THE INFLUENCE OF NATIONAL CULTURE ON GENDER DIVERSITY IN THE BOARDROOM

(1)

THE INFLUENCE OF

NATIONAL CULTURE

ON GENDER

DIVERSITY IN THE

BOARDROOM

(2)

THE INFLUENCE OF NATIONAL

CULTURE ON GENDER DIVERSITY IN

THE BOARDROOM

Master thesis BA O&MC

By Juliët Nanninga

University of Groningen

Faculty of Economics and Business

June, 2013

Supervisor: C.P.A. Heijes

Co-supervisor: J.S. Gusc

(3)

Abstract

Most existing research focuses on micro and meso-forces to explain the lack of women in the boardroom. Instead, this thesis investigates the influence of national culture on boardroom diversity. For the analysis two cultural models are used, the GLOBE model and the Hofstede model, to see whether different paradigms yield similar results. The results show that several cultural dimensions have a significant influence on board diversity. National culture, as a macro-environmental force, can thus complement the person- and organization-centered explanations for the lack of women in the boardroom. The two cultural models did not yield the same results, highlighting the limitations of these models as well as the complexity of cross-cultural research.

(4)

1 Introduction

A lot has been written about the underrepresentation of women in top management positions. Women are typically confronted with a so-called ‘glass ceiling’. This is an invisible barrier, preventing women from climbing to the top ranks of management (Oakley, 2000). Diverse initiatives are undertaken to increase female participation in the boardroom. Some countries, such as Norway and Spain, have introduced gender parity quotas. Other countries, including the Netherlands and Denmark, have more voluntary initiatives. They have charters that promote the participation of women in management (European Commission, 2011).

Nevertheless, recent figures show that the boardroom is still dominated by men. In 2010 only 12% of the board members at the largest European publicly listed companies were female (European Commission, 2011). With 3% the percentage of female board chairs was even lower. Between individual countries there are significant differences. In Norway about 40% of the board members are female, while in Italy this is less than 5% (European Commission, 2011). Why are there such large differences between these countries? Can cultural differences provide an explanation? Norway is known for its feminine culture, while Italy is known for its masculine culture (Hofstede, 2001).

Literature regarding women on corporate boards (WOCB) cannot give a clear-cut answer to this question. It is relatively recent that the upper echelon studies started addressing gender issues (Carpenter, Geletkanycz & Sanders, 2004). To date, WOCB research has mostly explored individual and firm factors (or: micro- and meso-level factors) in single country studies. These studies explain the lack of women on corporate boards mostly by differences in their functional background, age and tenure (Terjesen, Sealy & Singh, 2009). Individual characteristics can give insight into how specific women advance into the boardroom. However, it does not answer the question why some firms have female directors and others don’t (Hillman, Shropshire & Cannella, 2007).

There are wider external structures and processes that impact the enactment of women’s careers (Terjesen & Singh, 2008). According to Charles (2005:301) ‘all members of society are influenced by norms of gender difference in their preferences, behaviors, self-evaluations, and choices of employers and workers of both sexes’. However, little research addressed these macro-level issues or did cross-country comparisons (Terjesen & Singh, 2008; Terjesen, et al., 2009; Grosvold & Brammer, 2011). Board diversity is usually explored in a typical national context (Grosvold, Brammer & Rayton, 2007; Mateos de Cabo, Gimeno & Nieto, 2012). Few WOCB articles consider antecedents that are external to the individual or the firm (Shore et al., 2009). Researchers call for more research into the environmental context that influences board diversity (Terjesen & Singh, 2008; Ruigrok, Peck & Techeva, 2007; Shore et al., 2009). This thesis answers this call by addressing the role of national culture.

(6)

(2008) showed that culture influences the structure and composition of the corporate board. In this thesis, a macro-level, cross-country comparison between European countries will be done to explore the influence of national culture on boardroom diversity. The research question has been formulated as following:

What is the influence of national culture on gender diversity in the boardroom?

This question will be answered by comparing the culture from different countries with the female board participation rate of these countries. In contrast to most cultural research, not one but two cultural models are used: the Hofstede model and the GLOBE model. The former model is the most widely known cultural model, the latter is the most recent attempt to conceptualize culture at the national level (Parboteeah et al., 2008; Shi & Wang, 2011). It will be interesting to see whether different paradigms yield similar results.

This thesis contributes to the WOCB literature by exploring the influence of national culture on board diversity. It aims at improving the understanding of environmental factors that can be predictive of female board representation. Introducing the national environmental context can complement the person-centered and organization-centered explanations for the lack of boardroom diversity (Terjesen & Singh, 2008). Furthermore, if practitioners do not adequately understand which societal-level forces impede or facilitate women’s leadership, no programs and policies can be designed to address these issues (Bullough, Kroeck, Newburry, Kundu & Lowe, 2012). In order to manage diversity on corporate boards it is important to take national circumstances into account (Ruigrok et al., 2007). Finally, this thesis can be of practical interest for women seeking to advance in the corporate elite and for firms seeking to improve gender diversity (Hillman et al., 2007).

(7)

2 Literature review

This section contains the theoretical background of this thesis. First gender diversity is explored in more detail: what is the use of gender diversity in the boardroom, and what are common explanations for the lack of it. Next, national culture is discussed in more detail: why is it relevant in the context of board diversity, what is known about the relationship between culture and board diversity, and how is culture operationalized in this study. Finally, the hypotheses are formulated.

2.1 Gender diversity

There are two types of arguments in favor of board diversity: ethical arguments and economic arguments (Mateos de Cabo et al., 2012; Carter, D’Souza, Simkins & Simpson, 2010; Grosvold et al., 2007).

The ethical arguments make a normative case for board diversity and see board diversity as a desirable end in itself. Diversity is a positive attribute on its own right, rather than a means to an end. The ethical arguments emphasize that it is inequitable to exclude certain groups from the corporate elite based on their gender, race or other non-performance related characteristics (Mateos de Cabo et al., 2012; Grosvold et al., 2007). The underlying assumption of the economic arguments is that board composition influences board functioning, which influences firm performance (Mateos de Cabo et al., 2012). When a segment of society’s talent is systematically excluded from the board because of their gender, the board composition is not optimal. Systematic non-selection of able candidates damages firm financial performance. Many WOCB researchers investigated the impact of female directors on firm financial performance. The reported results are mixed, although more positive relationships are found in recent studies (Terjesen et al., 2009). Furthermore, a diverse board might better reflect stakeholder, including employees, constituencies (Grosvold et al., 2007).

While the importance of board diversity has been acknowledged, the majority of the boardroom is still male. Women have made only modest gains in terms of directorships on corporate boards (Terjesen & Singh, 2008). In 2010 only 12% of the board members at the largest EU publicly listed companies were female. The percentage of female board chairs was even lower: 3% (European Commission, 2011).

