How does female representation within a board of directors influence firm
performance, and what is the moderating effect of gender equality?
MSc IB&M Thesis
Irene van der Klooster Student number: S3413187 i.a.r.van.der.klooster@student.rug.nl
Supervisor: Dr. M.J. Klasing Co-assessor: Prof. Dr. J. De Haan
University of Groningen Faculty of Economics and Business
Duisenberg Building, Nettelbosje 2, 9747 AE Groningen, The Netherlands P.O. Box 800, 9700 AV Groningen, The Netherlands
ABSTRACT
The objective of this thesis is to enrich the existing research on board gender diversity within the board of directors and firm performance. Previous research on this topic has yielded contradicting findings, which could be the result of heterogeneity in effects because higher female representation within boards only produces benefits for firms under certain conditions. I assume that a nation’s gender equality, an unexplored moderator, could be such a condition influencing the effects of board gender diversity. This is because in societies with negative attitudes towards females, female board members are assumed to be unequal; therefore, the possible benefits of their presence such as high-quality decision-making are expected to diminish. Thus, I hypothesise that gender equality positively moderates the effect of board gender diversity on firm performance. Based on secondary data from 884 firms in 36 countries in the service sector, while utilizing ROA to measure firm performance, I find that the results are inconclusive, indicating that the results should be considered carefully and that future research on this topic is necessary. Nevertheless, the results provide a valuable foundation for future research.
ACKNOWLEDGEMENT
TABLE OF CONTENT
INTRODUCTION ... 6
LITERATURE REVIEW ... 8
Upper Echelon Theory ... 8
Board Diversity ... 10
Board gender diversity ... 12
Gender equality ... 16
Hypothesis development ... 20
METHODOLOGY ... 21
Data collection and sample ... 21
Variables ... 22 Dependent variable ... 22 Independent variable ... 23 Moderator ... 23 Control variables ... 24 Analytical strategy ... 26 Heteroscedasticity ... 27 Multicollinearity ... 27 RESULTS ... 28 Descriptive statistics ... 28
The evolution of female representation and gender inequality (2008-2017) ... 30
Correlation ... 32
Regression results ... 34
Robustness check ... 37
CONCLUDING REMARKS ... 45
Conclusion and discussion ... 45
Managerial implications ... 48
Limitations and future research ... 49
APPENDICES ... 52
Appendix I. Countries and regions in the service sector ... 52
Appendix II. Scatterplot of residuals ... 53
Appendix III. Breusch-Pagan/Cook-Weisberg test results ... 53
Appendix IV. VIF results ... 53
LIST OF TABLES
Table 1. Summary of all variables, indicators and data sources ... 26
Table 2. Descriptive statistics ... 29
Table 3. Correlation matrix ... 33
Table 4. OLS regressions ... 35
Table 5. Robustness check by restricting the sample ... 38
Table 6. Robustness check using Tobin's Q ... 41
Table 7. Robustness check by replacing the UN data with the EU data ... 43
LIST OF FIGURES Figure 1. Strategic choice under conditions of bounded rationality ... 9
Figure 2. Conceptual framework ... 21
Figure 3. Gender inequality index and indicators ... 24
Figure 4. Evolution of board gender diversity (2008-2017) ... 31
INTRODUCTION
Board gender diversity is a highly discussed topic, because female representation within boards is only 10.3% across 67 countries (Terjesen, Aguilera & Lorenz, 2015). This problem keeps countries and international organisations occupied while searching for solutions to increase gender diversity or increase the share of females in executive positions. Because boards are generally male-biased (Peterson & Philpot, 2007), gender diversity automatically refers to an increase in female representation in corporate boards. Therefore, the remainder of this thesis uses these terms interchangeably. The amount of research on gender diversity by the United Nations (2018) and OECD (2018) has recently increased. This has precipitated requirements to hire more women in organisations to increase female representation within boards (Adams & Ferreira, 2009). As a result of this, the U.S. and European countries have established quotas for a specific number of females within boards (Carter, D’Souza, Simkins & Simpson, 2010; Adams & Ferreira, 2009). For example, California requires all publicly listed firms to have at least one female on their board by the end of 2019 (Reuters, 2018). Altogether, this reflects the relevance of the subject of this thesis.
Because of the limited representation of female board members, various policy makers and researchers have recently been attracted to this topic (Conyon & He, 2017). Researchers have argued that an increase in female board members offers various advantages for firm performance (Conyon & He, 2017), such as stricter board monitoring, higher attendance during meetings (Adams & Ferreira, 2009) and high-quality decision-making. The latter is the consequence of an increase in the quality of reasoning as a result of females’ counterarguments (Milliken & Martin, 1996). Overall, the main benefits are that diverse boards enjoy enhanced information flows, new ideas, and greater innovation (Bantel & Jackson, 1989; Milliken & Martins, 1996). However, Ahern and Dittmar (2012) and Matsa and Miller (2013) demonstrated a negative relationship between board gender diversity and firm performance because female directors generally have less board experience and are risk-averse. Because the effect of increased female representation could be both positive and negative, the findings of Carter et al. (2010) and Erhardt, Werbel, and Shrader (2003), which provide insignificant results regarding this relationship, are not unexpected. Thus, the findings on the effects of board gender diversity on firm performance remain inconclusive (Ferreira, 2015; Post & Byron, 2015).
variables (Adams, Haan, Terjesen, & Ees, 2015; Ferreira, 2015; Conyon & He, 2017). By examining different conditions and contexts that are likely to influence the effect of female board members, these mixed findings could be adjusted (Post & Byron, 2015). Therefore, this study revisits the question of whether an increase in female representation in boards results in higher firm performance while also examining the influence of another condition: a new moderator. Firm strategic orientation and organisational culture are examples of moderators which have previously been studied and which strengthen the effect of gender diversity on firm performance (Dwyer, Richard & Chadwick, 2003). To my knowledge, gender equality has never been studied in this context; this thesis will contribute to the existing research, which suffers from certain imperfections.
A country’s level of gender equality is expected to influence the relationship between the number of female board members and firm performance. For example, in comparison to Western countries, countries such as India have certain cultural norms that produce higher gender inequality. In India, women are generally perceived as inferior to men, and they are expected to be ‘soft’ and ‘reserved’ and to look after the household (Benson & Yukongdi, 2005). Therefore, they have fewer opportunities to participate in board-related activities such as decision-making (Campbell & Mínguez-Vera, 2008). Some of the consequences of gender inequality are stereotyping and tokenism. The former corresponds to (partially incorrect) ideas regarding “a social group which are shared with other members of a culture” (Low, Roberts & Whiting, 2015, p. 382). Tokenism refers to females being appointed to board positions as a gesture rather than for their actual capabilities. Both of these consequences result in lower female contribution. Accordingly, it is expected that the benefits which derive from board gender diversity will be suppressed in societies with low gender equality (Nielsen & Huse, 2010). Thus, female representation within boards is expected to be less advantageous in some countries, while it is expected to be more advantageous in countries with positive attitudes towards females. Furthermore, since there is still a substantial difference between the highest (India: 0.581) and lowest (Sweden: 0.044) gender inequality index (United Nations Development Programme, 2019), this is a worthwhile avenue of investigation. This prompts the following research question:
The question whether female representation in corporate boards benefits firm performance and what the effect of a nation’s gender equality is will be investigated. By doing this, this thesis will provide valuable insights into the IB field and provide managerial implication. These managerial implications enable firms to actually benefit from the increased female representation within boards.
