MASTER
THESIS
Institutions and the board gender diversity – firm performance relationship:
A cross-country approach.
Theresa Kaufmann Student Number: 3803295 t.kaufmann.1@student.rug.nl
Master of Science International Business and Management University of Groningen, Faculty of Economics and Business
Supervisor: dr. E. Mendiratta (University of Groningen) Co-assessor: dr. S. Castaldi (Copenhagen Business School)
2 ABSTRACT
Much research has recently been conducted on the impact of board gender diversity on firm performance. Despite this increase in contributions, findings are still mixed and highly controversial. Scholars have mainly analysed the relationship between board gender diversity and firm performance in country-specific contexts, thereby failing to consider the potential moderating role of institutional set-ups at the country-level. This study thus aims to reconcile the inconsistency in findings by exploring the moderating effect of three distinct country-level institutions: 1) normative acceptance of women in decision-making positions, 2) gender equality in parental leave policies and 3) hard gender quotas. A cross-country panel analysis covering 1,712 firms in 35 countries indicates that the former two mechanisms do not significantly influence the board gender diversity – firm performance relationship. However, hard gender quotas yield a marginally positive moderating effect, suggesting that institutional mechanisms which directly address board dynamics might be more meaningful in the board gender diversity – firm performance relationship. Further research is needed to elucidate this moderating role in a more fine-grained manner.
3 TABLE OF CONTENTS
INTRODUCTION ... 5
LITERATURE REVIEW ... 8
Board Gender Diversity ... 8
An Institutional Approach ... 12
Normative Institutions ... 13
Regulatory Institutions ... 15
Parental leave policies. ... 15
Board Gender Quotas ... 17
The conceptual framework ... 19
METHODOLOGY ... 19
Data and Sample ... 19
Dependent Variable ... 20
Independent Variable ... 21
Moderators ... 22
Normative institutions ... 22
Parental leave policies ... 22
Gender Quotas ... 23
Control variables ... 23
Board size ... 23
Firm size ... 24
Policy on Board Diversity ... 24
Gender Quotas in Parliaments ... 24
GDP per capita and market capitalization to GDP ... 24
Analysis ... 25
Basic assumptions ... 25
Method of analysis ... 26
RESULTS ... 27
4
Regression Results ... 30
Robustness Test Results ... 31
Overview of findings ... 33
DISCUSSION ... 34
Implications ... 37
Limitations and future research ... 38
Conclusion ... 39 REFERENCES ... 40 Appendix A ... 47 Appendix B ... 48 Appendix C ... 49 Appendix D ... 50 Appendix E ... 51 LIST OF TABLES TABLE 1 VIFs before and after mean centering ... 255
TABLE 2 Desrciptive statistics ... 28
TABLE 3 Correlation matrix ... 29
TABLE 4 Predicting firm financial performance, 2008-2017 ... 322
TABLE 5 Overview of findings ... 333
TABLE 6 Descriptive statistics per country ... 4848
TABLE 7 Hard and soft board gender quotas per country ... 4949
TABLE 8 Robustness test including hard gender quota legislation ... 5050
TABLE 9 Robustness test including board gender diversity as percentage... 511
LIST OF FIGURES FIGURE 1 Conceptual framework ... 199
5 INTRODUCTION
Although women are still highly under-represented in the majority of boards worldwide, a growth of more than 10 percent from 2010 to 2017 in almost all developed countries (Catalyst, 2018) represents an important transformation process towards the inclusion of women on boards. The underlying premise of an increase in female board representation is – besides a general ethical reasoning of gender equality – that women improve decision-making processes with their distinct set of skills and characteristics. Based on the upper echelon theory (Hambrick & Mason, 1984) women are assumed to differ from men in terms of their knowledge backgrounds, cognitive frames and reasoning (e.g. Daily & Dawton, 2003; Ibrahim & Angelidis, 1994) and thus increase information variety on boards. In turn, a greater set of perspectives is said to positively influence firm performance (Byron & Post, 2015).
But while the theoretical value of board gender diversity on firm performance may be perfectly plausible, there is growing uncertainty as to whether this potential can always be realised. In fact, empirical findings on the effect of female board representation on firm performance are mixed and highly controversial. Even though some scholars identify women on boards as important determinants of superior firm value (Campbell & Mínguez-Vera, 2008) and performance (Erhardt, Werbel & Shrader, 2003), others find that board gender diversity has no effect on firm performance (Pletzer, Nikolova, Kedzior, & Voelpel, 2015; Chapple & Humphrey, 2014), or even negatively affects it (Adams & Ferreira, 2009; Darmadi, 2010; Mínguez-Vera & Martin, 2011).
The decidedly indecisive results indicate that the impact of female board representation is still unclear. This also puts the simplistic rationale of ‘higher board gender diversity promotes superior firm performance’ into question. Corporate governance scholars relying on the social identity theory (Tajfel & Turner, 1984) suggest that gender diverse boards may even negatively influence firm performance through social categorization processes (Nielsen & Huse, 2010). The formation of in- and outgroups within boards based on gender cues and gender-specific stereotypes (Eagly & Karau, 2002) could hinder group cohesion and trigger conflicts which may eventually be mirrored in inferior firm performance.
6 level, meaning that it is a social fact which does not depend on the judgement of individuals but can be seen as a collective validation formed by the institutions of a country (Suddaby, Bitektine & Haack, 2017). Whether women are seen as legitimate board members whose contributions are equally considered to those of men depends thus on the institutional set-up of a country characterized by both normative and regulatory mechanisms (DiMaggio & Powell, 1983). Accordingly, this study argues that the inclusion of both normative and regulatory institutions is crucial in determining whether the participation of women in boards promotes or reduces firm performance.
In light of the fact that institutional settings are important determinants of how organizations behave (DiMaggio & Powell, 1983), surprisingly little attention has been paid on the moderating role of institutions in the board gender diversity context. To the best of my knowledge, only one meta-analysis which explores the board gender diversity – firm performance relationship includes the moderating effects of country-level institutions (i.e. Post & Byron, 2015). Arguing that female representation on boards can only have a positive effect on firm performance if the contextual conditions in society are conducive to progressive female roles, Post & Byron (2015) explored the moderating impact of gender equality. Again, findings are mixed at best, as women’s representation on boards is found to positively influence market performance, but not the internal accounting performance of firms in countries with progressive gender attitudes. One reason for the inconsistent findings may be the rather broad measurement of the institutionalized gender equality as this approach ignores the fact that not all aspects of gender equality might be equally relevant in affording legitimacy of women on boards (Eagly & Karau, 2002; Hoobler, Masterson, Nkomo & Michel, 2018). Therefore, this study aims to further advance academic understanding of the moderating role of distinct normative and regulatory institutional set-ups.