What causes this persistent lack of women in the boardroom? In the literature a variety of explanations can be found. There are different ways to categorize these explanations. For example, Bilimoria & Piderit (1994) distinguish between explanations driven by an experience-based bias (differences in experience) and explanations driven by a sex-based bias (hindering women regardless of their qualifications). Alternatively, the explanations can be grouped into different levels: individual, board, firm or environmental (Terjesen et al., 2009).

(8)

individual’s gender-based perceptions. On the meso-level, explanations look at board and firm characteristics that influence board diversity. Also, meso-level research focuses on the relationship between diversity and firm performance (Terjesen et al., 2009).

According to the human capital theory, the lack of diversity is caused by gender differences in experience, education and skills. A commonly held assumption of board selectors is that women lack adequate human capital for board positions (Terjesen et al., 2009). However, evidence suggests that lack of human capital cannot explain the gender gap. For example, Bilimoria & Piderit (1994) found that after controlling for directors’ experience based characteristics, women were less likely than men to be represented on the more powerful board committees. Furthermore, Hillman, Cannella & Harris (2002) found that female directors are more likely to have an advanced degree than their male counterparts. Even when women possess the right qualifications, it seems harder for them to reach the top.

This might be caused by the different perceptions people have about men and women. Boardroom gatekeepers have views of gender-appropriate behaviors, roles, and expectations that may bias executive selection (Oakley, 2000). In other words, they have a certain perception about appropriate gender roles and stereotypes. Gender roles are the set of norms prescribing the behaviors and activities appropriate for each sex. Very broadly, men are seen as the ‘income provider’ and women as the ‘homemaker’ (Konrad, Ritchie, Lieb & Corrigall, 2000). Gender stereotypes are shared sets of beliefs about the psychological trait characteristics of women and men (Konrad et al., 2000). Men are expected to possess high levels of agentic qualities, such as being independent, masterful, assertive and instrumentally competent. While on the contrary, women are expected to possess high levels of communal attributes, such as being friendly, unselfish, concerned with others and emotionally expressive (Eagly & Karau, 2002).

(9)

evaluate people from the same group higher and make it more difficult for out-group individuals to join the group (Terjesen et al., 2009). In other words: men might prefer other men on the board. Westphal & Milton (2000) showed that the social network of a female director might mitigate out-group bias.

A consistent finding is that the larger the board, the larger the number of female directors (Terjesen et al., 2009). This is in line with the finding that firm size is positively related with board diversity. Other firm characteristics that might influence board diversity are: industry type, firm diversification strategy, and the network effects of linkages to other boards with female directors (Hillman et al., 2007). Also, shareholder distribution might play a role (Terjesen et al., 2009).

(10)

2.2 National culture & WOCB

Culture can be seen as the accumulation of shared meanings, rituals, norms, and traditions that distinguishes members of one society from another (Hofstede, 2001). The extent to which traditional gender roles are pronounced differs per society and is influenced by the values, beliefs and assumptions that are incorporated in a specific culture (Parboteeah et al., 2008). Norms embedded in a society’s culture affect the shape and functioning of organizations, including the composition of their boards (Li & Harrison, 2008). Yet, national culture has received little interest within board gender diversity research. What is known?

While not specifically focused on top management, Parboteeah et al. (2008) found that culture plays an important role in the managerial perception of gender. The authors found that national culture, amongst other institutional forces, influenced the gender role attitudes of managers. Furthermore, while not explicitly addressing gender, Li & Harrison (2008) showed that culture influences the structure and composition of the board (f.i. the percentage of outside directors). Bullough et al. (2012) found that national culture, amongst other institutional forces, influenced women’s participation in political leadership. However, as the authors mentioned, political leadership is different from business leadership and the results might not hold in a business context.

Grosvold & Brammer (2011) did address the relationship between national culture, again as part of the institutional context, and board gender diversity. The results of Grosvold & Brammer (2011) indicated that national culture is relevant for board gender diversity. However, the authors did not investigate dimensions of national culture in detail. They only looked at clusters of cultures (f.i. Anglo cultures and Nordic European cultures), which consisted out of a group of countries. This thesis addresses the relationship in more detail by looking at the national level and including several cultural dimensions per country.

(11)

2.3 Measuring national culture

In this thesis culture is operationalized with the models from Hofstede and GLOBE. These models are chosen because they both have dimensions that could be related theoretically to gender role attitudes (Parboteeah et al., 2008). The Hofstede model is the most widely applied cultural model. The GLOBE model is the most recent attempt to conceptualize and measure culture at the national level (Parboteeah et al., 2008; Shi & Wang, 2011). The GLOBE model has significant conceptual overlap with the Hofstede model (Parboteeah et al., 2008). Not surprisingly, since it was designed to replicate and expand on Hofstede’s work (Shi & Wang, 2011). Due to the similarities, using both models might lead to more robust findings. However, there are some significant differences between the two models. It is therefore also interesting to see whether these different paradigms yield similar results.

Hofstede model

As mentioned earlier, the Hofstede model is the most widely applied cultural model. With this model, Hofstede made a groundbreaking contribution to the field of cross-cultural research (Venaik & Brewer, 2010). Hofstede developed a series of cross-cultural dimensions based on a survey of a large number of IBM employees carried out between 1967-1973 in more than 40 countries. This work was updated and expanded in 1991 and 2001 (Venaik & Brewer, 2010). Cultural (distance) dimensions refer to national / societal values on which nations or societies tend to differ (Tung & Verbeke, 2010). The model has five dimensions (Hofstede, 2001): power distance (PD), individualism versus collectivism (ID), masculinity versus femininity (MAS), uncertainty avoidance (UA) and long-term versus short-term orientation (LTO). The last dimension was added later. The Hofstede model is certainly not without its critics (Venaik & Brewer, 2010). The validity and usefulness of operationalizing culture with a bundle of numerically measured dimensions has been questioned. Critics have argued for greater use of more qualitative analyses of culture (Smith, 2006). Additionally, the substance of the dimensions and their generalizability have been questioned. Also, regarding culture as a national-level characteristic might be problematic. Finally, there is a need to expand the scope of cross-cultural studies intra-nationally and over time (Tung & Verbeke, 2010). Despite these, and other, critics, the Hofstede model is still the dominant model, with over 25,000 citings in literature (Venaik & Brewer, 2010).

GLOBE model

(12)

psychological and behavioral tradition which assumes that cultures should be studied as they are interpreted by their members and that shared values are enacted in behaviors, policies and practices (House et al., 2004). The assessment of cultural values grows out of an anthropological tradition and deals more with how people value the actual practices (House et al., 2004).

The GLOBE study is less criticized than Hofstede, perhaps because it is much more recent and therefore researchers have not yet fully analyzed it (Venaik & Brewer, 2010). However, the complexity demanded of analyses built upon nine dimensions, even more when differentiating between practices and values, might defeat many research designs. For instance, there is a substantial risk of multicollinearity between dimensions (Smith, 2006).

