The Glass Cliff at Macro-level: Board Gender Diversity, Risk-taking and National Culture.

The Glass Cliff at Macro-level: Board Gender Diversity, Risk-taking

and National Culture.

ABSTRACT

This thesis investigates the effect of a crisis period on board gender diversity on number firms in 38 different countries over the period 2004-2013. Furthermore, the influence of women on risk-taking and the moderating effect of culture is investigated. Results indicate a

positive and significant relationship between a crisis period and board gender diversity. Women have a positive and significant effect on risk-taking, and masculine cultures have a

significant effect on board gender diversity.

Name: Julie van Neerbos Student Number: 2760061

TABLE OF CONTENTS

1. INTRODUCTION……… 4

2. LITERATURE REVIEW………. 8

2.1 The ‘Glass Cliff’ phenomenon……… 8

2.2 The effect of cultural masculinity on the ‘Glass Cliff’ phenomenon……. 11

2.3 Board gender diversity and risk-taking ………. 13

3. DATA ……….………. 17 3.1 Data collection……….17 3.2 Sample selection………..17 3.2.1 Dependent variable………. .18 3.2.2 Independent variable……….19 3.2.3 Moderator………...19 3.2.4 Control variables……….. 20 4. METHODOLOGY……….24 5. RESULTS……….. 27 5.1 Descriptive statistics……….27 5.2 Country summary statistics……… 5.3 Correlation matrix……….. 5.4 Empirical results………. 5.5 Additional analyses……….

6. ROBUSTNESS CHECKS……… 6.1 Robustness test excluding United States………. 6.2 Robustness test with alternative proxis for risk-taking……… 7. LIMITATIONS AND FUTURE RESEARCH……….

1. Introduction

The board of directors is a major factor influencing a firms’ performance. The board of directors generates strategic decisions, establishes important links with external stakeholders and stimulates diversity within the organization (Ali et al., 2014). The composition of the board in terms of diversity has an impact on the way the board performs, which indirectly determines the performance of the company (Carter et al., 2003). Board diversity hereby, refers to the internal variety in the composition of the board and can have multiple dimensions, such as nationality, age and ethnicity (Campbell and Minguez-Vera, 2007). This research focuses on board gender diversity, which is not only the most debated diversity issue in research (Campbell and Minguez-Vera, 2007), but is due to new legislative measures promoting female representation on corporate boards, also being heavily debated in today’s world. Despite the increased attention on board gender diversity, much still needs be done in order to be able to understand how and when board gender diversity influences the board and firm (Triana et al., 2014).

Palvia et al. (2015) state that gender-based differences between men and women manifest in decision-related choices that top executives and directors make, hereby influencing major strategic and financial decisions in corporations. However, gender diversity on boards is still limited as there remain several barriers towards women who are trying to obtain a higher position in a corporation. According to Wellington et al. (2003), barriers to women’s advancement in corporations include for example; the exclusion from certain networks, stereotyping, lack of mentoring, shortage of role models, commitment to personal responsibilities, lack of accountability on senior leadership, and limited opportunities for promotion.

appointed to board positions when a company experiences a state of crisis (Bruckmuller et al., 2014). Gender diversity leads to an increase in knowledge, creativity and innovation on the board, which has the potential to become a competitive advantage and as a result may improve board effectiveness (Watson et al., 1993). Therefore, during times of crisis, a gender diverse board of directors is able to create a more extensive process of solution generation for corporations. On the other hand, the downside of diverse boards seems to lead to a decrease in firm performance through for example an increased risk of conflict among directors (Erhardt et al., 2003). To diminish this negative effect on firm performance, boards could counteract this by reducing the percentage of women on board, during hectic periods of crisis. An important question here is whether there is empirical evidence to justify the impact of prosperity versus periods of crisis on board gender diversity, which may indirectly be related to corporate performance. Reviews of literature on this field have provided mixed results between the effect of either declining or improving firm performance on board gender diversity (Shehata et al., 2017), resulting in a remaining theoretical and empirical question. Next to that, Triana et al. (2014) acknowledge that an increase in the number of women and their power on boards gives rise to the importance of examining how diversity affects firm outcomes during times of threat. This is related to the famous “Lehman Sisters” hypothesis, stating that the crisis could have developed the way it did if ‘Lehman Brothers’ had been ‘Lehman Sisters’ (Adams, 2016). As such, when looking at gender diversity on corporate boards, it is of great importance to take both the stages of economic prosperity and economic crisis into account.

directors during an economic crisis compared to an economic prosperity stage. However, their research is restricted to Chinese firms and can therefore not be generalized to other countries. Elaborating on this research, we create a macro-level research on the glass cliff phenomenon including data on boards of directors from 38 different countries.