7 at men, I will attempt to explore how more gender equal parental leave policies may positively moderate the effect of female board representation on firm performance. Finally, also the impact of affirmative action laws in form of hard gender quotas will be examined. Although quotas aim to create greater gender equality on boards, their very existence underlines the fact that the voice of women on boards was previously not considered to be equal to that of men. Therefore, it will be argued that hard gender quotas undermine the legitimacy of women on boards such that it negatively moderates the relationship between board gender diversity and firm performance. In order to examine the proposed interactions, a sample covering 1,712 leading public firms from 35 countries over the period of 2008-2017 will be analysed to approach the following research question:
What impact do country-level normative and regulatory institutions have on the relationship between board gender diversity and firm performance?
8 LITERATURE REVIEW
This section examines and clarifies the concepts and findings that are most relevant for this research. After covering the most important literature on board gender diversity, firm performance and institutions, the underlying theoretical rationales of the upper echelon theory, the social identity theory and institutional theory are presented, which provide the basis for the formulation of three hypotheses. The conceptual framework, which illustrates the relationship between the relevant firm and country-level factors, completes this section.
Board Gender Diversity
Lately, the investigation of board gender diversity in management and corporate governance literature has gained momentum. Diversity generally comprises the “distribution of differences among the members of a unit with respect to a common attribute” (Harrison & Klein, 2007: 1200). Transferred to the board context, board gender diversity thus signifies diversity with respect to gender-specific demographic characteristics of board members.
As the number of women on boards slowly but continuously increased in the last years (Terjesen & Sealy, 2016), the academic interest in the effects of this development has risen significantly. Driven by ethical reasoning (e.g. Campbell & Mínguez-Vera, 2008), but also argumentatively based on the so-called business case (e.g. Carter, Simkins & Simpson, 2003), scholars have examined the impact of board gender diversity on multiple firm level outcomes. For instance, the link between female board representation and corporate social responsibility (Boulouta, 2013; Byron & Post, 2016), firm reputation (Bear, Rahman & Post, 2010; Brammer, Millington & Pavelin, 2009) or strategic change (Triana, Miller & Trzebiatowski, 2013) have been investigated. Notwithstanding these contributions, the undoubtably most dominant relationship analysis is the one between board gender diversity and firm performance. Generally, scholars in this field of research explore the effect of board gender diversity on both market-based firm performance and accounting-based firm performance (Post & Byron, 2015; Campbell & Mínguez-Vera, 2010). Despite the fact that numerous studies on the relationship between women on boards and firm performance have been conducted, findings are mixed and highly debated, for both market-based and accounting-based firm financial performances.
9 of Spanish public companies (Campbell & Mínguez-Vera, 2008) confirmed a positive relationship between female board representation and firm value. In line with these findings, Carter, D’Souza, Simkins and Simpsons (2010) as well as Ngyen and Faff (2012) found a positive relationship between women on boards and firm performance in the US and Australia, respectively. Furthermore, also Herring (2009) associated board gender diversity with increased sales revenue, higher customer numbers and greater relative profits.
On the other hand, findings indicate that female board representation is linked to inferior firm performance. For instance, Adams and Ferreira (2009) found that the average effect of women on US boards on both market-based performance (Tobin’s Q) and accounting-based performance (ROA, return on assets) is negative, although they manage to positively influence the monitoring task of boards. Consistent with this finding, an Indonesian exploration of women on boards revealed a negative relationship between board gender diversity and firm performance (Darmadi, 2011). Lastly, it was also found that the risk aversion of women on boards of Spanish small to medium sized enterprises had a negative impact on firm performance (Mínguez-Vera & Martin, 2011).
To shed even more confusion, studies by Rose (2007), Hassan and Marimuthu (2014) and Chapple and Humphrey (2014) suggest that board gender diversity has no significant impact at all on performance of Danish, Malaysian and Australian firms. Even the results of several meta-analyses could not reconcile the findings. For instance, a meta-analysis by Post and Byron (2015) indicates that board gender diversity has a slightly positive impact on firm performance including accounting returns, strategy involvement and monitoring processes. Similarly, a recent meta-analysis (Hoobler, Masterson & Nkomo, 2018) supports a direct positive impact of female board representation on firm performance. However, in contrast to these two studies, a third meta-analysis by Pletzer, Nikolova, Kedzior and Voelpel (2015) suggests that the impact of board gender diversity on firm performance is generally insignificant. The inconsistency in findings reveals that the relationship between female board representation and firm performance remains unclear, even though the most commonly used theoretical lens in the board context, the upper echelon theory (Hamrick & Mason, 1984), identifies positive associations with women on boards.
10 have an impact on choices made in top management teams. Thus, the upper echelon theory posits that executives’ cognition processes, formed by distinct knowledge, values and individual experiences, shape decision-making processes which eventually influence corporate strategy and firm performance. Although the upper echelon theory originally exclusively addresses executives in top management teams, it is commonly adopted by scholars examining the effect of board characteristics on firm performance. They argue that non-executives play a crucial role in strategic corporate planning mechanisms. Correspondingly, not only managers’ but also board members’ individual characteristics and cognitive processes determine corporate decision-making processes which eventually influences firm performance.
In the context of board gender diversity, scholars relying on the upper echelon theory predominantly propose a positive effect of women on boards. The theoretical reasoning for a positive impact of female board members on firm performance is based on two main arguments. First, women are assumed to differ from men in terms of their knowledge backgrounds, cognitive frames and reasoning (e.g. Daily & Dawton, 2003; Ibrahim & Angelidis, 1994; Byron & Post, 2016). Women amplify the pool of cognitive resources and thus bring new insights on boards. Consequently, a higher information variety and a greater set of perspectives might widen the scope of attention of board members. Second, it is argued that women positively change the nature of intra-board cooperation, as they show higher levels of commitment (Huse & Solberg, 2006) and communication (Groysberg & Bell, 2013). In combination with the argumentation of broadened human capital, it is thus claimed that the creation of a superior decision-making environment positively influences firm performance.
However, the upper echelon theory is based on the key implicit assumption that women are legitimate members on boards, meaning that their voice and input is heard and (equally to men) considered by the rest of the board. Even though the upper echelon theory recognizes varying levels of co-determination within boards (Hambrick, 2007; Finkelstein, 1992), it does not further elucidate interactive processes and dynamics which influence board performance. This is where the social identity theory (Tajfel & Turner, 1986; Tajfel 1979) comes in.
11 This cognitive process involves three steps (Turner, 1982). First, a cognitive evaluation and allocation of relevant individuals to different social categories takes place. There is a strong empirical evidence that categorization is a forceful and automatic socio-cognitive process that is hard to overcome (Hogg, 2013). The categorization procedure is followed by a process of identification during which individuals assign themselves to one or more groups, based on their social categorization. Lastly, comparative processes define and form inter-group relationships. Tajfel and Turner (1986) point out that individuals tend to favour the own ingroup over other outgroups in order to achieve high levels of self-concept. Besides the tendency of ingroup favouritism, group formations can additionally lead to outgroup aggression. Individuals who perceive themselves as part of the ingroup might behave in a hostile and discriminating manner towards outgroup members. By that, social categorization has the potential to provoke both discrimination and intergroup competition (Tajfel & Turner, 1986).