Differences

There are several differences between the two models (Shi & Wang, 2011; Smith, 2006; Hofstede, 2006). While there are overlapping dimensions, there are differences in the conceptualization of these dimensions. These differences will be explained later on, when the hypotheses are formulated. Furthermore, the GLOBE model consists out of more dimensions, each measuring two aspects of culture.

There are also methodological differences between the models. First, one person conducted the Hofstede research, whereas there were 170 researchers involved with the GLOBE research. Second, the GLOBE research surveyed 951 organizations, compared to only one (IBM) in the Hofstede research. By surveying different organizations, the GLOBE research addressed multiple industries. Furthermore, the GLOBE research focused on managers, while Hofstede looked at a broader group of employees. Additionally, the data used in the GLOBE study was new, whereas Hofstede re-analyzed existing data. Finally, the GLOBE research specifically focused on managers, whereas the Hofstede research did not (Shi & Wang, 2011).

(13)

2.4 Hypotheses

Following Parboteeah et al. (2008), not all cultural dimensions from both models are included in this research. Only the dimensions that are, according to theory, most relevant to the topic are included. Furthermore, researchers have found that some of the cultural dimensions are highly correlated. For example, there is a correlation between the power distance and individualism dimension from Hofstede. These dimensions were not empirically separable (Simith, 2006). Not including all dimensions reduces the risk of multicollinearity between the dimensions (Smith, 2006).

The question is now: which cultural dimensions from the two models are relevant when looking at boardroom diversity? Following Parboteeah et al. (2008), the following dimensions are selected from the Hofstede model: power distance, uncertainty avoidance and masculinity. Individualism is excluded due to its correlation with power distance (Smith, 2006). Additionally, it does not reflect the hierarchical or status nature of a society (Parboteeah et al., 2008). Finally, long-term orientation is also excluded. Long-term orientation was not in Hofstede’s original model and most researchers use the four ‘original’ dimensions (f.i. Li & Harisson, 2008; Parboteeah et al., 2008).

From the GLOBE model the dimensions power distance, uncertainty avoidance, assertiveness and gender egalitarianism are chosen. The main reason to select these dimensions is that they most closely resemble the dimensions chosen from the Hofstede model. This makes comparison possible. The selected dimensions are conceptually related and empirically correlated with the Hofstede’s dimensions (Leung, Bhagat, Buchan, Erez & Gibson, 2005). Another reason to exclude future orientation and performance orientation is because they are correlated with uncertainty avoidance (Smith, 2006). Furthermore, Parboteeah et al. (2008) excluded humane orientation because the included cultural dimensions have a parallel theoretical logic for links with traditional gender role attitudes. Finally, no relationships were expected between the collectivism dimensions and traditional gender role attitudes (Parboteeah et al., 2008).

In the Parboteeah et al. (2008) article the same cultural dimensions from both cultural models are used. Parboteeah et al. (2008) argued that these dimensions represent a more conservative and masculine dominated hierarchical society. The authors argue that such a society seems more likely to preserve traditional gender role attitudes (Parboteeah et al., 2008). Below the selected dimensions, and their proposed relationship with board gender diversity, are described in detail. For each selected dimension a hypothesis is formulated. When possible, the hypothesis for a dimension from both models is combined.

Power distance (PD)

(14)

In high power distance countries people accept inequalities as a legitimate basis of relationships, and the common perception is that power differences are basic to proper societal functioning (Parboteeah et al., 2008). High power distance societies tend to be more traditional, with rigid hierarchies and protocols. This may restrain the advancement of women into positions of authority (Caligiuri & Tung, 1999). In traditional societies women are typically lower in the hierarchy (referring to the status characteristics theory earlier mentioned: they are a low-status group), and people are more likely to accept such inequalities (Hofstede, 2001; House et al., 2004). In contrast, in low power distance societies, individuals tend to react negatively when they feel treated unfairly, and are more likely to find ways to minimize such inequalities (Lee, Pillutla & Law, 2000). Independent action by less-powerful actors (in this case: women) is valued and encouraged (Li & Harrison, 2008).

The expectation is thus that power distance is negatively related to the presence of women on boards. There is evidence that the power distance dimension of the Hofstede model and the GLOBE model are correlated (Leung et al., 2005). The first hypothesis, expected to hold for both models, is therefore formulated as following:

Hypothesis 1: Power distance is negatively related to gender diversity in the boardroom.

Uncertainty avoidance (UA)

(15)

focused exclusively on following of the rules. The measurement for Hofstede’s UA dimension was broader, it also included other items such as the respondents’ feelings of stress at work and the respondents’ intentions to stay with the company (Hofstede, 2006). The second hypothesis is formulated as following:

Hypothesis 2: Uncertainty avoidance is negatively related to gender diversity in the boardroom.

Masculinity (MAS), Assertiveness & Gender egalitarianism

The masculinity dimension from Hofstede and the GLOBE equivalents, gender egalitarianism and assertiveness, are expected to relate to traditional gender roles because they directly tap into societal values regarding gender roles (Parboteeah et al., 2008). Hofstede’s masculinity dimension deals with the distribution of roles between genders in a society. The assertive pole has been called ‘masculine’ and the modest, caring, pole ‘feminine’ (Shi & Wang, 2011). A masculine culture prefers achievement, heroism, assertiveness and material rewards for success, leading to a more competitive society. A feminine culture is more consensus-oriented and it favors cooperation, modesty, caring for the weak and quality of life (Hofstede, 2001). By definition a masculine society is ‘a society in which social gender roles are clearly distinct’ (Hofstede, 2001:297). The expectation is therefore that this dimension is positively related to traditional gender role attitudes (Parboteeah et al., 2008), and thus negatively related to the percentage of women in the boardroom. The third hypothesis is formulated as:

Hypothesis 3: Masculinity is negatively related to gender diversity in the boardroom.

The masculinity dimension is not similar in the GLOBE model and therefore additional hypotheses need to be formulated. In Hofstede’s model the dimension is a continuum, ranging from masculinity to femininity (Hofstede, 2001). The GLOBE researchers argued that the dimension should not be a continuum, but that masculinity and femininity should be two different cultural dimensions. They called these dimensions assertiveness and gender egalitarianism (House et al., 2004).

Assertiveness refers to the degree to which individuals in a society are assertive, confrontational and aggressive in social relationships (House et al., 2004). Similar to the MAS dimension from Hofstede, high assertive societies are characterized by more masculine values and norms, suggesting societal attitudes linked to more traditional gender role attitudes (Parbooteeah et al., 2008). The expectation is that in high assertive societies there are fewer women in the boardroom. This leads to the following hypothesis:

Hypothesis 3a: Assertiveness is negatively related to gender diversity in the boardroom.

(16)

boards, since role differences, or role incongruity, are minimized. This leads to the following hypothesis:

Hypothesis 3b: Gender egalitarianism is positively related to gender diversity in the boardroom.