Next to that, Hyland and Marcellino (2002) acknowledge that the interest in empirical research on the composition of corporate boards has mainly been restricted to US data. As such, these US focused articles promote a US corporate and academic view, which is questionable and cannot be generalized as a general and global paradigm. Moreover, through globalization, different cultures have influenced each other over the last few decades (Kostis, 2018), making it even more difficult to generalize the US culture to other countries. Therefore, these US focused studies may suffer from cultural blind spots when looking at other countries. With the ever-growing phenomenon of globalization, it is becoming even more important to investigate the impact of national culture on corporate boards. Our research responds to the increasing globalization by including a cultural analysis to our board diversity research.

post-

Our research will add to existing researches on board gender diversity by investigating the following research question:

2. Literature Review

There is general consensus that directors have the capability to influence firm performance (Hillman & Dalziel, 2003). However, in which way diversity has an effect on this is still greatly discussed in the field of corporate governance. Next to that, the appointment of directors by boards is influenced by differences in stages of economic performance such as downturns and upturns. As such, the composition of boards of directors in times of economic prosperity may create contradicting results compared to times of crisis.

Bruckmüller and Branscombe (2010) refer to the ‘glass cliff’ as the phenomenon of women being more likely to gain leadership positions in times of crisis, and men being more likely to gain these top positions in times of prosperity. However, unlike Ryan and Haslam (2007), they argue that the glass cliff phenomenon arises due to a ‘lack of fit’, in which women may be given leadership positions in times of crisis because their lack of competence for leading successful corporations is seen to qualify them only for second-rate opportunities.

This ‘lack of fit’ can be explained by theory on the managerial stereotype, where Ryan et al. (2007) found evidence on managers of unsuccessful companies being associated with ‘think crisis - think female’, creating a vision that women are seen as having certain capabilities that are suited for the management of crisis situations, and ‘think manager – think male’ associated with men leading companies in times of prosperity. They give the following examples; Tharenou (2001) mentions that these diverse skills that women bring to the corporation may create a competitive advantage, Kent and Moss (1994) position women as leaders in environments which require a great amount of social interaction crucial for today’s increasing global market, Rose (2007) and Hurst (1989) mention women’s characteristics related to confidence, sharing of information, cohersion and quick responsiveness to challenges, and Ruderman et al. (2002) acknowledge the skills that women acquire through their multiple roles, such as marital and parental roles, creating greater decision making processes and increased organizational performance. These increased skills that women bring with them into an organization are crucial for firm growth and the reducing of firm failure (Goedhuys & Sleuwaegen, 2016). Therefore, the skills and traits possessed by women may be seen as particularly useful in times of crisis and economic instability.

motivation to engage more in collective action, resulting in more acceptance towards out-group members, women.

Next to that, Shepherd et al. (2013) mention that high status groups, such as directors, working in stable environments already possess secure positions, creating a decrease in discrimination against out-group members. In this way, firms maintaining a stable economic performance during economic crisis, may be open to more female directors on boards.

We therefore assume that during crisis, men prefer to safeguard their positions of success by appointing women to these leadership positions. As such, firms in general, regardless of positive or negative economic performance, increase women to their boards of directors. We state the following hypothesis;

H1: More women are appointed on boards during a crisis period.

2.2 The effect of cultural masculinity on the ‘Glass Cliff’ phenomenon

A growing amount of literature has shown the importance of differences in culture on the economic decision-making in corporations (Li et al., 2011). As Kostis et al. (2018) state, culture and economics are two of the most powerful forces that shape human behavior, and thereby economic activity.

managerial roles will differ depending on the cultural norms and industry profile of the country, such as ‘masculine’ or feminine’ cultures.Masculine cultures are hereby described as valuing aggressive policies, competitive behavior and managerial decisiveness (Kanagaretnam et al., 2011). Also, societies dominated by a masculine culture stress material success and decisiveness and assign different roles to males and females, where males are expected to carry out the ambitious and competitive roles in society and “men should dominate in all settings” (Hofstede, 1984).