LITERATURE REVIEW
This literature review explains the theoretical concepts of this research. First, it discusses the foundation of this research, Upper Echelon Theory, followed by diversity and board gender diversity. Subsequently, it explains the moderating variable, gender equality. Finally, the conceptual model of this thesis is discussed.
Upper Echelon Theory
created a model (Hambrick & Snow, 1977) visualising the complex situation executives face when making a strategic decision (Figure 1). Since a decision is composed of more phenomena than an executive can possibly apprehend, the process is inherently complex. The psychological and observable characteristics create a disparity between the actual situation and an executives’ perception of it. This perceptual process consists of three sequential steps. First, it is likely that the situation extends beyond the cognitive base and values of executives or top management teams. Therefore, they are not capable of fully understanding the organisation and its environment. Second, because human beings selectively perceive a small portion of the phenomena included in their field of vision, executives’ perceptions are limited even further. Finally, information is interpreted based on one’s cognitive base and values. Based on this, Hambrick (2007) states that the theory is predicated on the premise of bounded rationality, meaning that situations are not objectively ‘knowable’, but rather interpretable (Cyert & March, 1963; March & Simon, 1958). Summarising the model, the basis for strategic choices is composed of a limited field of vision, followed by an executive’s perception and interpretation of a situation, which is based on an executive’s cognitive base and values.
Figure 1. Strategic choice under conditions of bounded rationality
Source: Hambrick and Mason (1984)
Despite the critics on UET (Lawrence, 1997; Pettigrew, 1992), the theory has been widely used in several works (Lee & Park, 2006; Hodgkinson & Sparrow, 2002) because it is reliable, objective, and uses a testability approach (Finkelstein & Hambrick, 1996) to predict organisational outcomes (Lee & Park, 2006). For instance, Lawrence (1997) claims that demographic characteristics contain useful features allowing validity and replicability which enables to prove the argument of the theory. Due to the fact that board members, as a result of their interests, values, and decisions, can impact firm performance, it is necessary to understand the influence of executives’ attitudes on various organisational outcomes (Hambrick, 2007). Because the theory stresses that managerial characteristics, and particularly a variety of executives, influence strategic decisions and in turn organisational performance (Caligiuri, Lazarova & Zehetbauer, 2004; Hambrick & Mason, 1984), it is utilised in this study to test the relationship between the female representation within boards and firm performance.
Board Diversity
As discussed, the BOD controls the firm, and a BOD’s effectiveness can significantly impact firm performance (Hambrick & Mason, 1984). According to Terjesen et al. (2009), the amount of research on board diversity and gender diversity has increased, but there are opposing views on this topic.
will increase the quality of problem-solving (Carter, et al, 2003). However, a diverse board may also hinder the decision-making or problem-solving process due to the variety of perspectives, norms, values, and experiences. Conflicts and longer, less efficient board meetings may arise as a result of this heterogeneity (Randøy, Thomsen & Oxelheim, 2006; Carter, et al, 2003). Nevertheless, diverse boards enable access to important resources such as suppliers, partners, or customers (Alexander, Fenell & Halpern, 1993; Randøy, et al, 2006). Due to these additional information sources, firms benefit from improved insight into their customers or key stakeholders, fostering a better understanding of the marketplace (Carter, et al, 2003; Randøy, et al, 2006).
To obtain a deeper understanding of the influence of diversity, the specific dimensions of diversity will be discussed. According to Barkema and Shvyrkov (2007), the composition or diversity of boards can be divided into groups based on demographic features such as nationality, age, and gender. These factors are reviewed in the following section.
provides significant evidence that multinational boards generally result in enhanced team and subsidiary performance.
The second dimension of diversity is age diversity. McIntyre, Murphy, and Mitchell (2007) and Richard and Shelor (2002) present significant findings that moderate age diversity is positively correlated with firm performance, while low and high levels of age diversity result in low firm performance. They also find that age diversity could result in more experience and a larger social network. Furthermore, older executives are more capable of coping with unexpected circumstances, typically operate more conservatively, and have an increased sense of responsibility due to their experience, while younger executives are more likely to be more innovative, dynamic, and less risk-averse (Timmerman, 2000). Thus, a ‘healthy’ age variety will improve firm performance (Li, Chu, Lam & Liao, 2011). On the other hand, Tsui, Egan, and O’Reilly (1992) stress a negative relationship between age diversity and firm performance, since they determined that executives whose ages differ from other board members experience a lack of cohesion and integration (Wiersema & Bird, 1993). Therefore, they are more inclined to abandon the organisation and more likely to be less devoted to their organisations, and absenteeism is higher amongst them. Finally, Wegge, Roth, Kanfer, Neubach, and Schmidt (2008) provide contradictory findings on the effect of age diversity. According to their research, age diversity will only have beneficial effects when coping with complex situations, while a board with easier tasks will be more effective if it is homogenous.
In summation, the theoretical arguments and existing empirical findings on some of the diversity dimensions are contradictory because negative, positive, as well as nonsignificant relationships have been observed. Additionally, a healthy diversity should optimise the results, whereas too much or too little heterogeneity might hamper performance. The same is expected for the third and final dimension of diversity, namely gender. Because this is the basis for this thesis, this dimension is discussed thoroughly in the following section.
Board gender diversity
in effective boards (Bianco, Ciavarella, & Signoretti, 2015). Therefore, gender diversity has been studied by several researchers. When more women are appointed to board positions, group and firm performance is influenced (Konrad, Kramer & Erkut, 2008). Examples of this include alterations in decision-making and altered perspectives of board members (Dahlin, Weingart & Hinds, 2005; Allini, Rossi & Hussainey, 2016). Whether these influences are positive or negative is debatable because there are contradictory theoretical arguments and findings on the effect of board gender diversity (Conyon & He, 2017; Carter, Simkins & Simpson, 2003; Adams & Ferreira, 2009).
