006) who also did not find a significant effect.

(1) (2) (3) (4) (5) (6) (7) (8) (9)

VARIABLES AGEB TQ AGEB ROE AGEB ROA GBX TQ GBX ROE GBX ROA NBX TQ NBX ROE NBX ROA

Board age diversity 0.005 0.474 0.303

[0.010] [0.970] [0.253]

Board gender diversity 0.093 10.449 -2.507

[0.197] [17.646] [3.778]

Board nationality diversity 0.241 7.911 3.482

[0.214] [14.973] [5.717]

Log total assets -0.176* 11.752 0.840 -0.167* 11.927 1.142 -0.260*** -18.618 -0.526

[0.095] [19.750] [1.491] [0.091] [18.976] [1.475] [0.095] [20.432] [1.691] Leverage -1.375*** -49.694 -6.657 -1.431*** -51.607 -7.552 -1.277*** 106.928 -1.083 [0.286] [149.947] [12.417] [0.282] [146.553] [12.656] [0.242] [123.388] [13.720] Board size -0.001 -1.156 -0.048 -0.001 -1.189 -0.152 -0.014 -2.018 0.070 [0.021] [3.120] [0.698] [0.021] [3.104] [0.657] [0.020] [2.644] [0.897] Constant 3.944** -162.225 -18.738 4.071*** -139.231 -6.731 5.386*** 239.707 13.417 [1.562] [177.934] [24.183] [1.371] [206.735] [19.844] [1.414] [239.280] [21.397] Observations 271 301 307 275 305 311 246 251 257 R-squared 0.294 0.026 0.068 0.297 0.026 0.055 0.319 0.081 0.039 Number of firm 55 59 59 55 59 59 49 49 49

Firm FE YES YES YES YES YES YES YES YES YES

Year FE YES YES YES YES YES YES YES YES YES

Industry FE NO NO NO NO NO NO NO NO NO

Robust standard errors in brackets *** p<0.01, ** p<0.05, * p<0.1

Table 5

Fixed Effect model: Relation between board diversity and firm performance

4.5 Fixed effects including lags

Despite the fact that the fixed effects method accounts for omitted variable bias, the possibility of reverse causality still exists. By adding lagged variables, the possibility that firm’s performance affects board diversity, and thereby the reverse causality is reduced as suggested by Liu et al. (2014) and Carter et al. (2010). As a robustness check, the previous models have been estimated again but now with the inclusion of one-year lags for the independent variables.

The estimation method does not change the main dependent variables of interest (Tobin’s Q and ROA) in equation (7) and R² is somewhat smaller in comparison to the previous model. Equation (8) changed the interaction effect of ROA to a negative one in comparison to the previous model although R² is somewhat lower and as with the non-lagged model and no significant result is found. The ROE equation reports a significant positive relationship in support of hypothesis 2. The coefficients of both Tobin’s Q and ROE are again positive and in conformation with the theory and empirical evidence suggesting that women are better monitors (Post and Byron, 2015; Adams and Ferreira, 2009).

Equation (9) reports positive coefficients for all performance variables. Including a lag turned the coefficient of ROE to a positive one. No significant results are found.

Note that the coefficients of the ROE equation (7), ROA equation (8) and ROE equation (9) show opposite results compared to the fixed effects model, indicating it is possible this relationship is driven by reverse causality.

4.6 Multiple regression: F-test

Table 7 reports the multiple linear regression that was calculated to predict the joint effect of age, gender and nationality. The coefficients of the independent variables are similar to the previous OLS model when using Tobin’s Q and ROA as the dependent variable in equation (10). The coefficient of board gender diversity changes from negative to positive when using ROE, which is in line with the presented theories.

Results of the fixed effects equation (11) are similar to the previous fixed effects model. The control variable total assets has increased its significance to the 1% significance level. With the exception of board gender diversity in the ROA equation, all coefficients are positive and in conformation with the presented theory.

The results of the F-test reports no joint significance for the independent variables after controlling for omitted and missing values. This means the variables age, gender and nationality are not able in predicting the performance of a firm.