Arena et al. (2015) predict a negative impact of the appointment of women directors on firm performance in masculine industries. As such, they pose that masculine cultures are less likely to appoint women to boards of directors in fear for negative economic consequences. Newman and Nollen (1996) state that in masculine cultures managerial compensation for good performance is relatively greater and the penalty for poor performance is relatively lower. As such, the relative low disadvantages for managers in masculine cultures, related to poor performance during times of crisis, may indicate that these masculine cultures are less likely to make drastic changes in the composition of boards, such as the appointment of more women on boards.

of directors. So, during times of crisis, when economic instability and threat is high, boards will prefer in-group members over out-group members.

As a result, we want to investigate what the cultural impact is on the percentage of women on boards, and whether the glass cliff on macro-level is strengthened or weakened when depending on a countries cultural setting. We hypothesize as follows:

H2: The increase or decrease in percentage of women on boards during a crisis can be explained by Hofstede’s dimension on masculinity.

2.3 Board gender diversity and risk-taking

Corporate risk-taking is fundamental to firm performance and survival (Li et al., 2013), and as such, understanding the risk-taking environment will increase a corporation’s future performance and help maintain its competitive position in the industry. Risk is defined as “the uncertainty that exists to what the eventual outcome will be, arising in any decision where there is some doubt about at least one of the possible outcomes” (Roszkowski, Davey, 2010). Risk tolerance is hereby the amount of risk chosen when making financial decisions.

Results by Soininen et al. (2012) indicate that the more risk-taking a firm is, the more the financial performance is affected by the crisis. As such, the effects of economic downturn during crisis are more severe on risk-taking firms. An explanation for this result may be that risk-takings firms operate in uncertain and precarious environments.

Next to that, Gerrans et al. (2013) state that the losses suffered by corporations during the crisis increased individuals risk aversion, as people become more averse to the possibility of future losses when they have experiences loss in the past. On top of this, their research indicates that risk tolerance in post-crisis period seems to be significantly lower than the pre-crisis risk tolerance, suggesting that after the pre-crisis the level of risk-taking will reduce. An important finding on this aspect is the fact that women are more risk sensitive in the context of losses than gains compared to men (He et al., 2007), because men and women differ in their estimation on probabilities of losses and gains (Olsen and Cox, 2001). Furthermore, Maxfield et al. (2010) mention that experiencing striking events may have a permanent impact on women investors’ perceptions and risk-taking behavior, where experiencing a number of losses reduces their willingness to take risk. In this way, women have shown to be more vulnerable to problematic and disastrous situations (Shane, 1993). As such, when corporations appoint more women to boards of directors, the amount of risk-taking will decrease.

Masculinity

In today’s globalized business environment, national culture significantly influences corporate risk-taking. Results by Li et al. (2013) demonstrate the important role of cultural influences on corporate risk-taking, showing that culture influences corporate risk-taking directly on risky corporate decision-making. Next to that, Maxfield et al. (2010) mention that gender gaps in risk propensity may be sensitive to national culture, again highlighting the importance of culture on risk behavior.

taking to achieve a position of power. Organizations must oversee the creation of these assumptions related to the stereotyping of men as risk takers, as these assumptions could cause classification of the wrong person to make choices related to risk-taking to achieve organizational goals. On the other hand, ‘glass cliff’ behaviors could result in an ‘over-gendered’ risk-related stereotype, where women are expected to have the capabilities to cope with organizations during periods of crisis.