which in turn enhances the quality of board decision-making (Conyon & Mallin, 1997). However, according to Conyon and He (2017), the distinctive perspectives and preferences of individual board members could lead to less effective board decisions due to trade-offs and bargaining, in turn lowering firm performance. Despite this, because of the distinctive experiences of women, different search-strategies are used to gather information (Erhart, et al, 2003), and information and knowledge can be drawn from different sources (Dahlin, 2005) because the frame of reference of females is different, which translates into a higher ability to assess information. Additionally, women are typically more open-minded and patient, which enables communication and supports information dispersion (Gul, Srinidhi, & Ng, 2011). Therefore, when considering the information supply in heterogenous groups in terms of gender, the available information on which the decision-making process is based is expanded (Dahlin, Weingart, & Hinds, 2005). Together, this complementary information that females provide could enhance the decisions made by the board. Another theoretical argument regarding why increased female representation on the board is beneficial for firms is because it fosters innovation and creativity (Erhardt, et al, 2003; Miller & Del Carmen Triana, 2009). Because women have different backgrounds, they approach problems distinctly, and alternative opportunities are taken into consideration. However, this might translate into time-consuming processes due to the heterogeneous in-group dynamics. Because females are better able to provide original perspectives and ideas on complex topics (Francoeur, Labelle & Sinclair-Desgagné, 2008), innovation and creativity is fueled (Erhardt, et al, 2003; Tsui, Ergran & Xin, 1995). Because innovation stimulates creative new ideas, which allows a business to distinguish itself from the competition, this is perceived as an indicator of firm performance (Hitt, Hoskisson & Kim, 1997). In summation, because men are not able to provide the competencies and capabilities women provide, women are vital in organisations (Benschop & Verloo, 2010).
industries and Tobin’s Q as outcome variable, also found a positive, significant relationship between the proportion of females on the board and firm performance, for the same reason that it is associated with a better understanding of the marketplace and enhanced creativity, innovation, and problem-solving. However, because board diversity is related to efficient leadership within firms and global relationships, it is also related to firm performance. On the other hand, Adams and Ferreira (2009) and Shrader et al. (1997), both studying US firms in various industries, determined significant negative relationships between female representation and firm performance. Adams and Ferreira (2009), which used ROA and Tobin’s Q to measure firm performance, found that relative to males, female board members have higher meeting-attendance ratios. Furthermore, male directors have less attendance problems as female representation on the board increases, and the likelihood that females contribute in monitoring committees is higher. However, extreme forms of monitoring will decrease firm performance. Schrader et al. (1997), which employed ROS, ROA, ROI, and ROE to measure firm performance, associate higher percentages of women with better capabilities of developing new and innovative ideas.1 Nevertheless, their negative findings could be the result of the fact that female board member have not been present long enough in board positions to have a positive impact (Shrader, et al, 1997). When considering these opposing results, the insignificant results of Carter et al. (2010) and Randøy et al. (2006) are not surprising. Carter et al. (2010) based their sample on US firms from various industries and used ROA and Tobin’s Q as measurements of firm performance. While their analysis does not support their propositions that women provide complementary characteristics, information, networks, human capital, and improved governance, which in turn increases firm performance, they also find no evidence for negative effects either. Randøy et al. (2006), using Nordic countries in their sample, while using market to book value ratio and ROA, also concluded that there are no significant positive or negative effects of gender diversity.
Whether these contradictory empirical findings are the result of the distinctive circumstances in which the studies have been executed is debatable (Adams, Haan, Terjesen, & Ees, 2015; Ferreira, 2015; Conyon & He, 2017). Omitted variables, the country in which the study was executed, different time frames, and differing measurement approaches could have influenced the results. For example, the majority of studies on this topic studied firms in American and Nordic countries, which are characterised by distinctive regulations and board systems. Furthermore, the
data was collected in distinctive time frames. Nevertheless, it is expected that additional factors cause the contradictory results, because while Carter et al. (2003) which focused on US firms found a positive relationship, Shrader et al. (1997), also focusing on US firms, observed a negative relationship. The legal or cultural context might be a moderating variable altering the effects of the female representation in the board (Randøy, et al, 2006; Low, Robert & Whiting, 2015). To be more specific, the level of gender equality in the country in which the study is conducted is expected to be a moderating variable influencing the relationship between the female representation on the board and firm performance.
Gender equality
Greater representation of female board members is likely to have a positive influence on firm performance. However, the cultural context which determines the attitudes, norms, and beliefs towards the acceptance of gender equality could influence this effect (Randøy, et al, 2006). For instance, a country with low gender equality could alter the positive effects on firm performance; therefore, this cultural aspect cannot be ignored (Low, et al, 2015).
Human development has been hampered by gender inequality for centuries. In terms of the labour market, education, and other societal areas, women are often discriminated against and seen as less valuable (United Nation Development Programme, 2019). However, over the past several decades in the developed world, serious efforts have been implemented to minimise gender discrimination and increase equality in all areas, particularly in the workplace (Armstrong, Flood, Guthrie, Liu, Maccurtain & Mkamwa, 2010). For instance, to conform to the Equality Act of 2004, the Irish government has prohibited discrimination against employees and those searching for a job with regards to characteristics such as gender, age, and ethnicity (Armstrong, et al, 2010).
access to the labour market, both men and women are able to participate in earning money. Secondly, “the equal sharing of money”: In an ideal and equal society, work of equal value should be rewarded by equal pay, which in turn should lead to equal incomes for men and women. Thirdly, “equal sharing of decision-making power”: The interests of the entire population will be considered in terms of a balanced participation of men and women in decision-making processes. In this way, the decision-making process will be based on a multiplicity of values and ideas. Finally, “equal sharing of time”: In a gender-equal society, all residents contribute in a balanced way in all domains of life, including work, care, and leisure. In other words, in a truly gender-equal society, the concept of unpaid time should gender-equally distributed.
Now that the concept of gender equality has been explained, previous research on this topic will be discussed to determine the possible influence of gender equality. No previous research has focused on the effect of equality on firm performance (Monks, 2007) and more specifically on the effect of gender equality on the relationship between female representation within boards and firm performance. Some empirical research has stressed generally positive effects; however, the outcomes are rarely generalizable (Armstrong, et al, 2010) and the capability to convert equality and diversity guidelines into enhanced performance is context-dependent (Jayne & Dipboye, 2003; Yasbek, 2004). Furthermore, according to Jackson, Joshi, and Erhardt (2003), only when diversity and equality are adequately accepted, adopted, and managed, it could result in enhanced firm performance.
The benefits of female representation within boards are expected to be moderated by a country’s cultural attitude towards women. In other words, an increase in female representation within boards might be advantageous in one cultural context, while in another context it might not be. A nation with a negative attitude towards females is expected to experience a preference for male executives, male dominated boards, and wage gaps for equal work (Hofstede, 2001; Low, et al, 2015). These societies are more likely to engage in tokenism and stereotyping, which can negatively impact females’ performance at work (Low, et al, 2015).
decision-making, which will be discussed thoroughly later in the paper, is impeded (Zimmer, 1988).