In Figure 1 the hypotheses are summarized in a conceptual model. For each dimension is mentioned whether the expected relationship with board diversity is positive or negative.

(17)

3 Data

To test the hypotheses a database has been created. This thesis is focused on European countries. This is done so that the data for the dependent variable could be obtained from a single database. If non-European countries were included, this data had to be collected from other databases. Measurement of variables might be different there. Within Europe there is enough ‘diversity’ between countries with regards to boardroom diversity. As mentioned in the introduction in Norway 40% of the board is female, compared to 5% in Italy.

Table 1 shows which countries are included. Starting point was including all the countries

from the European Commission (EC) database. Unfortunately, not for all countries values for the cultural dimensions were available. These countries were excluded. As can be seen in Table 1, this leads to a different sample size for the two cultural models. The final sample for the Hofstede model was larger than for the GLOBE model. This is not a problem when the two models are analyzed separately. However, when two models with a different sample size are compared, bias might occur in the results. Therefore, two Hofstede databases are created: one for the separate analysis (includes all 28 countries) and one for the comparison with the GLOBE model (includes the same 17 countries as the GLOBE database).

Table 1. Sample selection procedure Procedure Hofstede GLOBE

Countries included in EC-database

33 countries 33 countries

Minus: Countries without

values for cultural dimensions available

5 countries

Cyprus, Iceland, Latvia, Lithuania, Macedonia

16 countries

Belgium, Bulgaria, Croatia, Cyprus,

Czech Republic, Estonia, Iceland, Latvia, Lithuania, Luxembourg, Macedonia, Malta, Norway, Republic of Serbia, Romania, Slovakia

Final sample 28 countries

(266 country-year observations)

Austria, Belgium, Bulgaria, Croatia, Czech Republic, Denmark, Estonia, Finland, France, Germany, Greece, Hungary, Ireland, Italy, Luxembourg, Malta, Norway, Netherlands, Poland, Portugal, Republic of Serbia, Romania, Slovakia, Slovenia, Spain, Sweden, Turkey, United Kingdom

17 countries

(167 country-year observations)

(18)

Below, for each included variable is described how their data is collected and how they are measured.

Dependent variable

The dependent variable of this research is board gender diversity. Board diversity is measured, at country-level, by the percentage of members on the board of directors (or for countries with separated supervisory and management functions: the supervisory board) that is female. The data for this variable is obtained from the EC database ‘Women & men in decision making’.1_{This database contains, at country-level, female}

board participation rates for 33 European countries, and is updated annually. To calculate the participation rate for a country, the European Commission only looks at the situation in the largest publicly listed firms of that country. More specific: it looks at the firms that are members of the selected country’s blue-chip index This is an index maintained by the stock exchange covering the largest companies by market capitalization and/or market trades.

Independent variables

The independent variable of interest is national culture. As described earlier, culture is operationalized with two cultural models. From Hofstede the following dimensions are selected: power distance, uncertainty avoidance and masculinity. The dimensions consist out of a scale with values ranging from 1-120 (low to high). For each dimension, a country can be positioned somewhere on that scale. The score on a dimension is different for each country.

From the GLOBE research the power distance, uncertainty avoidance, assertiveness and gender egalitarianism dimensions are used. The measurement of these dimensions is somewhat different from Hofstede. The dimensions are measured using a scale with values ranging from 1-7. Furthermore, for the included countries, each GLOBE dimension contains not one but two scores. Each dimension measures two aspects of culture: values and practices. In this research, both aspects are included. Finally, the GLOBE research divided Germany into East and West Germany. Hofstede did not distinguish between East / West Germany (Hofstede, 2001). To eliminate this, the weighted-average of the scores for entire Germany have been calculated based on the number of inhabitants in East / West Germany at the time of the GLOBE data collection (1997). 2

(19)

benchmarks national gender gaps on economic, political, education and health criteria. It is calculated yearly. The GGGI is built out of four sub-indexes (World Economic Forum, 2012):

1. Economic participation and opportunity – this sub-index captures three concepts: the participation gap (measured by differences in labor force participation rates), the remuneration gap (measured by the ratio of estimated female-to-male earned income and by the ratio of wage equality for similar work) and the advancement gap (ratio of female legislators, senior officials and managers over male value and ratio of female professional and technical workers over male value).

2. Educational attainment – this sub-index measures the gender gap in access to education with ratios of women to men in primary-, secondary- and tertiary-level education. Furthermore, the longer-term view of a country’s ability to educate women and men equally is captured by a ratio of the female to male literacy rate. 3. Health and survival – this sub-index captures gender differences in health. First, it

includes the sex ratio at birth. This is done to capture the ‘missing women’ phenomenon prevalent in countries with a strong son preference (World Economic Forum, 2012). Next, it includes a measure for the gender gap between men and women’s healthy life expectancy.

4. Political empowerment – this sub-index measures the gap between men and women at the highest level of political decision-making by the ratio of women to men in minister-level positions and the ratio of women to men in parliamentary positions. Also, the ratio of women to men in terms of years in executive office (prime minister or president) for the last 50 years is included.

The GGGI is chosen for several reasons. First, choosing an index limits the number of control variables. Furthermore, the index covers all the macro-environmental variables used in the Terjesen & Singh (2008) research. As mentioned earlier, Terjesen & Singh (2008) explored as one of the first the relationship between the macro-environment and the presence of women on corporate boards. The authors addressed the influence of social (measured by the percentage of women in legislature, senior official and management positions), political (measured by the year that the first women was elected to political office) and economic structures (measured by the ratio of earned income by men and women).

(20)

The connection between culture and economic development, often measured by GDP per capita and the human development index (HDI), has often been mentioned in literature (Yeganeh & May, 2011). The GGGI measure itself eliminates direct impact of absolute levels of the included variables. For example, it does not look at the overall level of education, or overall level of income in a country. However, the GGGI is correlated with country level competitiveness, GDP per capita and human development (World Economic Forum, 2012). For this reason the ‘absolute level’ measures GDP per capita and HDI are not included in this research.

To improve the robustness of the findings and to control for time effects, the dataset covers multiple years (2003-2012). For most countries the data for the dependent variable was available over all these years. However, there were some missing country-year observations (14 in the Hofstede database and 3 in the GLOBE database). In total 266 year observations were included in the Hofstede dataset and 167 country-year observations in the GLOBE dataset.

In the analysis ‘year’ is included as a control variable. This variable contains a group of 9 dummy variables, since the dataset covers 10 years. Values for the independent variable (percentage of female directors) and control variable (GGGI) vary, only a little, per year. Values for the independent variable (the cultural dimensions) remained the same per country for the included years, since culture is a relatively stable construct (Hofstede, 2001).

In Table 2 all the variables and their definitions are summarized.