Which outgroups are perceived as relevant depends on situation-specific mechanisms. Social cognition scholars found that perceptual cues such as distinct visual features are particularly important drivers of social categorization processes (Brewer & Lui, 1989). In the board context, which is typically male-dominated (ILO, 2018) and oftentimes characterized as ‘old-boys-network’ (Perrault, 2015), women are therefore likely to be seen as outgroup members (Ashforth & Mael, 1989). A widely held assumption is that gender-specific stereotypes are a determining factor in the formation of in- and outgroups. Women are typically associated with communal attributes which belong to an affectionate, sensitive and nurturant behaviour (Eagly & Steffen, 1984). Contrary, men assume the role of ambitious, dominant and forceful agents – and are stereotypically prone to take over leadership. Thus, traditional gender roles which depict the “expectations about the actual and the ideal behaviour of women and men” (Eagly & Karau, 2002:574) may in fact stimulate intergroup bias and lead to higher discrimination and non-contribution of women on boards.
12 they could thus either afford or undermine the legitimacy of women on boards. The predominant country-specific approaches adopted by the majority of scholars may therefore be a reason for the present inconsistency in findings, and increased attention should be given to the different institutional frameworks of countries.
An Institutional Approach
This study adapts the institutional approach of new organizational institutionalism (DiMaggio & Powell, 1983), defining institutions as “granted ways of acting, which derive from shared regulative, cognitive and normative frames” (Morgan & Kristensen, 2006:1470). This includes that socially accepted forms of behaviours are embedded in a large system of shared beliefs which covers both regulatory (i.e. coercive) and normative institutions (DiMaggio & Powell, 1983; Suchman, 1995). Originally, normative institutions were primarily anchored in the professional setting (DiMaggio & Powell, 1983), but it is often extended to norms, values and beliefs that are established on a more general societal level (Archibald, 2004). Thus, while coercive institutions are commonly seen as formal rules and regulations that are promoted by states, normative institutions are characterized as norms and practices which assume a “rule-like status in social thought” (Ingram & Clay, 2000) but still have an informal character.
13 Transferred to the board context, legitimacy of women on boards is attained if female participation in decision-making positions is considered as a proper course of action consistent with overall societal norms and rules. That is, women on boards are seen as appropriate when leadership roles fit the role of women in society. A meta-analysis by Post and Byron (2015) indicates that countries with less distinct traditional gender roles are more likely to enhance female contribution on boards. Analysing the results of 140 studies, the scholars found that the effect of female board representation on performance is more positive in countries with higher institutionalized gender equality. However, the positive findings are limited in that they only apply to the market-based, reputational performance but not to internal accounting-based performance of firms. This signifies that overall gender equality positively affects investors’ evaluations regarding high female board representation more than board dynamics themselves (Post & Byron, 2015). Whether gender-related institutional set-ups actually influence board dynamics and categorization processes is still not clear.
A possible explanation why Post and Byron (2015) could not find a significant moderating effect of macro-level gender equality is the broad measurement of it. The scholars relied on the Global Gender Gap Index (World Economic Forum, 2018), which is an overall evaluation of gender equality regarding economic, educational and political participation and access to health-related services. As suggested by Hoobler et al. (2018), not all four pillars of gender equality might be equally important to minimize or emphasize gender stereotypes and their (mis)fit regarding leadership roles. For instance, the extent to which women have access to health care might not be as decisive in forming legitimacy of women in professional settings as political or economic integration. Examining the moderating effect of distinct institutional set-ups which particularly emphasize progressive attitudes towards women on boards might therefore reveal relevant new insights regarding the board gender diversity- firm performance relationship. Therefore, this study includes one normative and two regulatory institutions which I consider to be important factors for the legitimacy of women on boards.
Normative Institutions – the acceptance of women in decision-making positions
14 and the household (Grosvold & Brammer, 2011; Iannotta et al., 2016). Instead, women’s professional potential is recognized, which positively influences the relationship between board gender diversity and firm performance in two ways.
First, normative legitimacy of women on boards may deter the process of social categorization that is based on gender-related characteristics. Even though social categorization is a highly automatic process, the wider normative and legislative context in which the intergroup dynamics exist has a significant impact on it (Hogg, 2013). When there are only few gender-specific prejudices at the societal level and women are normatively accepted and appreciated in economic and political top positions, it is likely that this is reflected in inter-group relationships in corporate boards. That is, while social categorization processes based on gender cues may still persist, the assignment of traditional role-specific attributes which potentially foster tensions between the (male) ingroup and the (female) outgroup might significantly drop in environments in which gender-specific prejudices are rare (Lepore & Brown, 1997). Thereby, intergroup conflicts and isolation could significantly decrease, and a higher cohesion could stimulate decision-making processes. In turn, integrative and efficiently working boards could eventually improve firm performance.
Second, a normative acceptance of women in top positions is associated with an expansion of diverse resources. When women are professionally acknowledged, it is more likely that they increasingly demonstrate labour skills which are suitable for top positions. That is, due to normative acknowledgement of women in top positions, women are incentivized to leverage their distinct knowledge, experiences and values, and board members are more likely to appreciate female contribution (Bilimoria, 2006). The fact that women can incorporate their distinct human capital may positively alter decision-making processes. Not only do they enrich the spectrum of perspectives, which might provide better fit to the diverse set of stakeholders firms are confronted with (Daily & Dalton, 2003), they also potentially stimulate innovation and creativity (Campbell & Mínguez-Vera, 2008). Women’s legitimate informative value might therefore lead to superior decision-making processes, which – again – positively influences firm performance. Therefore, hypothesis 1 posits:
H1: Normative institutions moderate the relationship between a company’s board gender
15
Regulatory Institutions
I include two distinct types of institutional regulations in the study – welfare regulations (i.e. parental leave policies) and affirmative action laws (i.e. gender quotas). Broadly speaking, parental leave policies include regulations regarding maternity leave, paternity leave and parental leave1. These legislations thus refer to those national social policies which aim to legally frame the interception of parenting and economic participation of men and women. Contrary, affirmative laws do not constitute social regulations that frame the societal fabric and welfare of a state – instead they promote special rights to distinct (minority) groups whose positioning in specific situations is perceived to be suppressed and who therefore need special regulatory support. With regard to gender-specific discrimination in top economic and political positions, affirmative laws are implemented in the form of gender quotas.
Pointing to the fact that these two types of regulations are completely different in nature – welfare provisions frame the whole structure of the (welfare) state while gender quotas only address the inferior position on women in distinct economic and political contexts – I argue that they influence the legitimacy of women in powerful economic positions through different mechanisms.
Parental leave policies. Given the fact that welfare-related regulations constitute
important determinants for women on boards (e.g. Grosvold & Brammer, 2011; Iannotta, Gatti & Huse, 2016; Terjesen & Singh, 2008), surprisingly little (i.e. no) research has specifically focused on the moderating effects of such institutions on the relationship between board gender diversity and firm performance. Yet it makes intuitively sense to further elaborate the effect of parental leave policies in this context, as the institutionalized roles of women significantly influence the legitimacy of women in professional settings (Lucas, 2003).