In male dominated positions such as board functions, stereotyping against women is deeply rooted (Bergeron, Block & Echtenkamp, 2006). According to board selectors, females are lacking in required human capital for board positions (Burke, 2000). For example, for board positions, females are anticipated to be less capable because they are identified with being more social and service-oriented instead of achievement-oriented, which are assumed to be features of great executives (Eagly & Johannesen-Schmidt, 2001). Due to this preconception, the assessment of female skills and capabilities for a certain position, which is commonly seen as being better suited for males, is affected (Kanter, 1977). Therefore, women are given less opportunities to affect board decision-making when they are stereotyped as unequal and less valuable board members in comparison to men (Nielsen & Huse, 2010). In addition, another consequence of stereotyping is ‘stereotyping threat’, which refers to the anxiety of being judged and estimated according to a negative stereotype, and the worry of doing something that could support the negative stereotype (Roberson & Kulik, 2007). When females are executing their job, if they perform badly and correspond with this stereotyped image of women, this threat is present (O’Brien & Crandall, 2003). Because of this threat, an individual’s performance is affected by these anxieties, drawing the attention to problems that are not relevant to their job, which could reinforce the negative stereotype (Steele & Aronson, 1995), which will negatively affect firm performance. As a result of the stereotyping threat, it is not exceptional that females could conceal their personalities and act similar to their male colleagues to fit in and be accepted (Reciniello, 1999). Eventually, the effectiveness of female representation will decrease (Low, et al, 2015) as researchers have stressed the importance of a balance in gender diversity (Campbell and Mínguez-Vera, 2008).
minority perspective, in which females are less motivated to contribute. Both sides will be discussed.
In countries with low levels of gender equality, the likelihood that women will be discriminated or stifled is higher (Fernandez-Mateo & Fernandez, 2016). As a result of this, the qualifications that women provide, their complementary ‘voices’, or expressions and opinions that could be utilised are being suppressed because they are given less opportunities. In other words, the advantages that derive from an increase in female board members, for instance high-quality decision-making, creativity, or innovation (Carter, et al, 2003; Erhardt, et al, 2003), will not be utilised optimally as females are given less opportunities. On the other hand, in countries with higher gender equality, females are seen as equally valuable in comparison to men. Therefore, in such countries where female board members have an actual ‘voice’, the qualifications and benefits mentioned in the section on ‘Board gender diversity’ will be further appreciated and utilised, and in turn females are given more opportunities to contribute. Thus, the distinctive perspectives that arise from an increase in female board members are actually considered and valued, resulting in innovation (Low, et al, 2015).
In addition to the lack of opportunities and resulting lack of usage of the qualifications and benefits that derive from gender diversity in a country with low gender equality, the motivation of women within a country with such gender norms will also be lower, eventually resulting in lower firm performance. Nielsen & Huse (2010) discovered that the attitude towards gender equality is a substantial factor which affects the way in which female board members contribute to strategic involvement and decision-making. Female board members are less likely to be motivated to contribute to board tasks such as the decision-making process, when they are regarded as less important, unequal, or judged constantly (Armstrong, et al, 2010). Therefore, the possible benefits of greater female representation on firm performance deriving from their distinctive values, as discussed in the section ‘Board gender diversity’, are likely to be diminished when women are not viewed as equal, because they are less motivated and provided less opportunities to contribute to decision-making (Nielsen & Huse, 2010). On the other hand, greater gender equality will increase their motivation to contribute to board-related tasks and thus enhance firm performance, because gender equality is related to higher labour productivity and innovation (Armstrong, et al, 2010).
firm performance is a function of employee opportunity, motivation and ability (Appelbaum, Bailey, Berg & Kalleberg, 2000). The cultural attitudes and norms that determine gender equality influences the degree of tokenism and stereotyping, female representation within the board, and in turn the ability of women to perform at work. As these attitudes differ depending on the country, the benefits of female representation within board are also expected to vary.
Hypothesis development
Board gender diversity can be characterised as a ‘double-edged sword’ because it has positive and negative influences on firm performance depending on the context. However, this thesis emphasises the positive influence of a greater female representation within the board on firm performance, such as high-quality decision-making processes (Carter, et al, 2003), creativity and innovation (Erhardt, et al, 2003), and improved understanding of the marketplace (Randøy, et al, 2006) due to the fresh perspectives and ideas female board members provide (Simpson, et al, 2010). Erhardt et al. (2003) argue that these benefits offset the costs that come with board gender diversity, such as decreased speed of decision-making and greater potential for conflicts (Conyon & He, 2017). Thus, as a result of an increase of female board members, firm performance is expected to increase as well. Therefore, the following hypothesis has been constructed:
Hypothesis 1: There is a positive relationship between female representation within the
board and firm performance.
female representation within the board are also expected to vary. Therefore, the following hypothesis has been developed:
Hypothesis 2: The effect of female representation within the board on firm performance is
higher (lower) when a society’s attitude towards women is positive (negative). Conceptual framework
Figure 2. Conceptual framework
METHODOLOGY
The following section discusses the research design and methods which are necessary to answer the research question. First, the data collection and sample will be discussed, followed by a description of all variables. Finally, the analytical approach to test the relationship will be presented.
Data collection and sample
Development Programme (2019). This database provides information on the gender equality indexes from 1990 to 2017 (United Nations Development Programme, 2019). The United Nations database was chosen because it provides more information relative to the European Institute for Gender Equality (2019), which only includes European countries from the years 2005, 2010, 2012, and 2015.
The sample includes as many publicly listed firms as possible from various countries (Appendix I) in the service industry. Despite the fact that Thomas (2004) maintains that a minimum of 200 units is sufficient for a proper analysis, my aim was to gather as much as possible, because I expect that the final sample size will decrease due to the required conditions. Furthermore, since the companies are publicly listed, an abundant amount of information was available. Additionally, industries from the service sector were selected due to the relatively high representation of women on the boards of service-oriented industries (Harrigan, 1981). The service-oriented industries are derived from the standard industrial classification (SIC) codes 70-89, which are provided by Orbis. Finally, all data for the dependent, independent, and moderating variables was gathered from a ten-year time frame from 2008-2017. Because the goal of this thesis is to investigate the influence of gender diversity on firm performance over time, a panel data analysis was conducted. This corresponds with multiple previous studies (e.g. Adams & Ferreira, 2009; Campbell & Mínguez-Vera, 2008).
Out of all the gathered data from the databases, 884 firms from 36 countries met the required condition of providing data on the dependent, independent, moderating and control variables. There are 24 firms per country on average (Appendix I). This resulted in unbalanced panel data with 4,771 observations and a total of 37,184 board members. It was impossible to create a balanced dataset for the 10-year time frame with enough units, as recommended by Thomas (2004).
Variables
Dependent variable
al, 1997). Thus, in line with other research focusing on firm performance, this thesis uses ROA as a ratio. ROA is the net income divided by total assets (Erhardt, et al, 2003; Carter, et al, 2010; Finkelstein & Hambrick, 1996). In other words, “ROA is an indication of the ability of the firm to produce accounting-based revenues in excess of actual expenses from a given portfolio of assets measured as amortized historical costs” (Carter, et al. 2010, p. 403). Thus, when the ROA is higher, the firm has improved its performance and profitability.