Table 2. Variable definitions Dependent variable

%FEMALE Percentage of board participants that is female

Independent variables

Hofstede (values ranging 1-120)

PD Power distance

UA Uncertainty avoidance

MAS Masculinity

GLOBE (values ranging 1-7)

PD_val Power distance values PD_prac Power distance practices UA_val Uncertainty avoidance values UA_prac Uncertainty avoidance practices AS_val Assertiveness values

(21)

4 Methodology

To test the hypotheses a database was created, which consists out of the variables mentioned in the previous chapter. County-year observations with missing data on the cultural variable were directly excluded. The remaining cases with missing data were excluded pair wise during the actual analysis. Before the actual analysis, the descriptive statistics were calculated (mean, median, standard deviation, minimum and maximum). Three correlation matrixes were made: one for the Hofstede model, one for the GLOBE practices model and one for the GLOBE values model. The correlation matrix gives indications about relationships between variables. Furthermore, it gives indications about possible multicollinearity. The cut of correlation coefficient, as an indicator for multicollinearity, was set at 0,5 (Cohen, 1988). Additionally, collinearity diagnostics were performed in the actual analysis. Both VIF values and tolerance levels were calculated. VIF values should be less than 10 (Myers, 1990) and tolerance levels more than 0.2 (Menard, 1995).

To get the results, a multiple regression analysis was performed, since there are multiple variables that have an influence on one dependent variable (Blumberg et al., 2011). For each cultural model a separate regression analysis was done. Since GLOBE measures culture in two different ways there are three regression models in total. Furthermore, while there is one regression model for Hofstede, this model will be tested with two different databases. As mentioned in the previous chapter, the sample size of the Hofstede database was larger than the GLOBE database. This could bias the results when the two models are compared. To mitigate this bias the Hofstede model is analyzed with two databases: one database for the separate analysis (includes 28 countries) and one database for the comparison with the GLOBE model (includes the same 17 countries as the GLOBE database). Below the specification of the Hofstede regression model is shown (variable definitions are already described in Table 2):

%FEMALEi = i + 1 PDi + 2 UAi +3 MASi + 4 GGGIi +

∑

YEARi +



i

The GLOBE model looks similar, only the names of the cultural dimensions included are different. Below the GLOBE model is specified (not distinguishing between values / practices):

%FEMALEi = i + 1 PDi + 2 UAi +3 GEi +4 ASi + 5 GGGIi +

∑

YEARi +



i

The regression analyses were performed in SPSS. First a regression with only the control variables was run. Next, the cultural measures from a particular model were included. This was done to see whether the overall R2_{statistic improved significantly by adding the}

cultural dimensions. Before looking at the results, it was tested whether the assumptions required for regression analysis were met (Keller, 2009):

(22)

2. Normality of the error distribution

3. Independence of the errors, or: no autocorrelation 4. Homoscendasticity or: constant variance of the errors.

Linearity and homoscendasticity were confirmed with the help of a scatter plot of the residuals versus the predicted values. Normality was confirmed with the help of a histogram and normal P-P-plot of the regression-standardized residual. The scatterplots, histograms and P-P-plots are included in Appendix 1.

Independence was confirmed with the Durbin-Watson test. The value of this test should be around 2 (Keller, 2009). Because the dataset contains information about a single country for multiple years, the risk of autocorrelation exists. A preliminary regression analysis indeed confirmed that the value from the Durbin-Watson test was too low. To control for this a lagged version of the dependent variable has been included in the analysis (Keller, 2009). After including this lagged variable (called LAG), the values from the Durbin-Watson test were close to 2 in all the regression models. The precise values derived from the Durbin-Watson test are mentioned in the next chapter.

Finally, outliers in the data were detected. Values larger than ± 3 standard deviations were considered to be an outlier. The outliers were Winsorized at ± 3 standard deviations. Almost no outliers were present. Just a few cases of the independent variable were Winsorized. Furthermore, during the actual analysis overly influential cases were detected with the help of the ‘Cooks Distance’ measure. According to Cook & Weisberg (1982) absolute values greater than 1 may be cause for concern. During the analysis it became clear that none of the variables had a value greater than 1.

(23)

5 Empirical results

In this section the empirical results are presented. First the descriptive statistics are mentioned, including a correlation matrix for each cultural model. Next, for each cultural model the results of the regression analysis are described. There are two tables for each model: one with the overall summary of the regression model and one with the coefficients of the model.

5.1 Descriptive statistics

Table 3 shows the descriptive statistics. For each included variable the number of

observations, mean, median, standard deviation, minimum and maximum value are displayed. The descriptive statistics were calculated before any Winsorizing. For the Hofstede data the ‘large’ sample is used. The dummy variables for year were omitted from this overview.

Table 3. Descriptive Statistics

N Mean Median Std. Dev. Minimum Maximum

%FEMALE 266 12,199 11,000 7,529 1,0 44,0 PD 266 52,120 56,000 21,712 11 104 UA 266 71,450 75,000 23,219 23 112 MAS 266 47,240 45,000 24,674 5,00 110 AS_prac 167 4,227 4,230 ,343 3,41 4,71 GE_prac 167 3,556 3,620 ,319 3,02 4,02 PD_prac 167 5,193 5,320 ,414 4,14 5,68 UA_prac 167 4,384 4,250 ,694 3,26 5,36 AS_val 167 3,580 3,590 ,466 2,68 4,61 GE_val 167 4,883 4,840 ,244 4,46 5,20 PD_val 167 2,629 2,570 ,225 2,23 3,19 UA_val 167 4,279 4,500 ,559 3,34 5,16 GGGI 242 ,7124 ,7018 ,0529 ,5711 ,8451

In Table 4, the correlation matrix for the Hofstede model is shown. For each variable the Pearson Correlation Coefficient is displayed. After each coefficient the corresponding significance level is shown. All correlations between the independent variables and control variables higher than 0,5 are marked bold. Again, the year dummies are omitted from this overview.

Table 4. Correlations Hofstede Model

%FEMALE PD UA MAS GGGI

%FEMALE 1 -,093 -,415** -,419** ,635**

PD 1 ,579** ,245** -,559**

UA 1 ,151* -,615**

MAS 1 -,449**

GGGI 1

(24)

Table 4 shows that the sign of all included variables is as predicted. Power distance,

uncertainty avoidance and masculinity are negatively related to the percentage of female directors. Furthermore, GGGI is positively related to the percentage of female directors. All correlations, except power distance, with the dependent variable are significant. In Table 5, the correlation matrix for the GLOBE practices model is displayed. All variables are significantly correlated with the dependent variable. Surprisingly, the sign of the correlation coefficient for the uncertainty dimension is in the opposite direction than expected.