Firm performance

Atkinson et al. (2003) mention that women achieve comparable performance as men, regardless of differences in their risk-taking strategies. Thus, even with a more cautious approach to risk-taking, performance is still likely to increase in a more steady and consistent manner. Perryman et al. (2016) include that an increase in the percentage of females on boards may indirectly result in less risk-taking and therefore lower returns, while simultaneously having fewer huge losses due to an increased stable performance. Their results show that an increase in board gender diversity reduces firm risk and improves firm performance. Olsen and Cox (2001) add to this by stating that women tend to perform better than men due to their lower risk-taking behavior and the taking of time to respond to an uncertain situation, compared to men. Next to that, after a period of crisis, a board more open to female directors can recover more quickly from the crisis than boards without such openness (Sun et al., 2015). Due to the increased knowledge, diversified experiences and contrasting backgrounds, boards with such openness to female directors are more likely to seize opportunities and are more prepared to interpret the crisis. Miller et al. (1998) hereby add that learning from a more diverse set of experiences creates opportunities related to greater skill capacities (Ferrero et al., 2012), which in turn generates greater decision processes related to the managing of immediate opportunities and threats, such as periods of crisis. As such, women may create favorable corporate circumstances after periods of downturn.

3. Data

3.1 Data

To address our research question, we gather data from the Wharton Research Data Services, with Compustat Global and Boardex as our main data sources. We excluded financial firms (SIC codes between 6000 and 6990) from our dataset, because this industry is highly regulated and reasonably different from other industries (Masulis and Mobbs, 2014).

Our dataset structure is multilevel. At the country level, we have firms from 38 different countries. At the firm level, we have 505 different firms. It is important to be able to distinguish the effects that take place at the country level from those that take place at the firm level, both to understand the role of country- versus firm-level determinants. As such, we are able to properly model their interactions.

In the end, our dataset consists of 13363 observations in total, with which we eventually perform cross-country firm-level analyses for the period of 2004 to 2008, the pre-crisis period, and 2009 to 2013, the post-crisis period. Next to that, all continuous variables are winsorized to the 1th and 99th percentile in order to alleviate the impact of outliers and to reduce the skewness of variables (Gujarati and Porter, 2009). Dummy variables included in the sample are not winsorized to the 1th and 99th percentile, as it is not possible for these values to skew the distribution.

3.2 Sample selection

As has been shown in previous research, board gender diversity may not be a random exogenous variable (Adams and Ferreira, 2009). In order to reduce the issue of reversed causality, natural logarithm is used whereby all independent variables, excluding masculinity variable, are lagged by t-1. Lagged variables are preferred in our analysis, as Liu et al. (2014) mention that female directors, and board characteristics in general, require considerable time before generating an impact on corporations (Liu et al., 2014). Data on Hofstede’s dimensions are collected over the period 1967 to 1973, and as such, do not vary over time in our time period ranging from 2004 to 2013. Therefore, our variable on masculinity is not lagged.

3.2.1 Dependent variable

In our analysis on hypothesis 1 and 2, we construct two different measures for the dependent variable board gender diversity. We firstly follow Sun et al. (2015) and create a dummy variable which equals 1 if a board has w directors and 0 otherwise (BrdGender). Secondly, we create a more advanced model by looking at the percentage of women on board of directors (Perryman et al., 2016; Adams and Ferreira, 2009) as a measure for board gender diversity (PWomen). This variable was calculated following Mulcahy and Linehan (2013) by dividing the number of women on board by the total number of directors on the board. By comparing the simplified model on board gender diversity with the more advanced model, we are able to investigate the differences between having a woman on the board of directors in general and the effect of an increase in the number of women directors on boards.

an economic shock caused by a crisis. We follow Bruno and Shin (2014) by computing the country-adjusted volatility of firm-level earnings, calculated as the standard deviation of the country-adjusted ROA for each firm. We calculate earnings volatility by calculating the standard deviation of the difference between a firm’s EBITDA/assets and the country average for each year (Risk).

Next to that, in order to investigate the relationship between the amount of risk-taking in post-crisis period compared to pre-post-crisis and post-crisis period for hypothesis 3, we again create a dummy variable (PostCrisis) which equals 1 for the post-crisis period from 2009-2013 and 0 otherwise.

3.2.2 Independent variable

To be able to investigate the relationship between our different models on board gender diversity in crisis period for hypothesis 1 and 2, we create a dummy variable (Crisis) which equals 1 for the crisis period and 0 otherwise. We hereby identify the crisis period ranging from 2007 to 2009.

The independent variable for hypothesis 3 is board gender diversity, in which we use our more advanced model on board gender diversity by taking the percentage of women on boards (PWomen). For all independent variables, excluding masculinity, we use lagged variables to mitigate the possible effect of endogeneity.