Initially, parental leave policies were exclusively related to the job protection of mothers. The fact that this distinct focus on motherhood legislation has been highly resistant to change (Iannotta et al., 2016) indicates that it is first and foremost women and not men who are institutionally assigned to the responsibility of non-economic childcare activities. Although there is evidence that high maternity leave increases women's employment continuity, it is also
1 Maternity leave refers to those paid weeks of absence from work before and after the birth of a child that are
16 discussed that this tendency promotes gender inequality in the labour market (Ruhm, 1998) particularly in decision-making positions, through several mechanisms.
First, extended labour protection rights may encourage women to defer their return to their jobs (Ruhm, 1998; Baker & Milligan, 2016). Limited labour experience could eventually result in a reduction or even loss of distinct labour skills that are necessary to reach top economic positions (Edin & Gustavsson, 2001). The argumentation of the upper echelon theory that women on boards amplify the pool of eminent human resources is thus difficult to justify if exclusively the development of women’s (and not men’s) labour skills are significantly detracted by non-economic child care activities. Consequently, female human capital depreciation potentially weakens the legitimate claim of women to become part of decision-making boards.
Second, extended maternity leave policies entail the risk of “statistical discrimination against women” (Farré, 2016:52). There is no doubt that, in environments in which primarily women are institutionally incentivised to dedicate themselves to child care, the assigned role of women in society differs considerably from that of men (Eagly & Steffen, 1984). Not only can this be seen as a major reason for the under-representation of women on boards (Iannotta et al., 2016), it is also a significant determinant for intra-board dynamics between men and women. The institutionalised role of women as communal caregivers could have a negative impact on the position of women on boards (Eagly & Karau, 2002). That is, social categorization processes which are based on gender cues may be intensified by the specification of gender attributions. Women's contributions on boards could be perceived as inferior in that they fall beyond their primarily assigned stereotypical role as affectionate caregiver (Eagly & Karau, 2002). The reduced perception of legitimacy might provoke intra-board tensions which eventually impede efficient decision-making.
17 androgynous (Eagly & Karau, 2002). Such trends can positively influence the position of women in top economic positions.
Higher gender equality in parental leave policies might therefore stimulate women’s legitimacy in boards. First, pronounced paternity leave might reduce the tendencies of women to defer their return to their jobs (Farré, 2016). Following this line of thought, a possible lack of appropriate leadership qualities that women are accused of due to prolonged parenting would be reduced. Secondly, when men and women incorporate multiple roles and are not confronted with gender-specific stereotypes, they are not restricted to formerly ascribed tasks (Eagly & Karau, 2002). Thus, women might feel empowered to demonstrate their economic competencies and thereby maximize and enhance the human capital of the board. Accordingly, the institutionalization of gender equal parental leave policies might reduce potential board conflicts based on demographic cues. Similar to the argumentation in H1, stereotypical gender attributions which potentially promote inter-group tensions might decrease with regulatory changes in parental leave policies towards more gender equality. Consequently, board dynamics could improve and make decision-making processes more valuable, which results in superior firm performances. I therefore posit the following hypothesis:
H2: Country-level welfare institutions moderate the relationship between a company’s board
gender diversity and its firm performance. That is, board gender diversity has a positive effect on firm performance in countries in which gender equality in parental leave policies is pronounced.
Board Gender Quotas. Although the gender gap in labour force participation worldwide
18 Black, Jensen & Lleras-Muney, 2018), this market reaction indicates that the principle of quotas might not be conducive to the legitimacy of women on boards due to several reasons.
First, quotas entail the danger that women on boards are not perceived as a promising source of innovation but as a mere projection of politically correct behaviour. Scholars argue that quotas undermine the achievements of women on boards (e.g. Kakabadse et al., 2015; Nielsen & Huse, 2010). Instead of recognizing that women have undertaken commitment and hard work to reach the top of companies, board members could simply label women to “sideline the merit principle that ensures the best person for the job” (Kakabadse et al., 2015:275). Being perceived as an imposed requirement rather than as a valuable and contributing asset, women might reduce their commitment which could negatively influence decision-making processes on boards.
Additionally, quotas might also emphasize ingroup/outgroup formation within boards. Quotas underline the fact that gender equality within boards is not realizable without regulatory intervention of states. Thus, men are likely to consider themselves to be in a more legitimatised and superior position in upper echelons. Although not empirically tested, Nielsen and Huse (2010) argue that affirmative action laws signify a burden to gender equality in that they undermine the value of women through the strengthening of existing prejudices. Outgroup derogation and resulting tensions between men and women could eventually weaken board efficiency and impede the board’s objective to constantly improve its strategic focus. Thus, the delegitimizing nature of gender quotas for women on boards might eventually decrease firm performance.
19
H3: Affirmative action laws moderate the relationship between board gender diversity. That is,
board gender diversity has a negative effect on firm performance in countries in which hard gender quotas are mandated.
The conceptual framework
The conceptual framework visualized in figure 1 illustrates the proposed interconnections.
FIGURE 1
Conceptual framework.
METHODOLOGY
This section comprises all information about the sample, the variables and the method of the study. After giving an overview of the composition of the sample, the dependent and independent variables as well as the moderators and control variables will be described. The section concludes with the explanation of test assumptions and the statistical method.
Data and Sample
In order to create an internationally diverse sample, I have combined several indices, using the S&P Global Index as the basis for the sample. The S&P Global 1200 consists of 7 individual indices (i.e. S&P 500 (US), S&P Europe 350, S&P TOPIX 150 (Japan), S&P TSX 60 Canada, S&P/ASX All Australian 50, S&P Asia 50 and S&P Latin America 40). Covering
Normative Institutions:
- Acceptance of women in decision-making positions
Regulatory Institutions:
- Parental Leave Policies - Hard Gender Quotas
20 1200 leading firms in 30 countries, the index represents nearly 70 percent of worldwide market capitalization and therefore constitutes an adequate base for a cross-country analysis.
However, following Zhang’s (forthcoming) approach, I have also included 19 leading country-specific indices in order to counter the under-representation of the following countries: Australia (ASX 200 Index), Austria (ATX Index), Belgium (Bel 20), Brazil (Bovespa Index), China and Hong Kong (Hang Seng Index), Colombia (COLCAP Index), Denmark (OMXC 25), Finland (OMX Helsinki 25 Index), India (BSE Sensex Index), Indonesia (JSE LQ45 Index), Israel (Tel Aviv 25 Index), Korea (KOSPI 50 Index), Malaysia (KLCI Index), New Zealand (NZX 50 Portfolio), Norway (OBX Index), Philippines (Philippines SE Index), Russia (MICFX Index), Singapore (FTSE Straits Times Index) and Thailand (Thai SET 50 Index).