Table 5. Correlations GLOBE practices model

%FEMALE AS_prac GE_prac PD_prac UA_prac GGGI

%FEMALE 1 -,338** ,352** -,315** ,484** ,635** AS_prac 1 -,227** ,235** -,191* -,412** GE_prac 1 -,267** -,030 ,196* PD_prac 1 -,520** -,443** UA_prac 1 ,753** GGGI 1

** Correlation is significant at the 0.01 level (2-tailed) * Correlation is significant at the 0.05 level (2-tailed)

The final correlation matrix is shown in Table 6. This is the correlation matrix for the GLOBE values model. Interesting is that the signs of the assertiveness dimension and the power distance dimension are in the opposite direction than predicted. For the other variables, the signs are as predicted. Also interesting is that for two of the five included variables, the correlations with the dependent variable are insignificant.

Table 6. Correlations GLOBE values model

%FEMALE AS_val GE_val PD_val UA_val GGGI

%FEMALE 1 ,152* ,048 ,074 -,332** ,635** AS_val 1 -,010 ,019 ,334** ,245** GE_val 1 ,019 -,513** ,406** PD_val 1 -,014 ,038 UA_val 1 -,624** GGGI 1

(25)

5.2 Regression analysis

First the regression analysis for the Hofstede model was performed. The large Hofstede dataset was used for this (including 28 countries). In Tabel 7a, the overall model summary is given. Including the cultural dimensions adds significantly to the explanatory power of the model (∆R2_{is 0,023, 1% significance level). Furthermore, the score for the Durbin}

Watson test is close to 2, showing no autocorrelation problems.

In Table 7b the coefficients of the Hofstede regression model are displayed. Year dummies and the lagged variable are excluded from this overview. All VIF values and tolerance levels stayed below 10 and above 0.2 respectively, indicating no multicollinearity problems.

Table 7a. Summary Hofstede model

Model R R Square Adjusted R Square Std. Error of the Estimate Change Statistics Durbin-Watson R Square Change F Change df1 df2 Sig. F Change 1 ,876 ,768 ,757 3,475 ,768 69,211 11 230 ,000 2 ,889 ,791 ,778 3,323 ,023 8,193 3 227 ,000 1,758

Model 1: Controls only (Year dummies, LAG, GGGI)

Model 2: Controls + Cultural dimensions Hofstede (MAS, UA, PD)

Table 7b. Coefficients Hofstede Model

Unstand. Coefficients Stand.Coefficient

t Sig. Tolerance VIF

B Std. Error Beta (Constant) -20,763 5,667 -3,664 ,000 PD ,062 ,014 ,192 4,453 ,000 ,496 2,017 UA -,040 ,013 -,132 -3,030 ,003 ,488 2,049 MAS -,025 ,010 -,089 -2,481 ,014 ,720 1,388 GGGI 38,956 6,825 ,293 5,708 ,000 ,351 2,850

(26)

for this, the regression analysis for the Hofstede model is repeated with a smaller sample, including the same 17 countries as the GLOBE dataset, to see whether the results change. Using this smaller sample yielded similar results. At a 5% significance level the PD and UA dimensions were negatively related to %FEMALE. The PD dimension was again positively related to %FEMALE, however in this case not significant. Since this additional analysis resulted in no significant differences, the tables are not included here but can be found in Appendix 2.

Table 8a shows the overall results for the GLOBE practices model. Again, adding the

cultural variables improved the explanatory power of the regression model, yet a little less than for the Hofstede model (∆R2_{is 0,020, 5% significance level). Furthermore, the}

Durbin-Watson score was again within the allowed range (Keller, 2009).

Table 8a. Summary GLOBE practices model

Model R R Square Adjusted R Square Std. Error of the Estimate Change Statistics Durbin-Watson R Square Change F Change df 1 df2 Sig. F Change 1 ,843 ,711 ,690 3,470 ,711 32,721 11 146 ,000 2 ,855 ,732 ,703 3,392 ,020 2,689 4 142 ,034 1,480

Model 2: Controls + Cultural dimensions GLOBE practices (AS, GE, PD, UA)

In Table 8b the coefficients of the GLOBE practices model are displayed. Again, year dummies and the lagged variable are excluded from this overview. Also, again the VIF values and tolerance levels were within allowed range. As predicted, the GE_prac dimension is significantly (1% significance level) related to %FEMALE. The sign of the GE_prac coefficient is positive. This is in line with the expectations formulated in hypothesis 3b. The UA_prac dimension is also significantly (10% significance level) related to %FEMALE. However, its sign is in the opposite direction than was predicted by theory, but similar to the sign shown in the correlation matrix (Table 5). Furthermore, GGGI was again positively related to board diversity at a 5% significance level.

Table 8b. Coefficients GLOBE practices model

Unstand. Coefficients Stand.Coefficients

B Std. Error Beta

(27)

The last regression analysis was performed for the GLOBE values model. Table 9a shows the overall results. Unfortunately, adding the cultural dimensions did not add significantly to the explanatory power of the model. To provide a complete overview, Table 9b displays the coefficients of the GLOBE values model. None of the cultural variables showed a statistically significant relationship with board diversity. The GGGI variable was again positively, at a 1% significance level, related to gender diversity in the boardroom.

Table 9a. Summary GLOBE values model

Model R R Square Adjusted R Square Std. Error of the Estimate Change Statistics Durbin-Watson R Square Change F Change df1 df2 Sig. F Change 1 ,843 ,711 ,690 3,470 ,711 32,721 11 146 ,000 2 ,847 ,717 ,687 3,483 ,006 ,720 4 142 ,580 1,533

Model 2: Controls + Cultural dimensions GLOBE values (AS, GE, PD, UA)

Tabel 9b. Coefficients GLOBE values model

Unstand.Coefficients Stand.Coefficients

(28)

6 Discussion

Table 10 summarizes the results of this research, with regards to the defined hypotheses.

There is partial support for hypothesis 2 (UA, Hofstede). The regression analysis with the Hofstede model confirms hypothesis 3 (MAS). The regression analysis with the GLOBE practices model confirms hypothesis 3b (GE). No support was found for hypotheses 1 and 3a. Furthermore, the entire GLOBE values model provided no significant results.

Table 10. Overview results

Hypothesis GLOBE values GLOBE practices Hofstede

1 Power distance Reject Reject Reject*

2 Uncertainty avoidance Reject Reject* Confirm

3 Masculinity - - Confirm

3a Assertiveness Reject Reject -

3b Gender egalitarianism Reject Confirm -

The selected dimensions from both cultural models were similar to the ones used in the Parboteeah et al. (2008) research. Very broadly, the findings of this thesis are in line with the Parboteeah et al. (2008) findings. The authors found a significant relationship between the normative context, or national culture, and managers’ gender role attitudes. In line with this, this thesis finds a significant relationship between national culture and board diversity. However, the specific cultural dimensions that showed significant results are somewhat different in this research. For example, gender egalitarianism was also significant in the Parboteeah et al. (2008) research, but the masculinity dimension was not. A possible explanation for these differences is that Parboteeah et al. (2008) did not limit their research to the boardroom population and thus formulated the dependent variable differently. Their dependent variable of interest was managers’ traditional gender role attitudes. This attitude was measured through four items from the World Values Survey. Another difference is that Parboteeah et al. (2008) performed a multilevel analysis, also including individual level factors. Finally, their sample focused on different countries. As cultures differ per country, including other countries might lead to different results.