3.2.3 Moderator

0 to 100. These country scores indicate the relative position of a country compared to other countries.

3.2.4 Control variables

To account for potential alternative drivers of board gender diversity and risk-taking, we include control variables that might be theoretically correlated with our predictors. In order to address this correlated omitted variable bias problem, both firm- and country-specific control variables were included in our research. Firm performance, firm size, board size, cash flow and firm value were used as potential firm-specific control variables, and gender quota, real gross domestic product, and the log of real gross domestic product as country-specific control variables.

Firm-level control variables

Firm performance

(ROA). Compared to other performance ratio’s, this ratio is less subject to managers’ judgement on policy choices (Jiraporn et al., 2011), and therefore provides a more reliable measure on a company’s actual performance.

Firm size

We expect a positive coefficient for firm size, since large firms may have the ability to perform better due to economies of scale compared to small organizations (Carter et al., 2003) and may offer a mixture of diversity training which may result in the appointment of a more diverse board (Hyland et al., 2002). As such, firm size has a positive influence on board diversity. When looking at risk-taking, Fernandes et al. (2012) argue that larger firms may be prone to take on less risk. Moreover, Hillman et al. (2007) mention that larger firms are more visible to the public and as such experience an increasing pressure from a variety of stakeholders to increase the amount of female representation on boards. As such, firm size (Size) is included as a control variable, which may influence board gender diversity and risk taking. Following Conyon and He (2017) and Sun et al. (2015), firm size is measured by the log of total assets.

Board size

taking board size into account as a control variable. Next to that, Golden and Zajac (2001) mention a curvilinear relationship between board size and degree of risk-taking. This implies that as board size increases, risk-taking increases as well, but only up to a certain point, after which, as board size increases the degree of risk-taking decreases. Nakano and Nguyen (2012) mention a negative relationship between board size and risk-taking, due to the difficulty of reaching an agreement in bigger boards. Board size (BrdSize) is measured by taking the log of the total number of directors in a corporation (Carter et al., 2003).

Firm value (Property, plant, equipment)

Shin and Stulz (2000) state that an asymmetric relation exists between firm value and risk, indicating that an increase in firm value has a smaller impact on risk than a decrease in firm value. Next to that, Carter et al. (2003) found a positive association between firm value and the percentage of women on boards of directors. As such, we include firm value (FirmValue) as a control variable, following Gompers et al. (2010) by taking net property, plant and equipment and dividing it by lagged assets.

Country-level control variables

Dummy Quota

Corporate Governance Institute (ECGI). A dummy variable was created (QUOTA), equaling 1 if a gender code/quota is implemented in the country and 0 if there is no code.

Gross Domestic Product

4. Methodology

In order to test hypothesis 1, we focus on the coefficient 𝛽_!, which measure the sensitivity of crisis period on the board gender diversity. A positive and significant coefficient indicates that during a crisis period, gender diversity on boards increases. We include time, country and industry fixed effects in our models. Fixed effects models diminish the effects of possible reverse causality as they omit time-invariant firm characteristics (Adams and Ferreira, 2009). As our sample covers the period from 2004 to 2013 including the crisis period which affects all firms in our research, it seems reasonable to control for time fixed effects. Our sample consists of observations covering 38 different countries, so country-fixed effects are included, and industry-fixed effects are included as our sample covers multiple industries. Our first model on hypothesis 1, with our dummy variable as board diversity measure, is estimated with the following regression equation:

𝐵𝑟𝑑𝐺𝑒𝑛𝑑𝑒𝑟 !,! (!"#$#$) = 𝛽! + 𝛽!𝐶𝑟𝑖𝑠𝑖𝑠!,!!!+ 𝛽!𝑅𝑂𝐴 !,!!!+ 𝛽!𝐹𝑖𝑟𝑚𝑆𝑖𝑧𝑒!,!!!+

𝛽_! 𝐵𝑜𝑎𝑟𝑑𝑆𝑖𝑧𝑒_!,!!!+ 𝛽_!𝐹𝑖𝑟𝑚𝑉𝑎𝑙𝑢𝑒_!,!!!+ 𝛽_! 𝐺𝐷𝑃 _!,!!!+ 𝛽_!𝐿𝐺𝐷𝑃_!,!!!+ 𝛽_! 𝑄𝑢𝑜𝑡𝑎_!.!!!+ 𝑌𝑒𝑎𝑟 𝑓. 𝑒. +𝐼𝑛𝑑𝑢𝑠𝑡𝑟𝑦 𝑓. 𝑒. +𝐶𝑜𝑢𝑛𝑡𝑟𝑦 𝑓. 𝑒. + 𝜀_!,!,!