Nine countries which were part of the initial sample (i.e. Bermuda, Cyprus, Isle of Man, Macau, Malta, Panama, Papua New Guinea, Portugal and Taiwan) have less than 5 firms and are therefore excluded. The final sample captures data on 1712 leading publicly traded firms in 35 countries from 2008-2017.
Data on the dependent as well as independent and firm-level control variables was retrieved from Thomson Reuters’ EIKON database for the period of 2008-2017. EIKON contains global economic, financial and corporate data and thus is a suitable source for firm specific data collection.
Dependent Variable
The dependent variable for this study is firm performance. In general, there are two widely accepted firm financial performance indicators: market-based performance measures and accounting-basedperformance measures.
Market-based performance metrics such as Tobin’s Q, market to book-ratio or market returns encompass external market reactions based on external information and therefore conceptualize firm financial performance as an aggregation of stakeholder expectations and long-term perspectives (Gentry & Shen, 2010). Numerous scholars who explore the effect of board characteristics on corporate performance rely on Tobin’s Q as dependent variable (e.g. Campbell & Mínguez-Vera, 2010; Carter et al., 2003, Adams & Ferreira, 2009; Rose, 2007) as it provides a conclusive long-term measure of firm financial performance.
21 performance context (e.g. Erhardt et al., 2003; Adams & Ferreira, 2009; Hassan & Marimuthu, 2016; Darmadi 2011; Post & Byron, 2015), is an internal indication of how profitable and efficient a company performs in relation to its capital. Unlike market-based performance metrics such as Tobin's Q, ROA reflects how firm specific characteristics and processes influences firm efficiency and profitability rather than external corporate reputation and market positioning. Since the theoretical reasoning of this study promotes the idea that gender diversity on boards affects board dynamics and the nature of decision-making structures, ROA thus appears to be the more appropriate measure of corporate performance and is thus applied in this study.
Independent Variable
The independent variable of this study is board gender diversity. Former studies on board gender diversity have measured women on boards in various ways. Some scholars include board gender diversity as percentage of women on boards (e.g. Campbell & Mínguez-Vera, 2008; Erhardt et al., 2003; Darmadi, 2011), others rely on dummy variables to simply capture the presence of women on boards (Mínguez-Vera & Martin, 2011; Rose, 2007; Ngyen & Faff, 2007). Another frequently applied way of measuring board gender diversity is the Blau’s Index. The Blau’s Index was specifically designed to assess varieties within groups (Harrison & Klein, 2007). The main difference to simple percentages is that its highest value does not correspond to the highest share of women on boards, but to the highest gender equal distribution on boards. Thereby, the Blau’s Index particularly rewards gender equality. The Blau’s Index ranges from 0 to 0.5 and is denoted as:
𝐵𝐺𝐷 = 1 − (% 𝑝𝑤𝑏)2− (% 𝑝𝑚𝑏)² (1),
where
% pwb = percentage of women on boards, % pmb = percentage of men on boards.
22
Moderators
Normative institutions. Normative institutions are summarized in a score which is
composed by three distinct indicators.
The first indicator is part-time employment (PTE) which is the ratio of the share of female part-time employment of the total female employment to the share of male part-time employment of the total male employment. Numerous studies which explore the effect of county normative institutions on board gender diversity have already relied on this figure as it provides useful hints on the distribution of economic and domestic work (Adams & Kirchmaier, 2013; Iannotta et al., 2016). Consequently, it signifies a conclusive indication of how integrated women are in labour market, and which role they are asserted to in professional settings. Information on part-time employment was retrieved from the World Bank database.
The two remaining normative indicators reflect the normative legitimacy of women in top economic and political positions in a more direct way than the PTE. More precisely, the two remaining indicators embody the percentage of women in senior and middle management (which were previously related to board gender diversity (Bilimoria, 2006)) and the percentage of women in parliaments. The primary source for the share of women in senior and middle management was the ILOSTAT database2, while information on women in parliaments was retrieved from the World Bank database3.
The final index which ranges from 0-1 is the averaged sum of the indicators per country.
Parental leave policies. The second moderating variable is parental leave policies
(PLP), which is the ratio of paternity leave policies to maternity leave policies. Maternity leave includes all paid weeks that are available for women both through distinct maternity leave or general parental leave regulations. The composition of paternity leave policies is slightly different in that it covers only the paid paternity and parental leave weeks that are exclusively reserved for fathers and non-transferable to mothers.
Data on both maternity leave and paternity leave were primarily retrieved from the OECD employment database. For a set of countries which were not covered by the OECD database, World Bank’s Women, Business and Law database which provides access to the overview of the legal provisions for individual countries and the ILO Travail legal database
2 Data of countries for which there was no ILOSTAT data available was collected through World Economic
Forum’s Global Gender Gap Reports.
3 Since there was no parliamentary data given for Hong Kong, the Hong Kong Women in Figures Reports were
23 were used4. Despite precise data collection, the occurrence of missing values could not be prevented. Since laws have the property of being constant over a long period of time and change relatively abruptly, missing values were replaced with the last value applicable.
The score ranges from 0-1, whereby a higher number signifies higher gender equality in parental regulations. For two countries an exception to the given calculation had to be made. As Australia did not provide paid maternity, paternity or parental leave from 2008 to 2010 and the US did not provide paid maternity, paternity or parental leave throughout the whole period, the score calculation as described above was mathematically impossible. Seeing the non-existing paid leave policies as an equal treatment of men and women, the missing values for these countries were replaced by 1.
Gender Quotas. The concluding variables within the regulatory environment refers to
gender quotas in publicly traded companies. Generally, one can distinguish between hard and soft quotas. While hard quotas are defined by law, soft quotas are codes of conduct that often contain a non-compliance clause but are not binding (Terjesen & Sealy, 2016). Even though this study primarily alludes hard gender quotas, soft gender quotas are additionally added for the sake of completeness.
Data on gender quotas in publicly traded firms were retrieved from Terjesen, Aguilera and Lorenz (2015), Terjesen & Sealy (2016) and Devnew, Le Ber, Torchia & Burke (2018). The variable is categorically coded and turns 2 if there is a hard gender quota (as of the compliance date), 1 if there is a soft gender quota and 0 if there is no quota at all.
Control variables
Board size. Board size is a common control variable for studies on corporate governance
(e.g. Carter et al., 2003; Adams & Ferreira, 2009; Campbell & Mínguez-Vera, 2008) as it is likely to affect board dynamics and decision-making processes. Although some scholars argue that larger boards improve the access to critical resources, it is commonly suggested that too many board members negatively influence board performance and firm outcomes (Boivie, Benar, Aguilera & Andrus, 2016). For instance, Yermack (1996) found that large boards deteriorate value since they are associated with higher costs of coordination. Since it is argued
4 The countries which were not covered by the World Bank database were Brazil, China, Colombia, Hong Kong,
24 that an increasing presence of women on boards is often accompanied by the growth of board size (Campbell & Mínguez-Vera, 2008; Ben-Amar, Chang & McIlkenny, 2017), it constitutes a critical control variable. In order to match board gender diversity, also board size was time-lagged by one year. Board size which is measured as the total number of board members over the sample period is derived from Thomson Reuters’ EIKON database.