(29)

Furthermore, this research confirms the findings from Terjesen & Singh (2008). The GGGI was clearly positively related to boardroom diversity. In all regression analyses the GGGI variable had a significant positive relationship with board diversity. The index contains dimensions that are similar to the dimensions investigated by Terjesen & Singh (2008). By including cultural dimensions, this thesis expands their findings.

There were some unexpected findings. Two dimensions (marked with an * in Table 10) turned out to have a significant positive relationship with the dependent variable, while a negative relationship was predicted by theory. The two dimensions are PD from the Hofstede model and UA from the GLOBE practices model (UA_prac). First the unexpected PD findings are discussed in more detail.

The unexpected sign of this beta coefficient might be caused by multicollinearity (Kennedy, 2002). However, the VIF values and tolerance levels are within the allowed range (see Table 7b). Still, there may be unobserved relationships between PD and other predictor variables that mediate the relationship between PD and %FEMALE. Looking at the relevant correlation matrix in Table 4, it shows a significant correlation of PD with UA and GGGI. Interesting is that when UA and GGGI are excluded from the regression analysis, which decreases the overall R2_{statistic of the model, the sign indeed}

switches. However, the coefficient becomes not significant. The corresponding regression results can be found in Appendix 3. It indicates that there indeed might be unobserved relationships between PD and the other predictor variables. Another indication for this is that in the univariate analysis (the correlation matrix, Table 4) the sign of the correlation coefficient was negative, as predicted. Only in the multivariate analysis the sign was not as predicted. Finally, it is worth noting that when using a smaller sample, the positive relationship of PD becomes not significant.

(30)

No other literature could be found documenting a positive relationship between power distance and (board) diversity. The dominating view in existing literature is that high power distance negatively influences the advancement of women into positions of authority (f.i. Caligiury & Tung, 1999; Parboteeah et al., 2008). If the relationship between PD and %FEMALE really works different than expected, this could have implications for practitioners designing programs to stimulate diversity. Creating an organizational culture with a certain degree of power distance might actually stimulate board diversity instead of prohibiting it. Research is thus needed to explore the relationship between power distance and board diversity in more detail.

Looking at the UA_prac findings, already in the univariate analysis (the correlation matrix, Table 5) the sign of the coefficient was ‘wrong’. Also, excluding other predictor variables that are to some extent correlated to UA_prac in the regression analysis (PD_prac and GGGI) does not change the findings. It might be that the relationship between UA_prac and %FEMALE really is different than expected. In high UA countries, people prefer orderliness, consistency, and structure to deal with uncertainty and ambiguity (House et al., 2004). These countries tend to have strict laws and rules. While uncertainty avoidance might prohibit the appointment of female board members by perceiving them a ‘risky hire’ (Bilimoria and Piderit, 1994), it might also stimulate boardroom diversity when existing laws and rules are aimed at improving boardroom diversity. Awareness of the importance of boardroom diversity is growing. At the European level, equality in business leadership is high on the agenda of the European Commission (European Commission, 2011). More and more countries are implementing measures, ranging from ‘soft’ voluntary measures to ‘hard’ gender parity quotas, to improve boardroom diversity (European Commission, 2011).

(31)

incomes and higher standards of living. Therefore, these countries would score high on UA_val, since the value aspect measures how things ‘should be’. In richer countries (generally high on UA_prac), lower-order needs are met. These countries may aspire less rules, because they want more opportunities for self-actualization. For example, more opportunities for experimentation and innovation (House et al., 2004). This line of reasoning might provide an explanation for the different UA findings in the two cultural models.

Other findings from both models are more in line with each other. For both models, hypothesis 1 had to be rejected. When using the ‘small’ Hofstede sample, for both models PD was positively, yet not significantly, related to %FEMALE. Also, the GE and MAS findings are in line with each other. Both dimensions look at how gender role differences are present in a society. As expected, GE was positively related to %FEMALE, whereas MAS was negatively related to %FEMALE.

What becomes clear from the results is that while the two cultural models might seem similar on some dimensions, they are not. A different operationalization of national culture can lead to different results. Both models showed that, overall, national culture influences board diversity. However, sometimes the models provide different explanations for how this relationship works in detail. The relationship between the two models seems to be more complex than suggested by the similar definition of some dimensions. Unfortunately, there has been little analysis or discussion of the consistency of results produced by these two models (Venaik & Brewer, 2010). On one hand this may be due to the relative ‘newness’ of the GLOBE model. On the other hand, due to the somewhat similar formulations of dimensions, fundamental underlying differences may be overlooked (Venaik & Brewer, 2010).

(32)

Baskerville, 2003). Dimensions are treated as if they are equally important across countries. However, the centrality or intensity of certain cultural norms and values may differ between countries (Harrison & McKinnon, 1999; Tung & Verbeke, 2010). Furthermore, a high score on a certain dimension might have a different effect in different countries (Harrison & McKinnon, 1999). Also, an exclusive focus on cultural values is insufficient to capture the complexity of national culture (Harrison & McKinnon, 1999). Additionally, the values (or dimensions) are described at a very generalized level, thereby failing to provide a detailed picture of the internal workings of the cultural environment (Harrison & McKinnon, 1999). This, combined with problems in transferring country-level findings to the individual-level, leads to a lack of specific recommendations that managers need (McSweeney, 2009; Leung et al., 3005). National averages cannot be readily transferred to the individual level (McSweeney, 2009).

Taking it a step further, the notion of one national culture has been questioned (McSweeney, 2009; Baskerville, 2003). Researchers wrongly assume cultural homogeneity within a country. There are intra-national cultural differences, which these models do not capture (Tung & Verbeke, 2010; McSweeney, 2009). In fact, these intra-national differences may be even more substantial than cross-national differences (Tung, 2008). Furthermore, by assuming that culture is constant, the models ignore that cultures change and evolve over time (Tung, 2008).

(33)

7 Conclusions, limitations & future research

In the introduction, an example was given of Norway and Italy. Norway has a high level of board diversity, Italy a very low level. Furthermore, Norway has a feminine culture, while Italy has a masculine culture. The question was, whether these cultural differences could influence the differences in levels of board diversity. The results of this research indeed show a significant negative relationship between masculinity and board diversity. Summarizing the results, this thesis makes clear that national culture influences gender diversity in the boardroom. The results show that national culture is one of the wider external structures that impact the enactment of women’s careers. National culture, as part of the macro-environment, can complement the person-centered and organization-centered explanations of, the lack of, board diversity. During the analysis several dimensions showed a statistically significant relationship with board diversity. Depending on the specific cultural dimension, this relation was either positive or negative. Most of these relationships worked in the expected direction. However, not all relations worked in the way that was expected. These findings highlight the complexity of cross-cultural research. Furthermore, the results showed that the nature of the relationship might depend on the cultural model that is used to measure it. In contrast to most cultural research, this thesis used two cultural models, the GLOBE and the Hofstede model. It became clear that the GLOBE model and the Hofstede model do not always yield similar results. For example, the UA findings showed that while the definition of a dimension might seem similar in both models, it is in fact not. The relationship between the two models is thus more complex than initially anticipated. Furthermore, the findings contribute to a broader discussion about the use of either one of these cultural models in cross-cultural research.