(7) (8) (9)

VARIABLES AGEB TQ AGEB ROE AGEB ROA GBX TQ GBX ROE GBX ROA NBX TQ NBX ROE NBX ROA

Board age diversity 0.011 -0.603 0.029

[0.008] [0.477] [0.122]

Board gender diversity 0.222 24.787** 1.928

[0.196] [10.060] [4.523]

Board nationality diversity 0.300 -8.127 0.291

[0.196] [10.347] [3.280]

Log total assets 0.003 -1.484 0.153 0.010 -2.239 0.008 -0.031 -0.660 0.086

[0.022] [1.424] [0.382] [0.021] [1.515] [0.446] [0.030] [1.172] [0.562] Leverage -0.565** 11.748 -1.675 -0.626** 9.279 -2.596 -0.640** 4.139 -4.601 [0.235] [9.494] [8.632] [0.270] [10.166] [8.130] [0.288] [9.595] [9.836] Board size -0.051** -0.652 0.228 -0.053** -0.702 0.189 -0.080*** -2.220** 0.085 [0.021] [0.986] [0.346] [0.021] [1.001] [0.321] [0.020] [1.006] [0.376] Constant 0.580 37.198 2.785 1.047** 41.888* 7.000 1.734*** 29.792 7.584 [0.524] [22.914] [8.826] [0.428] [21.973] [6.315] [0.567] [18.018] [7.829] Observations 250 276 282 254 280 286 227 233 239 R-squared 0.249 0.038 0.040 0.249 0.045 0.038 0.295 0.054 0.029 Number of firm 52 58 58 52 58 58 49 55 55

Firm FE YES YES YES YES YES YES YES YES YES

Year FE YES YES YES YES YES YES YES YES YES

Industry FE NO NO NO NO NO NO NO NO NO

All independent variables are lagged one period. Robust standard errors in brackets. *** p<0.01, ** p<0.05, * p<0.1

Table 6

Lagged Fixed Effect model: Relation between board diversity and firm performance

(10) (11) (12)

VARIABLES OLS TQ OLS ROE OLS ROA FE TQ FE ROE FE ROA FE w/ lag TQ FE w/ lag ROE FE w/ lag ROA

Board age diversity -0.013 0.747 0.166 0.007 1.964 0.322 0.011 -0.203 0.066

[0.012] [0.681] [0.137] [0.012] [1.375] [0.307] [0.010] [0.649] [0.177]

Board gender diversity -0.733** 3.339 -2.923 0.052 10.624 -3.192 0.075 20.583 -0.322

[0.300] [24.347] [5.350] [0.261] [19.100] [5.250] [0.225] [18.419] [7.329]

Board nationality diversity 0.200 -18.919 -2.153 0.210 0.527 2.329 0.266 -7.915 0.090

[0.183] [18.580] [2.862] [0.248] [17.401] [5.904] [0.217] [10.688] [3.536]

Log total assets -0.074*** 1.554* 0.146 -0.272*** -23.709 -1.288 -0.050 0.080 -0.019

[0.025] [0.879] [0.312] [0.092] [22.607] [1.675] [0.032] [1.331] [0.513] Leverage -1.160*** 12.676 -1.736 -1.271*** 115.808 0.057 -0.623*** 4.473 -3.377 [0.258] [24.128] [5.396] [0.233] [124.985] [13.399] [0.213] [11.294] [9.215] Board size 0.077** 3.289 0.511 -0.013 -1.751 0.021 -0.058** -1.717 0.140 [0.036] [3.257] [0.485] [0.021] [2.778] [0.930] [0.026] [1.262] [0.330] Constant 2.436*** -57.377 -4.484 5.201*** 209.770 8.005 1.496* 26.966 4.457 [0.706] [52.793] [9.146] [1.535] [216.320] [22.641] [0.806] [34.380] [14.334] Observations 250 276 282 254 280 286 227 233 239 R-squared 0.249 0.038 0.040 0.249 0.045 0.038 0.295 0.054 0.029 F-statistic 0.1044 0.5725 0.5398 0.4939 0.5058 0.5410 0.1068 0.2349 0.9685 [2.07] [0.67] [0.72] [0.81] [0.79] [0.73] [2.15] [1.46] [0.08] Number of firm 49 49 49 49 55 55

Firm FE NO NO NO YES YES YES YES YES YES

Year FE YES YES YES YES YES YES YES YES YES

Industry FE YES YES YES NO NO NO NO NO NO

All independent variables are lagged one period. Robust standard errors in brackets. F-statistic value in brackets. *** p<0.01, ** p<0.05, * p<0.1

Table 7

Multiple regression: Joint significance testing

5. Conclusion

Over the past few years, diversity on the board of directors has become a much debated topic. The purpose of this paper is to present empirical evidence on the effect of board diversity on firm performance and complement the existing literature. Firm performance was defined as financial performance and measured by one market-measure (Tobin’s Q) and two accounting measures (ROE and ROA). The scope of this study consisted of a sample of Dutch listed firms, which were examined over the period 2010 to 2015. The combination of the upper echelons theory , agency theory, resource dependency theory, and recent empirical literature have not been able to clearly demonstrate a relationship between board diversity and firm performance. Panel data was used to estimate both OLS and fixed effects models and take partially take into account the potential endogeneity issues.