Firm size. In addition to board size, firm size in considered to be an important control
variable. Previous studies have shown that large firms face high pressure to emphasize female representation on boards (Hillman & Shropshire, 2007). Accordingly, firm size has an impact on decision-making processes on boards which eventually are reflected in firm performance. Therefore, it is included as a control variable. Like board gender diversity and board size, the variable is time-lagged for one year.
Policy on Board Diversity. Policy on Board Diversity constitutes the third firm-level
control variable. Companies which autonomously define a governance code are more likely to aim an increase in board gender diversity than firms which do not define such rules at all. In addition, the proactive behaviour towards higher diversity standards also indicates that the role of women on boards is positively rated. Therefore, a positive effect of policies on board diversity is expected. The control variable which is time-lagged by one year is binary coded and assumes a value of 1 if policy on board diversity is implemented and 0 if not.
Gender Quotas in Parliaments. Gender quotas were first applied in the political
environment and thus paved the way to regulations on gender diversity on boards (Terjesen et al., 2009). In many countries, gender quotas in parliaments are important mechanisms in national legislations. Consequently, it is the decision-makers in these national legislations themselves who may experience the effect of quotas in their own context which might influence regulations on quotas in boards. Therefore, a binary variable is included as control variable and is coded as 1 if a quota on the parliament is existent. Data on gender quotas in parliaments were retrieved from the IDEA gender quota database.
GDP per capita and market capitalization to GDP. GDP per capita constitutes the
25 corporate governance. Following Terjesen, Couto & Francisco (2016), also the proportion of market capitalization to GDP is included as a country-level control variable, as advanced financial markets in highly developed countries may be reasonable predictors for advanced corporate governance instruments. Data on both variables were retrieved from the World Bank database.
Analysis
Basic assumptions. Before conducting the final analysis, basic assumptions for
statistical methods were tested.
First, the predictors were tested for multicollinearity. Multicollinearity is defined as the existence of correlations between independent predictor variables (Iacobucci et al., 2016). The linear dependency of variables can cause analytical issues such as biased parameter estimates and has therefore to be eliminated. To detect potential multicollinearity, variance inflation factors (VIF) were used. It is commonly suggested that the VIFs should not be higher than 10 (Tabachnick & Fidell, 2014). The left columns of Table 1 indicate that board gender diversity (BGD) and its interaction term with normative institutions (BGDxN_Inst) clearly exceed this threshold.
TABLE 1
VIFS before and after mean centering.
Before mean
centering VIF 1/VIF
After mean
centering VIF. 1/VIF
BGD 38.10 0.026208 cenBGD 3.44 0.290995
N_Inst 2.27 0.453754 cenN_Inst 2.43 0.411427
BGDxN_Inst 41.90 0.023867 cenBGDxN_Inst 1.75 0.571732
PLP 4.41 0.227000 PLP 1.42 0.704661
BGDxPLP 5.30 0.188794 cenBGDxPLP 2.52 0.396962
Soft Quota 5.24 0.190689 Soft Quota 2.44 0.410564
Hard Quota 6.02 0.197216 Hard Quota 1.60 0.624926
BGDxSoft Quota 6.00 0.166749 cenBGDxSoft Q. 2.11 0.473468 BGDx Hard Quota 5.83 0.171472 cenBGDx Hard Q. 1.60 0.625079
Board Size 1.14 0.713590 Board Size 1.40 0.713590
Policy BGD 1.36 0.732981 Policy BGD 1.36 0.732981
Total Assets 1.39 0.716917 Total Assets 1.39 0.716917
Quota Parliament 1.56 0.640115 Quota Parliament 1.56 0.640115
GDP 2.69 0.372106 GDP 2.69 0.372106
MCap/GDP 1.59 0.629449 MCap/GDP 1.59 0.629449
26 Since the two problematic VIFS belong to two key variables of this study, it is not possible to simply leave them out (which is the easiest way and regularly done). Therefore, another approach called mean centering is applied. Mean centering can be defined as the substraction of a variable’s mean from all observations on that variable in the given data set so that the newly created variable mean turns zero (Iacobucci et al., 2016). This way of reducing multicollinearity has the power to significantly lower correlations between predictor variables but has no effect on analysis outcomes such as confidence intervals (Jaccard, Turrisi & Wan, 1990). As Table 1 reveals, the VIF score significantly dropped after the mean centering such that multicollinearity was eliminated.
Secondly, it was tested whether the residuals were normally distributed or not. Normal distribution of residuals is an important assumption as it alters the significance of the overall model (Tabachnick & Fidell, 2014). It was assessed by a histogram and a standardized normal probability plot (P-P plot). Figure 2 in appendix A shows that the residuals were non-normally distributed. Although the large number of observations (i.e. 14,476) somewhat alleviates the problem of skewness, it is still reasonable to reduce non-normal distribution. A closer look at the dependent variable ROA revealed that the distribution suffered from six negative outliers which were excluded. In addition to the correction of outliers, the logarithm of the control variable TA was calculated. Logarithms are a common approach for reducing skewness (e.g. Hillman et al. (2007); Nielsen & Huse, 2010) and also ensure the basic assumption of normality in this study, as figure 2 in appendix A shows.
Lastly, heteroskedasticity was tested which, if present, has the potential to lower the precision of coefficient estimates (Tabachnick & Fidell, 2014). Testing heteroskedasticity was done by using a modified Wald Test. The null-hypothesis of homoskedasticity was rejected, indicating that the variance of the residuals was not constant and heteroskedasticity present. In order to handle heteroskedasticity, the regression model of choice had to be tested with robust clustered errors. This adjustment also stemmed potential autocorrelation.
Method of analysis. In order to assure that the predictor variables were not biased by
27 can be assessed without potentially disturbing time-invariant firm characteristics (Gelman & Hill, 2006). This also implicates that standard control variables such as industry dummies are not necessary in fixed effects models (e.g. Adams & Ferreira, 2009).
Before performing the analysis, it had to be verified whether the sample was appropriate for a fixed effects analysis or whether its characteristics better corresponded to a random effects model. Typically, a Hausman fixed-vs-random effects test is conducted for this. However, since the Hausman test could not be applied with clustered and robust standard errors in Stata (which was the statistics program of choice), a robust test version (xtoverid) equivalent to the Hausman test was applied (Arellano, 1993). A small p-value (0.000) indicated rejection of the H0, and
thus supported a firm-fixed effects approach.
In addition, a time-fixed effects test was conducted in order to explore if years had an effect on the predicting variables or the outcome variable. The p-value (0.000) was sufficiently small to reject H0 which suggested that time-fixed effects were required. Therefore, the analysis
included both firm and time fixed effects.
Several variables at firm-level had missing observations. I assumed that the missing data were completely missing at random (MCAR) and therefore conducted a complete case analysis. This method uses listwise deletion of cases which miss variables, leaving only complete cases (Pigott, 2001). One criticism about complete case analysis is the reduction of observations. The limitation of observations based on the complete case method was also true for this study as 2,225 observations were excluded. A final observation of 14,402 was still assessed as being sufficiently high to neglect this constraint.