Independent variable
The female representation in the board is the independent variable of this research, which is defined as the number of female directors divided by the total number of directors. In other words, a ratio of the percentage of female directors has been used. This measure has also been employed by many studies (Shrader, et al, 1997; Ossorio, 2017; Ahern & Dittmarr, 2012) because it is regarded as the most inclusive indicator to measure the influence of female representation within the board (Torchia, Calabrò & Huse, 2011). This independent variable is lagged by one period, because as Carter et al. (2010) indicate, the influence of female representation within the board on firm performance becomes clear over time rather than immediately. For example, the board characteristics such as the independent variable from 2012 are related to the firm performance data in 2013.
Moderator
Because there is no previous research on ‘gender equality’, it was impossible to find an ultimate indicator to measure the level of gender equality. Nevertheless, the United Nations Development Programme (2019) and the European Institute for Gender Equality (2019) provide a ranking in terms of gender inequality per country, which is based on an extensive list of indicators. Per year, the gender inequality data does not vary dramatically. However, there is significant variation over a period of a decade. For this reason, time varying gender inequality data provided by the United Nations from 2008 to 2017 were used.
the amount of births per 1,000 women between the age of 15 and 19, while the maternal mortality ratio is constructed from the amount of deaths per 100,000 live births. The empowerment dimension is measured by the indicators “the proportion of adult females and males aged 25 years and older with at least some secondary education” and “the proportion of parliamentary seats occupied by females”. The proportion of females and males with at least a secondary education is defined as the percentage of people with at least a secondary education out of the total population, while the share of parliamentary seats is defined as the percentage of seats held by women. Finally, the labour market is measured by the “labour force participation rate of female and male populations aged 15 years and older”, which is logically constructed from the percentage of females/males participating in the labour force. Based on these indicators, the three aforementioned gender-based disadvantages in three dimensions (health, empowerment, and labour market) are created. These dimensions can be scaled from 0 to 1, where a GII value close to 1 means more inequality between men and women.
This rank makes it possible to compare values to other countries, and the usage of this GII measure consists of a mixture of indicators which together create the index. Therefore, the GII is more accurate and is characterised as an added value rather than a sole measure (United Nation Development Programme, 2019).
Figure 3. Gender inequality index and indicators
Source: United Nations Development Programme (2019)
Control variables
general firm characteristics. Board characteristics include board size and board national diversity, while firm size and firm age are general firm characteristics. The dependent, independent, moderating, and control variables are conditions for the sample. If these variables were not available, the firm was removed from the sample. Furthermore, like the other data from this research, the control variables were gathered from the timeframe 2008-2017, and the board characteristics were lagged one year in comparison to the firm performance data.
Board size is measured as the total number of board members (Conyon & He, 2017;
Ilaboya & Ashafoke, 2017). Board size is one of the control variables which is used in almost every study concerning board diversity and firm performance, because board size influences organisational outcomes. This is due to the fact that larger boards typically contain an abundant amount of diversified resources and knowledge (Hambrick & D’Aveni, 1992), which in turn result in enhanced decision-making (Jacklin & Johl, 2009; Adams & Ferreira, 2009). Additionally, the likelihood that female representation within the board increases as board size increases (Carter, et a, 2010).
Board national diversity has seldom been used as a control variable in prior studies.
Nevertheless, many researchers stress the influence of this variable on firm performance, whereas a board with a higher national diversity translates to higher performance (Caligiuri, et al 2004). This is due to the fact that a diverse board in terms of nationality, when considering the complexities of the decision-making process, has a greater variety of cultural attitudes. This should result in better decisions because the decision-making process requires creativity and innovation (Milliken & Martins, 1996; Sessa & Jackson, 1995). Therefore, it is important to control for this variable. Board national diversity was measured by taking the number of foreign directors divided by the total number of directors.
Firm size has been widely used as a control variable in diversity and firm performance
Firm age was measured by the number of years since the firm’s incorporation (Jackson,
Brett & Sessa, 1991). Because the sample consists of panel data, the year of incorporation was subtracted from the annual report date. This control variable was included because prior literature demonstrates that older firms are more capable of coping with uncertain situations (Jacklin & Johl, 2009). Furthermore, older firms have more experience, and therefore they are more constant and have more knowledge on how to cope with and how to avoid organisational disasters. On the other hand, younger firms might be better able to adapt to a dynamic environment in comparison to older firms (Thornhill & Amit, 2003). In short, firm age is likely to improve the performance of an organisation.
Table 1. Summary of all variables, indicators and data sources
Variables Indicator Data source
Main variables
Firm performance Net income / total assets Orbis Board gender diversity Number of women / total number of
board members
Orbis/BoardEx
Gender equality Gender inequality index Human development reports
Control variables
Board size Total number of board members Orbis/BoardEx Board national diversity Number of foreign board members /
total number of board member
BoardEx
Firm size Total number of employees Orbis
Firm age Number of years since incorporation Orbis
Analytical strategy
this thesis is to investigate the interaction effect of gender equality, some OLS models have to be executed and compared. Additionally, due to the fact that diversity may be endogenous as a result of omitted time-invariant firm characteristics such as corporate culture, fixed effects (FE) estimators are considered (Joecks, Pull & Vetter, 2013; Adams & Ferreira, 2009). For example, in some firms the likelihood of appointing female board members is larger than in other firms, depending on the culture. Additionally, the Hausman test provides statistical evidence that the FE model is more useful than the random effect (RE) model. Furthermore, in all regressions, year dummies are added to take the time fixed effects into consideration. After the OLS regressions were executed, different robustness checks were performed to ensure the robustness of the results. By restricting the sample in distinctive ways, the possibility that the results are based on coincidence or influenced by other factors are prevented (Adams & Ferreira, 2009).
Heteroscedasticity
The data were also checked for heteroscedasticity, meaning that the error terms do not have a constant variance, and in turn the standard errors and parameters might be biased by heteroskedasticity (Breusch & Pagan, 1978). By using a residual plot analysis (Appendix II) and the ‘the Breusch-Pagan/Cook-Wesiberg test’ statistical method (Appendix III), the data were tested. The plot analysis displays a cone pattern, while the statistical test, which tests the null hypothesis indicating the error variances are all equal, has a p-value of 0.000. This signifies that the null hypothesis must be rejected. Thus, both illustrate heteroskedasticity (Breusch & Pagan, 1978). Therefore, all standard errors are corrected for the problem of heteroscedasticity in all regression models by using robust standard errors and clustering the data at the firm level (Joecks, Pull & Vetter, 2013).
Multicollinearity
are strongly related to each other. If multicollinearity occurs, this will affect the standard errors. In this case, the standard errors will become larger, and as a result the likelihood that variables seem to be insignificant and a type II error will be made increases, meaning that the researcher determines that there is no relationship when there actually is (Schmidt, 1992). Therefore, if multicollinearity occurs, the variance inflation factors (VIFs) were considered to determine if multicollinearity is present. The acceptable VIF values are inconclusive, as researchers propose distinctive ‘acceptable’ VIF values. For example, Nachtsheim, Neter, and Kutner (2004) propose a maximum VIF value of 10. Rogerson (2001) states that the VIF should not exceed a value of 5, while Pan and Jackson (2008) propose a maximum value of 4. Because the VIF values contained in Appendix IV did not exceed these proposed values, there is no problem with multicollinearity.