The second more advanced model, with the percentage of women on boards as measure for board gender diversity, is estimated with the following regression equation:

𝑃𝑊𝑜𝑚𝑒𝑛 _{!,! (!"#$#$)} = 𝛽_! + 𝛽_!𝐶𝑟𝑖𝑠𝑖𝑠_!,!!!+ 𝛽_!𝑅𝑂𝐴 _!,!!!+ 𝛽_!𝐹𝑖𝑟𝑚𝑆𝑖𝑧𝑒_!,!!!+ 𝛽! 𝐵𝑜𝑎𝑟𝑑𝑆𝑖𝑧𝑒!,!!!+ 𝛽!𝐹𝑖𝑟𝑚𝑉𝑎𝑙𝑢𝑒!,!!!+ 𝛽! 𝐺𝐷𝑃 !,!!!+ 𝛽!𝐿𝐺𝐷𝑃!,!!!+

Board Gender is a dummy variable equaling 1 if a board has female directors and otherwise; PWomen is the percentage of women on boards of directors for company i at year t (Torchia et al., 2011); Crisis is a dummy variable equaling 1 for the crisis period and 0 otherwise; ROA is measured by taking the earnings before extraordinary income and preferred dividend in financial year t and dividing it by the average of book values of total assets at the end of the financial year t (Haslam, 2010); Firm Size measured as the log of total assets (Conyon and He, 2017; Sun et al., 2015); Board Size is measured by taking the log of the total number of directors in a corporation (Carter et al., 2003); Firm Value is measured by taking net property, plant and equipment and dividing it by lagged assets (Gompers et al., 2010); GDP is measured using the Maddison Project Database and taking the real GDP per capita across countries and LGDP is measured as the log of real GDP (LGDP) per capita (Kingsley and Graham, 2017); Quota is a dummy variable equaling 1 if a gender quota is implemented in the country and 0 otherwise (ECGI).

For our second hypothesis, we examine the effect of cultural differences on the relationship of a crisis period on the percentage of women on boards of directors. From now on, we will continue to use our more advanced variable on board gender diversity (PWomen). The regression equation is as follows:

𝑃𝑊𝑜𝑚𝑒𝑛 _{!,! (!"#$#$)}

= 𝛽! + 𝛽!𝑀𝐴𝑆 !,!+ 𝛽!𝐶𝑟𝑖𝑠𝑖𝑠!,!!!+ 𝛽!𝑀𝐴𝑆 !,!∗ 𝐶𝑟𝑖𝑠𝑖𝑠 + 𝛽!𝑅𝑂𝐴!,!!!

+ 𝛽_!𝐹𝑖𝑟𝑚𝑆𝑖𝑧𝑒_!,!!! + 𝛽_! 𝐵𝑜𝑎𝑟𝑑𝑆𝑖𝑧𝑒_!,!!!+ 𝛽_!𝐹𝑖𝑟𝑚𝑉𝑎𝑙𝑢𝑒_!,!!! + 𝛽_! 𝐺𝐷𝑃 _!,!!!+ 𝛽_!𝐿𝐺𝐷𝑃_!,!!!+ 𝑌𝑒𝑎𝑟 𝑓. 𝑒. + 𝐼𝑛𝑑𝑢𝑠𝑡𝑟𝑦 𝑓. 𝑒.+ 𝜀_!,!,!

following Hofstede’s scores from 0 to 100 (Hofstede, 2001). Hereby, we want to measure whether the results in hypothesis 1 can be explained by a cultural dimension.

A significant positive interaction coefficient of Crisis*MAS is an indication that high masculine countries strengthen the results on the relationship between crisis period and the percentage of women on boards. A negative interaction coefficient would indicate that high masculine countries weaken the results on the relationship between crisis period and the percentage of women on boards.