First of all, the results of the OLS regressions show a positive- but non-significant relationship between the board age diversity and firm performance. When Tobin's Q was used as a measure of financial performance, the coefficient of the explanatory variable is only marginally different from zero and not significant. To take care of endogeneity, fixed effects models have been employed. The results from the fixed effects regression equations indicate a positive but again non-significant relationship between a board average age and firm performance. An explanation for this positive coefficient is provided by the human capital theory arguing that age comes with greater knowledge and experience. At the same time, the resource dependency theory predicts that on average older boards have increased linkages to other stakeholders. These findings are in line with previous resource by Ferrero-Ferrero et al. (2015), Engelen et al. (2012), and Hasan and Wu (2012).

Finally, the third hypothesis arguing that the presence of foreign nationals has a positive effect on firm performance was tested. The OLS regression result shows only a coefficients marginally different from zero when it is tested on the forward-looking performance indicator Tobin's Q. However, both ROA and ROE indicate that the presence of foreign nationals has a negative impact on firm performance. All OLS estimations are non-significant. Contrary results were found when using fixed effects models which could indicate the OLS model suffers from the omitted variable bias. The fixed effects model coefficients indicate higher board nationality diversity comes with better firm performance. The resource dependency theory argues that differences in nationality diversity increases international network (Oxelheim et al., 2013; Oxelheim, 2003).Lastly, the joint significance of the independent variables (age, gender and nationality) could not be proven. However, the individual t-statistics of the joint regression model including age, gender and nationality reported similar results to the other models. As a last note, it is possible that the used models suffer from reverse causality as is reported by the fixed effect model with lags.

Overall, the positive coefficient results can, by and large, be attributed to the theories introduced. Increased diversity bolsters independence and lessens agency problems within the firm. In addition, higher levels of diversity expand boards’ linkages taking into account stakeholders’ needs and limit a firms’ dependence on strategic resources. Lastly, the human capital theory predicts an increase in different skills and experiences as diversity increases. The differences between the coefficients of the forward- and backward-looking dependent variables can be attributed to signals to the market (Miller and Triana, 2009). Increased diversity may lead companies towards higher performance and competitiveness. Shareholders and its management can therefore not only benefit from greater diversity but at the same time act in line with moral and societal norms.

This study has some limitations. First of all, results can suffer from contingencies because this study was conducted on firms listed on the Dutch stock exchanges. Other countries have different rules and regulations, business culture, and potentially other factors that have an influence on board diversity and firm performance. Conducting this study across multiple countries and with a larger sample of firms would enhance the understanding of board diversity and its potential effect on firm performance. In addition, it is advised to take into account the survivorship bias which is not addressed in this study. Second, the scope of this research was purely on economic theories. Since no significant results could be found it can be that a possible link between board diversity must be sought not only in economic theory but jointly with (e.g.) social theories.

Appendix

Appendix A: Variables overview

Table 1 Variable Definitions Financial performance (dependent variables) Definition

Tobin’s Q Calculated by dividing book value of total assets by market

capitalisation at the end of a year.

Return on assets (ROA) (%) Calculated by dividing total assets by net income at the end of the

year.

Return on equity (ROE) (%) Calculated by dividing shareholders equity by net income at the end

of the year.

Diversity (independent variable)

Board age diversity Sum of the ages of all executive board directors divided by total size

of executive board.

Board gender diversity Gender diversity of the executive board. Calculated as a percentage

of total board members of a firm.

Board nationality diversity Nationality diversity of the executive board. Calculated using the

Blau-index.

Control variables

Natual log of firm total assets Natural logarithm of total assets measured by the book value at

year-end.

Board size Number of directors on the executive board.

Leverage ratio Leverage ratio calculated by; (short-term debt + long-term debt)

Appendix B: Manually adjusted or corrected data

Some of the data have been manually adjusted, supplemented or adjusted in order to take care of missing values or increase the total amount of observations.

With regard to 168 positions (directors) data was missing with regard to their nationality and/or gender. Via the use of public sources such as annual reports, management scope and company websites this information was added. In addition, some companies were added as a whole because the information was not available in either Orbis or BoardEx. These companies are:

● Value8 ● Amsterdam Commodities ● Stern Group NV ● Ballast Nedam NV ● Corio NV ● Exact Holding ● Grontmij NV ● Hunter Douglas NV ● Kardan NV

● Macintosh Retail Group NV ● Nutreco NV

● Pharming S.A. ● TKH Group N.V. ● Unit4

● USG People NV