RESULTS
After the discussion of descriptive statistics and the correlation matrix, the test results are considered and explained. Finally, two robustness tests and the outcomes are depicted.
Descriptive statistics
28 value was 0.272.5 The variable normative institution was mean centred as well, resulting in a mean of zero (0.000) with a standard deviation of 0.086 and a range of -0.23 to 0.2726. The average value of parental leave policies was 0.483 with a standard deviation of 0.465 and a range from 0 to 1. Board Gender Quota, which was a categorically coded variable and turned 1 if there was a soft quota and 2 if there was a hard quota, had a mean of 0.4182 with a standard deviation of 0.566, revealing that soft codes are more often applied than hard gender quotas. The average number of board size was 11.067. The fairly high standard deviation of 3.636 indicated that the number of seats vary greatly, from a minimum of 1 to a maximum of 44 seats. The mean of policy on BGD was 0.386 with a standard deviation of 0.487. The logarithm of total assets had a mean of 23.213 with a standard deviation of 1.900, and ranged from 11.855 to 29.900. Quota in parliaments had an average of 0.151 and a standard deviation of 0.358. GDP per capita had an average score of 42,100.434. The extremely high standard deviation of 19,103.46 revealed that the variable varied significantly among countries, from a minimum score of 1001.46 to a maximum of 119,000, indicating that economic activity among the countries in the sample differ significantly. Lastly, the ratio of market capitalization to GDP per capita has an average of 0.006 and a standard deviation of 0.015, ranging from 0.000 to 0.290.
TABLE 2
Descriptive statistics.
Variable Mean S.D. Min Max
ROA 0.057 0.085 -1.373 1.401
BGDAB -0.000 0.157 -0.23 0.272
Normative InstitutionB 0.000 0.086 -0.32 0.186
Parental Leave Policy 0.483 0.465 0.000 1.000
Board Gender Quota 0.4182 0.566 0 2
Board Size 11.067 3.636 1.000 44.000
Policy BGD 0.386 0.487 0 1
Total AssetsC 23.213 1.900 11.855 29.020
Quota in Parliament 0.151 0.358 0 1
GDP per capita 42100.434 19103.94 1001.46 1.19e+05
MarketCap/GDP 0.006 0.015 0.000 0.290
Note: BGD = board gender diversity; BGD, Board Size, Total Assets and Policy BGD are time lagged (t-1); AMeasured as Blau’s Index; BCentred; CLogarithm
5 The average of the uncentred BGD measured as Blau’s Index (not depicted in the table) was 0.228 with a standard
deviation 0.1568 and a range of 0-0.5. The average percentage of women on boards (also not depicted in the table) was 15.1 percent with a standard deviation of 0.121, and ranged from 0 to 85.7 percent.
6 The mean of the uncentred normative institution (not depicted in the table) was 0.376 with a standard deviation
30 The correlation matrix in Table 3 shows that many variables are significantly correlated, albeit with variations in the levels of correlation. Most of the correlations exhibit only a low correlation level with coefficients of less than 0.2, suggesting that the change in one variable only leads to minor changes in another variable (Tabachnick & Fidell, 2007). No correlation coefficient exceeds the 0.6 level which is why the levels of correlations are seen as acceptable for the analysis (e.g. Nielsen & Huse, 2010).
Regression Results
The stepwise regression analysis which comprises 7 models in total is depicted in table 4. Model (1) only includes the control variables. It can be seen that policy on board gender diversity has a small yet highly significant effect on ROA (β = 0.0056, p<.01). In addition, total assets have a negative effect on ROA at the 0.01 significance level. The ratio of market capitalization to GDP constitutes the third control that influences the dependent variable (β = 0.2204, p<.05). The rest of the control variables are not significant.
Model (2) incorporates the main effect of the independent variable board gender diversity into the analysis. Its marginal negative effect on ROA is not significant (β = -0.0069, p=.388) and stays insignificant throughout the rest of the stepwise regression as depicted in model (7) (β = -.0099; p=.324). The impact of the control variables policy on board gender diversity, total assets and the market capitalization ratio to GDP remain similar.
Model (3) introduces the remaining independent variables. Interestingly, the variable normative institution has a slightly negative effect on ROA (β = -0.029, p<.05). The marginal negative standardized coefficient of parental leave policies (β = -0.0026, p=.422) and hard gender quotas (β = -0.0027, p=.250) are not significant. However, soft quotas have a minimal positive effect on ROA (β = 0.0085) at the .01 significance level. With the introduction of the moderator variables, the coefficient of the market capitalization – GDP ratio becomes insignificant. For the rest of the variables, no meaningful changes occur.
31 Model (5) incorporates the moderation of parental leave policies of hypothesis 2 which predicts a positive relationship between board gender diversity and firm performance in environments in which gender equality in parental leave policies is pronounced. The positive coefficient for this interaction term is insignificant (β = 0.0041, p=.792). This result is also confirmed in model (7) (β = 0.0033, p=.835), which is why hypothesis 2 is not supported.
Model (6) includes the interaction term of board gender diversity with affirmative gender laws including both hard and soft gender quotas. Hypothesis 3 predicts a negative effect of board gender diversity on firm performance in environments in which hard gender quotas are legally anchored. Contrary to the hypothesis, hard gender quotas seem to positively influence the board gender diversity – firm performance relationship (β = 0.0217, p=0.055). The positive coefficient of the interaction term in model (7) at the 0.071 significance level supports this result. Therefore, hypothesis 3 is rejected.
Robustness Test Results
In order to test whether the results of the study are structurally valid, two robustness checks are conducted. First, the measurement of the variable gender quota is changed. In the main analysis, hard gender quotas are measured as of the compliance date on which the quota was introduced into boards. To assess whether a positive quota effect could already be determined from the time at which the affirmative law was legislated, the time data is postponed to the legislation date. The regression results for robustness test 1 can be found in table 8 in appendix D. Additionally, a typically applied approach to replace gender diversity measured as Blau’s Index by the percentage of women on boards (e.g. Campbell & Mínguez-Vera, 2011) constitutes the second robustness check. Table 9 in appendix E depicts the results of the second robustness test.
In the first robustness test, no significant changes could be detected for all variables except for the interaction term of board gender diversity and hard gender quotas. That is, the positive coefficient is not significant anymore (β = 0.015, p=.270 in model (6); β = 0.0147, p=.278 in model (7)), indicating that hard gender quotas only significantly influence the board gender diversity – firm performance relationship when they are actually legislated.
32
TABLE 4
Predicting firm financial performance, 2008-2017.