RESULTS
The following section presents the results of all regressions. The descriptive statistics are presented first, followed by the correlation matrix. Next, the OLS regression results of the relationship between board gender diversity and firm performance and the interaction effect of gender equality are discussed. Finally, the results section concludes with the performed robustness checks.
Descriptive statistics
The descriptive statistics of all variables which are utilised in this thesis are displayed in Table 2. Because the sample consists of a time frame of 10 years from 2008 to 2017, there are some missing years, and thus unbalanced data is present.
important to note that the large difference between the minimum value of -99.4 and a maximum value of 92.13 of the ROA indicates that there are still firms which were highly profitable during the specified time frame.
Moreover, the independent variable is female representation within the board. This variable displays a mean of 11.9% with a maximum value of 75%. This average percentage is lower than other gender diversity research focusing on the service sector (Erhardt, et al, 2003 (20%)). In comparison to this research, the higher percentage of females within the board in the study by Erhardt et al. (2003) is striking. Due to mandatory quotas several countries have implemented over the past few years and the increasing importance of the gender equality within corporate boards, a higher percentage of women was expected because of the recent timeframe of this study (Ossorio, 2017). Furthermore, the relatively high maximum value of female representation corresponds with the argument that the service sector is more likely to be female dominated (Harrigan, 1981; Farrell & Hersch, 2005; Erhardt, et al, 2003). Nevertheless, the minimum value of 0 demonstrates that there are firms with no female representation within their boards.
Table 2. Descriptive statistics
Variable Obs Mean Std. Dev. Min Max
Dependent variable
ROA (%) 4,771 -.2305701 17.5164 -99.4 92.13
Independent variable
Board gender diversity (% of female directors)
4,771 .1190918 .1228132 0 .75
Moderating variable
Gender equality 4,771 .8454177 .0907549 .476 .96
Control variables
In terms of the moderating variable of gender equality, the large variance in values of the moderating variable can be explained by the large amount of countries (36) which were included in this study, and thus was expected. For example, Sweden, which is recognised for its prominent position in terms of gender equality, has a gender equality index of .96, while India has an index score of .476.
For the first control variable (board national diversity), the mean is .145, which indicates that on average 14.5% of total board members have a nationality other than the home country of the firm. The maximum value demonstrates that in the firm with the widest variety of nationalities, 80% have another nationality than the home country of the firm. Furthermore, the difference between the smallest and largest board size is quite large, because the smallest board is composed of only 2 members, while the largest board consists of 21 members. Moving on to firm size, which is measured by the total number of employees and converted to the logarithm of employees: This variable has a value between 0 and 13.3820, with a mean of 7.225. Finally, the last control variable to discuss is firm age: Out of a total sample of 884 firms, the youngest firms are less than one year old. The oldest firm ‘Whitbread PLC’ from the UK was incorporated in 1742, which translates to an age of 275 years.
The evolution of female representation and gender inequality (2008-2017)
Figure 4. Evolution of board gender diversity (2008-2017)
Figure 5. Evolution of gender inequality (2008-2017)
Correlation
By using the Pearson correlation test, the correlation between all variables has been tested. The correlation matrix is visualised in Table 3. The correlation lays in a range between -1 and 1. A value of 0-0.1 reveals no correlation between the variables, while a value of 0.1-0.3 corresponds to a medium effect. A value of 0.3-0.5 demonstrates a large effect (Field, 2009). From a value of 0.5 and higher, the rules of thumb have been contradicting in stating from which point the correlation becomes problematic for the analysis. Some state that a correlation from 0.7 and higher is problematic while others state that a value higher than 0.9 becomes problematic (Tabachnick & Fidell, 2007). To ensure there are no highly correlated variables in this analysis, the value of 0.7 will function as the indicator of a problematic correlation.
The first conclusion that can be drawn from the correlation matrix in Table 3 is that most variables are significantly but not highly correlated with one another. The only exception is the strong linear correlation between firm size and board size (0.5431 at the 1% significance level), which indicates that an increase in firm size will result in a likewise increase in board size.
the rules of thumb, this value below 0.1 indicates no correlation. These inconclusive results align with the research of Carter et al. (2010) who find no correlation between board gender diversity and firm performance.
Table 3. Correlation matrix
ROA Board gender diversity Gender equality Board National diversity
Board size Firms size Firm age ROA 1.0000 Board Gender diversity 0.0744*** (0.0000) 1.0000 Gender equality -0.0540* (0.0014) 0.1604*** (0.0000) 1.0000 Board national diversity 0.0084 (0.5628) 0.0630* (0.0000) -0.0919* (0.0000) 1.0000 Board size 0.1589*** (0.0000) 0.1265*** (0.0000) -0.1211*** (0.0000) 0.1846 *** (0.0000) 1.0000 Firm size 0.3506*** (0.0000) 0.1530*** (0.0000) -0.1605*** (0.0000) 0.0832*** (0.0000) 0.5431*** (0.0000) 1.0000 Firm age 0.1563*** (0.0000) 0.1860*** (0.0000) -0.0784*** (0.0000) 0.0700*** (0.0000) 0.2041*** (0.0000) 0.2350*** (0.0000) 1.0000
The same holds for the correlation between gender equality and ROA (-0.0540 at a significance level of 10%), in which the correlation is significant, but according to the rules of thumb there is no correlation. Moreover, the negative correlation contradicts previous research which states that increased gender equality should lead to superior decision-making (Armstrong, et al, 2010). Despite the fact that there is no direct linear correlation between the independent and dependent variable and the moderator and dependent variable, there is a small positive correlation between gender equality and gender diversity (0.1604 at a significance level of 1%), which corroborates the study by Sugarman and Straus (1988), who state that when a firm is located in a country with relative gender equality, this has a positive influence on the female representation on the BOD.
higher ROA. When considering the correlation with gender diversity, all control variables demonstrate a positive and significant correlation. Board size (0.1265), firm size (0.1530), and firm age (0.1860) have a slight correlation, whereas national diversity (0.0630) has no correlation. This indicates that gender diversity will be higher with a larger and older firm with a larger and more nationally diverse board. In terms of gender equality, a slight negative and significant correlation between all control variables and this variable can be seen. Board size (-0.1211) and firm size (-0.1605) display a negative correlation while national diversity (-0.0919) and firm age (-0.0784) display no correlation.
Regression results
The OLS regression is used to test the two hypotheses from this thesis: the relationship between female representation within boards and firm performance and the moderating effect of gender equality. The results from a stepwise OLS analysis are depicted in Table 4, and as already explained in the analytical strategy section and the heteroscedasticity section, in all models the robust standard errors are considered, firm FE are applied, and the sample is clustered at the firm level. Furthermore, in all models, year dummies are used. After this, the validity and reliability of the results were tested by performing a variety of robustness checks.