This thesis contributes to the board diversity literature by exploring the role of national culture. To date national culture has received little interest within board gender diversity research (Terjesen et al., 2009). By exploring its role, this thesis answers the call for more research into the environmental context that influences board diversity (Terjesen & Singh, 2008; Ruigrock et al, 2007; Mateos de Cabo et al., 2012). When exploring antecedents of board diversity, national culture is one of the macro-level forces researchers should take into account. The findings are also relevant for practitioners. An understanding of the influence of national culture on board diversity can help practitioners to design programs and policies which stimulate diversity. However, it must be noted that care should be taken when translating the findings to the individual level, since this research explicitly addressed the national level.

(34)

analysis models might have yielded different results. Perhaps when using different variables and / or different models, the results with regards to the PD dimension would be different. Finally, it is acknowledged that culture is a complex concept and using models like the Hofstede and GLOBE model is only one way to operationalize culture. In the discussion several limitations of these models were mentioned. Analysis with a different operationalization of culture might lead to different results. Due to these restrictions, it is not possible to generalize the results.

(35)

References

Baskerville, R. F. 2003. Hofstede never studied culture. Accounting, Organizations and

Society, (28):1-14.

Bilimoria, D. & Piderit, S. 1994. Board committee membership: Effects of sex-based bias.

Academy of Management Journal, 37(6): 1453-1477.

Blumberg, B. 2011. Business Research Methods (3rd ed.). Berkshire: McGraw-Hill Education.

Bullough, A., Kroeck, K. G., Newburry, W., Kundu, S. K. & Lowe, K. B. 2012. Women’s political leadership participation around the world: An institutional analysis. The Leadership Quarterly, 23: 398-411.

Caligiuri, P.M. & Tung, R.L. 1999. Comparing the success of male and female expatriates from a

US-based multinational company. International Journal of Human Resource Management,

10(5): 763-782.

Carpenter, M. A., Geletkanycz, M. A. & Sanders, G. 2004. Upper Echelons Research Revisited:

Antecedents, Elements, and Consequences of Top Management. Journal of Management,

30(6): 749-788.

Carter, D. A., D’Souza, F., Simkins, B. J. & Simpson, W. J. 2010. The Gender and Ethnic

Diversity of US Boards and Board Committees and Firm Financial performance. Corporate

Governance: An International Review, 18(5): 396-414.

Charles, M. 2005. National skill regimes, post industrialism, and sex segregation. Social Politics,

12(2): 289-316.

Cohen, J. 1988. Statistical power analysis for the behavioral science (2nd ed.). Hillsdale, NJ:

Lawrence Erlbaum Associates.

Cook, R. D. & Weisberg, S. 1982. Residuals and influence in regression. New York, NY: Chapman & Hall.

Eagly, A. H. & Karau, S. J. 2002. Role congruity theory of prejudice toward female leaders.

Psychological Review, 109(3): 573-598.

European Commission. 2011. Report on Progress on Equality between Women and Men in

2010: the gender balance in business leadership. Luxembourg: Publications Office of the European Union.

Grosvold, J. & Brammer, S. 2011. National Institutional Systems as Antecedents of Female

Board Representation: An Empirical Study. Corporate Governance: An International Review,

19(2): 116-135.

Grosvold, J., Brammer, S. & Rayton, B. 2007. Board diversity in the United Kingdom and

Norway: an exploratory analysis. Journal of Business Ethics, 16(4): 344-357.

Harrison, G. L. & McKinnon, J. L. 1999. Cross-cultural research in management control systems

design: a review of the current state. Accounting, Organizations and Society, 24: 483-506.

Heilman, M. E., Wallen, A. S., Fuchs, D. & Tamkins, M. M. 2004. Penalties for success: reactions

to women who succeed at male gender-typed tasks. Journal of Applied Psychology, 89(3):

(36)

Hillman, A. J., Cannella, A. A. & Harris, I. C. 2002. Women and racial minorities in the

boardroom: How do directors differ? Journal of management, 28(6): 747-763.

Hillman, A. J., Shropshire, C. & Cannella, A. A. 2007. Organizational predictors of women on

corporate boards. Academy of Management Journal, 50(4): 941-952.

Hofstede, G. 2001. Culture’s Consequences: Comparing Values, Behaviors, Institutions,

and Organizations Across Nations (2nd ed.). Thousand Oaks CA: Sage Publications.

Hofstede, G. 2006. What did GLOBE really measure? Researchers’ minds vs. respondents’

minds. Journal of International Business Studies, 37(6): 882-896.

House, R. J., Hanges, P.J., Javidan, M., Dorfman, P.W. & Gupta, V. 2004. Culture, Leadership,

and Organizations: the GLOBE Study of 62 Societies. Thousand Oaks, CA: Sage Publications.

Keller, G. 2009. Managerial Statistics (9th ed.). South-Western Cengage Learning.

Kennedy, P. 2002. Oh No! I Got The Wrong Sign! What Should I Do? Burnaby: Simon

Fraser University.

Konrad, A. M., Ritchie, j. E., Lieb, P. & Corrigall, E. 2000. Sex differences and similarities in job

attribute preferences: A meta-analysis. Psychological Bulletin, 126: 593-641.

Lee, C., Pillutla, M. & Law, K. S. 2000. Power distance, gender and organizational justice.

Journal of Management, 26(4): 685-704.

Leung, K. Bhagat, R. Buchan, N. Erez, M. & Gibson, C. 2005. Culture and international

business: Recent advances and their implications for future research. Journal of International

Business Studies, 36(4): 357-378.

Li, J. & Harrison, J.R. 2008. National Culture and the Composition and Leadership Structure of

Board of Directors. Corporate Governance: An International Review, 16(5): 375-385.

Mateos de Cabo, R., Gimeno, R. & Nieto, M. 2012. Gender Diversity on European Banks’ Board

of Directors. Journal of Business Ethics, 109: 145-162.

McSweeney, B. Dynamic Diversity: Variety and Variation Within Countries. Organization

Studies, 30: 933-957.

Menard, S. W. 1995. Regression analysis; Logistic distribution. Thousand Oaks CA: Sage

Publications.

Myers, R. H. 1990. Classical and Modern Regression with Applications. PWS-Kent

Publishing Company.

Oakley, J.G. 2000. Gender-based Barriers to Senior Management: Understanding the Scarcity of