Our third hypothesis investigates the impact of a gender diverse board on risk-taking in post-crisis period. We estimate the following regression equation:

𝑅𝑖𝑠𝑘 _{!,! (!"#$!")}

= 𝛽_! + 𝛽_!𝑃𝑊𝑜𝑚𝑒𝑛 _!,!!!+ 𝛽_!𝑃𝑜𝑠𝑡𝐶𝑟𝑖𝑠𝑖𝑠_!,!!!+ 𝛽_!𝑃𝑊𝑜𝑚𝑒𝑛 _!,!!! ∗ 𝑃𝑜𝑠𝑡𝐶𝑟𝑖𝑠𝑖𝑠 + 𝛽_! 𝑅𝑂𝐴_!,!!!+ 𝛽_!𝐹𝑖𝑟𝑚𝑆𝑖𝑧𝑒_!,!!! + 𝛽_! 𝐵𝑜𝑎𝑟𝑑𝑆𝑖𝑧𝑒_!,!!! + 𝛽!𝐹𝑖𝑟𝑚𝑉𝑎𝑙𝑢𝑒!,!!!+ 𝛽! 𝐺𝐷𝑃 !,!!!+ 𝛽!𝐿𝐺𝐷𝑃!,!!!

+ 𝑌𝑒𝑎𝑟 𝑓. 𝑒. + 𝐼𝑛𝑑𝑢𝑠𝑡𝑟𝑦 𝑓. 𝑒. + 𝐶𝑜𝑢𝑛𝑡𝑟𝑦 𝑓. 𝑒.+ 𝜀_!,!,!

Risk is measured as the average risk-taking within a given country for each year (John et al., 2008).A positive and significant interaction term PWomen*PostCrisis would imply that crisis periods strengthen the results on the relationship between the percentage of women on boards and risk-taking.

5. Results

5.1 Descriptive statistics

Table 2 provides the descriptive statistics of the full sample used in our research. In the table an overview is shown of the number of observations, the mean, standard deviation, minimum and maximum. As mentioned before, we correct for outliers in all independent and control variables by winsorizing at the 1st and 99th percentile, as outliers may result in biases and problems in our regression analyses. Our control variables are summarized, excluding the dummy variables as these do not show significant interpretable results.

5.2 Country summary statistics

Table 3 reports descriptive statistics in mean values by country. In general, it is observable that Germany, Great Britain and United States dominate in firm-year observations (n = 1380, 1224, and 2922 respectively). In terms of the percentage of women, it is worth noting that Argentina, Indonesia, New Zealand, Panama and Peru have no women on boards of directors, whereas Norway shows the greatest percentage of women (≈ 34%), followed by Finland (≈ 20%). Next to that, Israel, Peru, and Poland present the highest ratio of risk-taking (.106, .109, and .145). Norway presents the lowest score on the cultural masculinity index (8), whereas Japan scores the highest on masculinity (95). Looking at ROA, Israel and Peru show a negative sign (-.043 and -.029 respectively), indicating that the companies in these countries for our sample are inefficiently using its assets to generate earnings. China presents the Table 2. Descriptive statistics

VARIABLES N mean sd min max

pWomen 13,363 0.127 0.115 0 0.444 Risk 13,363 0.0413 0.0473 0.00144 0.247 MAS 13,363 56.57 17.70 8 95 BoardSize 13,363 1.045 0.170 0.699 1.462 FirmSize 13,363 9.422 2.565 3.402 16.30 GDP 13,363 37,778 13,641 3,936 61,905 LGDP 13,363 10.42 0.590 8.278 11.03 ROA 13,363 0.0422 0.0885 -0.427 0.278 FirmValue 13,363 0.268 0.208 0.00186 0.823

Furthermore, 8 countries in our sample experience the use of gender quota’s; Belgium,

Finland, Germany, Great Britain, Israel, Italy, Norway and Spain. As such, it is quite plausible that Finland and Norway display the highest percentage of women on boards. Lastly, it is quite remarkable to notice that United States, without gender quota legislation, has a

Table 3. Descriptive statistics mean values by country.

Country N pWomen Risk MAS ROA GDP LGDP Quota

5.3 Correlation matrix