(1) (2) (3) (4) (5) (6) (7)
VARIABLES ROA ROA ROA ROA ROA ROA ROA
BGDAB (t-1) -0.0069 -0.0046 -0.0054 -0.0064 -0.0082 -0.0099 (0.0080) (0.0081) (0.0081) (0.0099) (0.0089) (0.0101) BGDAB (t-1) x Norm. Institutions 0.0529 (0.0494) 0.0447 (0.0499) BGDAB (t-1)
x Parental Leave Policies
0.0041 (0.0154) 0.0033 (0.0160) BGDAB (t-1) x Hard Quota 0.0217* (0.0113) 0.0207* (0.0115) BGDAB (t-1) x Soft Quota 0.0060 (0.0094) 0.0051 (0.0098) Norm. InstitutionsB -0.029*** (0.0092) -0.0215* (0.0123) -0.029*** (0.0091) -0.030*** (0.0092) -0.0230* (0.0123)
Parental Leave Policies -0.0026 -0.0024 -0.0021 -0.0023 -0.0017
(0.0033) (0.0033) (0.0036) (0.0033) (0.0035) Hard Quota -0.0027 -0.0027 -0.0026 -0.0040 -0.0039 (0.0024) (0.0024) (0.0024) (0.0025) (0.0025) Soft Quota 0.0085*** 0.0085*** 0.0085*** 0.0087*** 0.0087*** (0.0018) (0.0018) (0.0018) (0.0018) (0.0019) Board Size (t-1) 0.0003 (0.0003) 0.0002 (0.0003) 0.0002 (0.0003) 0.0002 (0.0003) 0.0002 (0.0003) 0.0002 (0.0003) 0.0002 (0.0003) Policy Board Diversity (t-1) 0.0056*** 0.0058*** 0.0058*** 0.0058*** 0.0058*** 0.0058*** 0.0057***
(0.0020) (0.0020) (0.0020) (0.0020) (0.0020) (0.0020) (0.0020) Total AssetsC (t-1) -0.043*** (0.0039) -0.043*** (0.0039) -0.043*** (0.0039) -0.043*** (0.0039) -0.043*** (0.0039) -0.043*** (0.0039) -0.043*** (0.0039) Gender Quota Parliament -0.0117 -0.0118 -0.0131 -0.0127 -0.0130 -0.0131 -0.0127
(0.0128) (0.0128) (0.0127) (0.0128) (0.0127) (0.0127) (0.0128) GDP 0.0000 (0.0000) 0.0000 (0.0000) -0.0000 (0.0000) -0.0000 (0.0000) -0.0000 (0.0000) -0.0000 (0.0000) -0.0000 (0.0000) MCap 0.2204** (0.1060) 0.2245** (0.1059) 0.1235 (0.1053) 0.1199 (0.1047) 0.1234 (0.1053) 0.1068 (0.1039) 0.1046 (0.1033) Constant 0.091*** 0.091*** 0.098*** 0.097*** 0.098*** 0.098*** 0.097*** (0.009) (0.009) (0.009) (0.009) (0.009) (0.009) (0.009) Observations 14,570 14,507 14,476 14,476 14,476 14,476 14,476 R-squared 0.5908 0.5909 0.5923 0.5923 0.5923 0.5923 0.5924 Number of Firms 1,712 1,712 1,712 1,712 1,712 1,712 1,712
Firm FE YES YES YES YES YES YES YES
Year FE YES YES YES YES YES YES YES
Note: Robust standard errors in parentheses; clustered by firms; *** p<0.01, ** p<0.05, * p<0.1
33 quota interaction term. The complete analysis in model (7) supports this finding (β = 0.0193, p=.161). Thus, even though the two robustness tests suggest a high validity of the majority of the variables, they indicate that the significant moderating effect of hard gender quotas may be only applicable in distinct cases, which will be further explored in the discussion part.
Overview of findings
The empirical analysis yields the following conclusions for each of the hypotheses. First, hypothesis 1 could not be supported, indicating that the normative acceptance of women on economic and political positions does not influence decision-making processes and dynamics on boards. Second, no significant support was found for hypothesis 2. It therefore cannot be confirmed that the degree of gender equality in parental leave policies significantly influence the board gender diversity – firm performance relationship. Lastly, H3 is rejected as hard gender quotas marginally positively moderate the relationship between board gender diversity and firm performance. However, as the robustness tests indicate, the positive effect should be considered cautiously.
TABLE 5
Overview of the findings.
Hypotheses Results
H1 Board gender diversity has a positive effect on firm performance in countries where gender equality in economic and political top positions is normatively more pronounced.
Not supported
H2 Board gender diversity has a positive effect on firm performance in countries in which gender equality in parental leave policies more pronounced.
Not supported
H3 Board gender diversity has a negative effect on firm performance in countries in which hard gender quotas are mandated.
34 DISCUSSION
This section discusses the results and relates them to the hypotheses that were formulated. The answer to the research question is completed by theoretical and managerial implications. Before the final conclusion is drawn, limitations of the study and further research recommendations are pointed out.
Discussion of Results
The primary objective of this paper was to examine the moderating role of country-level institutions in the relationship between board gender diversity and firm performance. More precisely, the moderating role of distinct normative and regulatory institutions were taken into consideration.
Hypothesis 1 explored the normative aspect of institutions and predicted that environments which normatively emphasize the role of women in economic and political decision-making positions positively influence the impact of board gender diversity on firm performance. The moderating effect of normative institutions turned out to be insignificant, indicating that the degree to which women in decision-making positions are normatively legitimized does not have an impact on the relationship between board gender diversity and firm performance in the given sample. In turn, these findings suggest that board dynamics and cooperation between men and women on boards are not influenced by broader socio-structural factors which frame the acceptance of women in political and top economic positions. The results are therefore – contrary to the expectations – in line with the findings by Post and Byron (2015) which could not find a significant moderating impact of overall gender equality on the board gender diversity – accounting-based firm performance relationship in their meta-analysis.
35 that parental leave policies which support a more gender-neutral role distribution between men and women might influence the processes and dynamics in the general workforce, but do not reach the top of companies. In light of both non-significant findings concerning the moderating role of normative legitimacy and parental leave policies, it could therefore be suggested that institutions which determine the socially accepted role of women on a superordinate level are not relevant for the core processes and structures of boards.
Although these findings are in line with former research (Post & Byron, 2015), the measurement of the two institutions which afford legitimacy of women in professional settings on a superordinate level should still be critically evaluated. Normative legitimacy of women in the professional setting was measured as a score covering three distinct indicators (i.e. the percentual share of women in senior and middle management, the percentual share of women in parliaments and the ratio of the share of female part-time employment to the share of male part-time employment). While the representation of women in economic as well as political top positions might adequately reflect the hypothesized institutional legitimacy of women in decision-making positions, the part-time employment ratio might be too vague and not sufficiently be tailored to the particular characteristics of the board context. In fact, previous studies have pointed out that equality of the sexes in the general labour markets do not necessarily correspond to the equality in leadership positions (Eagly & Karau, 2002). The moderating power of an index which includes the ratio of part-time employment per se could therefore be too vague and should be reconsidered. Future research could address this issue by considering additional measurements of normative legitimacy of women in decision-making positions.