Table 4. OLS regressions
(1) (2) (3)
VARIABLES ROA ROA ROA
Board gender diversity (BGD) 6.140 -20.30
(3.915) (33.56)
Gender equality (GE) -11.79 -10.35
(30.52) (30.75) BGD x GE 31.54 (39.75) Board size -0.244 -0.283* -0.280 (0.172) (0.171) (0.171) Firm size -1.600*** -1.603* -1.610* (0.396) (0.864) (0.865) Firm age 0.126 0.141 0.125 (0.139) (0.233) (0.238)
Board national diversity -2.232 -2.200 -2.255
(1.927) (2.275) (1.927)
Constant 9.472** 18.41 17.68
(3.987) (20.98) (20.98)
Observations 4,771 4,771 4,771
R-squared 0.012 0.013 0.013
Firm fixed effects Yes Yes Yes
Year dummies Yes Yes Yes
Number of firms 883 883 883
Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1
Model 2 is used to test the first hypothesis: ‘There is a positive relationship between female
representation within the board and firm performance’. When considering Model 2, where the
on decisions (Westphal & Milton, 2000). Additionally, despite the fact that a diverse board fosters innovation and creativity, the decision-making process will be hindered due to contradicting ideas. Due to the contradicting findings, Carter et al. (2010) predict neither a positive or negative relationship. The insignificant results could also be explained by the fact that gender diversity could be endogenous in the regressions. This is discussed in the limitations section of this thesis. Additionally, the direct effect instead of the moderating effect of gender equality is added in this model. Contrary to expectations, gender equality (-11.79) displays a negative and insignificant effect on ROA. This negative relationship could indicate that an indicator for gender equality such as an increase in female participation in the labour market is just a superficial gesture (Low, et al, 2015). Furthermore, the R-squared value has slightly increased to 1.3%, in comparison to Model 1. This could imply that these independent variables partially explain a company’s performance. However, the difference in R-squared values between Models 1 and 2 is extremely small.
Finally, model 3 represents the full regression; the effect of the independent variable on
the dependent variable, including the moderating effect; gender equality. In other words, model 3 analyses the second hypothesis: ‘The effect of female representation within the board on firm
performance is higher (lower) when a society’s attitude towards women is positive (negative)’.
When considering the control variables in all three models, there can be concluded that board size remains negative in all models and only becomes significant in model 2. This negative relationship is in line with previous research (Conyon & He, 2017; Adams & Ferreira, 2009). Furthermore, contrary to previous research stating that larger firms have higher performance (Conyon & He, 2017; Adams & Ferreira, 2009), firm size remains negative and significant, indicating that larger firms show lower performance. Moreover, corresponding to Campbell and Mínguez-Vera’s (2010) research, firm age remains positive and insignificant. Board national diversity remains negative and insignificant in all models.
Robustness check
is depicted in Table 6. Because ROA and Tobin’s Q both measure firm performance but in distinctive ways, this could influence the results. Furthermore, in the final robustness check for this thesis the United Nations data for the gender equality variable will be replaced by data provided by European Institute for Gender Equality (Table 7). For the same reason as explained in the firm performance proxy, all previously mentioned robustness checks were performed using this new data.
Table 5. Robustness check by restricting the sample
(1) (2) (3) (4) (5) (6)
VARIABLES ROA ROA ROA ROA ROA ROA
Board gender diversity (BGD) -24.79 -28.36 -15.66 -45.58*** -35.34 -65.08* (33.90) (49.48) (34.48) (15.61) (40.87) (35.69) Gender equality (GE) -9.247 -25.10 -12.59 4.437 -19.94 -66.88 (30.41) (36.31) (30.76) (24.27) (29.91) (42.61) BGD x GE 36.74 41.88 24.91 59.92*** 54.09 89.13** (39.97) (58.43) (40.64) (19.08) (48.22) (43.92) Board size -0.0916 -0.533** -0.292* -0.0649 -0.397* -0.0173 (0.159) (0.247) (0.159) (0.110) (0.217) (0.196) Firm size -1.424 -1.693* -2.556*** -1.854** -1.451 -2.638*** (0.919) (0.975) (0.840) (0.752) (0.991) (0.905) Firm age 0.133 0.282 0.209 -0.0610 -0.238 0.161 (0.226) (0.286) (0.238) (0.140) (0.326) (0.277)
Board national diversity -2.719 -3.737 -1.658 -1.573 -2.016 -1.708 (2.069) (2.785) (2.194) (2.272) (2.406) (2.203)
Constant 15.82 26.71 25.00 20.46 33.92 69.68**
(22.43) (26.94) (22.47) (16.56) (20.66) (30.37)
Observations 4,438 3,845 4,396 3,356 3,607 2,891
R-squared 0.013 0.016 0.020 0.030 0.014 0.028
Firm fixed effects Yes Yes Yes Yes Yes Yes
Year dummies Yes Yes Yes Yes Yes Yes
Number of firms 815 764 795 680 880 638
Sample type FS>50 FS<10000 FA>5 ROA>0 EFCY BGD>0, <0.5 Robust standard errors in parentheses
*** p<0.01, ** p<0.05, * p<0.1
As displayed in Models 1, 2, 3 and 5 in Table 5, there are no major differences in comparison to the results in the full model (Model 3) in Table 4. In the four aforementioned models, the relationship between board gender diversity and ROA remains negative and insignificant. Additionally, the relationship of the moderating variable stays more or less the same in terms of sign and insignificance. Despite the fact that the signs of the coefficients stay the same, the magnitude of both relationships do vary slightly, which could be explained by the adjustment in the sample. Because these results are relatively comparable to the original model in Table 4, and these have already been discussed, no further comments will be made on these results. Thus, when only considering Models 1, 2, 3 and 5, it is possible to conclude that the results are reliable.
in countries with higher gender equality, females are better able to do their tasks efficiently. Their values and capabilities are appreciated and utilized and thus the possible benefits of gender diversity are used optimally. The findings in this model could be explained by the argument of Armstrong et al. (2010), who state that especially high performing firms have implemented diversity and equality management systems, which allows firms to properly manage gender diversity and equality. This way they are better able to enjoy the advantages of female representation. Even though this corresponds with these results, it is important to mention that gender equality in a country does not reveal anything about a firm’s gender diversity management system; however, it may be one of the underlying processes explaining the findings of these models.
In Model 6, the extremes of board gender diversity have been excluded. These findings are similar to the results of Model 4. Board gender diversity (-65.08) becomes significant at the 10% level. Because the extremes of board gender diversity have been excluded, a positive relationship between the independent and dependent variable was expected because Campbell and Mínguez-Vera (2008) and Adams and Ferreira (2007) indicate that diversity should be in balance to be beneficial. Boards that lack or experience too much heterogeneity will not be able to benefit from, for instance, a healthy amount of fresh perspectives. The likelihood that such boards have to deal with problems is high (Conyon & He, 2017). Despite the fact that the findings do not align with expectations, it does in fact correspond with the studies by Adams and Ferreira (2009) and Shrader et al. (1997), which both found a negative and significant relationship. According to these findings, hypothesis 1 must be rejected. In addition, the moderating effect of gender equality (89.13) has also become significant at the 5% level. One can interpret these results to mean that female representation has a negative effect on firm performance, and that these negative effects will be weakened in nations with a positive attitude towards women. These findings provide substantial support for hypothesis 2.
previous robustness checks, these will not be discussed extensively. The only difference between this model and the models from Tables 4 and 5 is the magnitude of the coefficients, which could be explained by the fact that the sample size has decreased.
Table 6. Robustness check using Tobin's Q
(1) (2)
VARIABLES ROA Tobin’s Q
Board gender diversity (BGD) -4.224 -2.779
(54.50) (2.038)
Gender equality (GE) 3.622 4.672
(50.56) (4.509) BGD x GE 17.96 3.499 (62.31) (2.560) Board size -0.0141 0.00960 (0.274) (0.0174) Firm size -2.932*** -0.0599 (1.066) (0.0468) Firm age -0.111 0.0391* (0.308) (0.0233)
Board national diversity -2.099 -0.491*
(3.785) (0.267)
Constant 19.59 -3.611
(38.20) (3.208)
Observations 1,716 1,716
R-squared 0.027 0.060
Firm fixed effects Yes Yes
Year dummies Yes Yes
Number of firms 263 263
Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1
were able to find evidence for a significant relationship while using Tobin’s Q. The differences in results could be explained by the fact that Randøy et al. (2006) focused their research on Nordic countries, while Carter et al. (2003) used US-based firms. Despite this, because both coefficients for this research are insignificant, hypotheses 1 and 2 are not valid. Finally, it is important to mention that similar to the research of Adams and Ferreira (2009), after using Tobin’s Q, the R-squared value of the model has increased to 0.060.
Finally, the last robustness check replaces the gender equality data provided by the United Nations with the data provided by the European Institute for Gender Equality (Table 7). Both databases have used different indicators to construct their gender equality indices, and therefore a change in results is expected. Because Model 1 in Table 7 functions as a reference point using the United Nations data, which already has been explained extensively, it will not be discussed thoroughly. The most significant difference between this model and Model 3 in Table 4 is that the signs have switched. However, this is expected to be the result of the extreme decrease in sample size, distinctive countries, and time frame.
Table 7. Robustness check by replacing the UN data with the EU data
(1) (2) (3) (4) (5) (6) (7) (8)
VARIABLES ROA ROA ROA ROA ROA ROA ROA ROA
Board gender diversity (BGD) 285.2 201.1** 224.7** 198.9** 203.1* 217.9*** 116.8 165.5 (180.1) (93.53) (95.56) (100.9) (103.7) (82.20) (118.9) (112.8) Gender equality (GE) -62.09 0.196 0.252 0.0650 0.185 0.347 0.220 0.132 (84.29) (0.272) (0.272) (0.502) (0.284) (0.223) (0.307) (0.396) BGD x GE -323.7 -2.835** -3.233** -2.764* -2.904* -3.101*** -1.665 -2.664 (205.2) (1.357) (1.370) (1.471) (1.492) (1.181) (1.732) (1.650) Board size 0.0699 0.0740 0.272 0.126 0.109 -0.00968 -0.0943 -0.0302 (0.467) (0.462) (0.413) (0.692) (0.403) (0.204) (0.577) (0.483) Firm size -0.760 -0.658 -3.238 -0.628 -3.418 -0.984 -6.017 -10.68* (2.587) (2.607) (2.845) (2.915) (2.268) (1.297) (4.219) (5.497) Firm age 0.396 -0.0721 0.0693 0.0189 -0.131 0.169 0.0274 0.232 (0.553) (0.352) (0.354) (0.516) (0.349) (0.216) (0.457) (0.463) Board national diversity 1.333 1.870 4.765 -0.863 8.606 -1.288 1.796 4.768 (7.836) (7.506) (5.503) (9.166) (7.313) (3.723) (8.405) (10.70) Constant 49.27 49.27 4.502 -1.547 14.65 -13.53 25.75 70.85 (65.34) (65.34) (26.98) (36.59) (25.41) (17.43) (34.72) (47.98) Observations 695 695 624 575 643 519 507 398 R-squared 0.009 0.009 0.034 0.012 0.042 0.075 0.066 0.163
Firm fixed effects Yes Yes Yes Yes Yes Yes Yes Yes
Year dummies Yes Yes Yes Yes Yes Yes Yes Yes
Number of firms 289 289 265 251 260 242 287 205
Sample type full UN full EU FS>50 FS<10000 FA>5 ROA>0 EFCY BGD>0,<0.5 Robust standard errors in parentheses
*** p<0.01, ** p<0.05, * p<0.1
Note: FS UN= full sample using UN data, full EU =full sample using EU data, FS= firm size, FA= firm age, EFCY = excluding financial crisis years (2008-2010)
All in all, the opposing findings in Table 7 are likely to be explained by the fact that the United Nations and the European Institute for Gender Equality use different indicators to compose their gender equality indices. Furthermore, only European countries and the years 2005, 2010, 2012 and 2015 are covered in the European Institute for Gender Equality Data. This is just a subsection of the sample is this thesis. Because of this limited coverage of the database, the sample has decreased tremendously. So, the distinctive results between Tables 5 and 7 corresponds to the argument of Ferreira (2015) and Conyon and He (2017), who argue that results are prone to differences in measurements, time frames, and so on.
when only considering the results using Tobin’s Q (Table 6), one can conclude that the results are reliable. Despite the fact that Models 4 and 6 in Table 5 provide valuable findings indicating substantial support for hypothesis 2, these results have to be treated with caution, because the original model (Model 3) in Table 4 determined insignificant results. Caution is also necessary because there are changes in significance and signs between this model and the other models in the robustness checks.
CONCLUDING REMARKS
The final section of this thesis provides a conclusion on all results by discussing related research. After the conclusion and discussion, the managerial implications are described. Finally, the limitations and suggestions for future research are provided.
Conclusion and discussion
The aim of this thesis was to make a contribution to the existing research on board gender diversity and firm performance because many researchers provide contradictory or insignificant findings. Additionally, this thesis is of added value to the IB field, because a new moderating variable ‘gender equality’ has been investigated. Based on a thorough literature review comprised of articles on related topics, a positive relationship between board gender diversity and firm performance was expected, because the mentioned benefits of board gender diversity overshadow the costs that accompany it (Erhardt, et al, 2003). Therefore, the following hypothesis (1) was composed: ‘There is a positive relationship between female representation within the board and
firm performance’. Moreover, there was no previous research available on gender equality as a
moderating variable. However, after analysing related articles a strengthening effect of gender equality on the first hypothesis was expected, and in turn the second hypothesis was composed: ‘The effect of female representation within the board on firm performance is higher (lower) when
a society’s attitude towards women is positive (negative)’. The UET and several notable studies
