• No results found

The relationship between diversity in board composition and risk-taking behavior

N/A
N/A
Protected

Academic year: 2021

Share "The relationship between diversity in board composition and risk-taking behavior"

Copied!
45
0
0

Bezig met laden.... (Bekijk nu de volledige tekst)

Hele tekst

(1)

The relationship between diversity in board composition and

risk-taking behavior

Name: Matthijs Bredenoord Student number: 10580646

Thesis supervisor: prof. dr. V.R. O’Connell Date: June 23th, 2018

Word count: 12,027

Msc Accountancy & Control, specialization Control

(2)

Statement of Originality

This document is written by student Matthijs Bredenoord who declares to take full responsibility for the contents of this document.

I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.

The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

(3)

Contents

Abstract ... 5

1 Introduction ... 6

2 Literature ... 9

2.1 Resource-based theory... 9

2.2 Upper echelons theory ... 9

2.3 Diversity ... 10

2.4 Risk-taking ... 12

2.4.1 Corporate risk-taking ... 12

2.5 Hypothesis / theory development ... 13

2.5.1 Masculinity versus femininity (MAS) ... 13

2.5.2 Uncertainty avoidance index (UAI)... 13

2.5.3 Individualism versus collectivism (IDV) ... 14

3 Research method... 15 3.1 Conceptual model ... 15 3.2 Regression equation ... 15 3.3 Explanation of variables ... 16 3.3.1 Independent variables ... 16 3.3.2 Dependent variables ... 17 3.3.3 Control variables ... 18 4 Data ... 21 4.1 General ... 21 4.2 Sample ... 21 4.3 Descriptive statistics ... 22

4.3.1 Core statistic values ... 22

(4)

4.3.3 Multicollinearity ... 25

5 Results... 26

5.1 Main results ... 26

5.2 Discussion of results ... 27

5.2.1 Model 1: Association with R&D intensity ... 27

5.2.2 Model 2: Association with the Altman Z-score ... 29

5.2.3 Model 3: Association with standard deviation ROA ... 30

5.2.4 Model 4: Association with volatility ... 31

5.2.5 Summary of results ... 32

5.3 Sensitivity test ... 33

6 Conclusion ... 36

6.1 Conclusions ... 36

6.2 Limitations... 37

6.3 Contributions and avenues for further research ... 38

7 References ... 39

(5)

Abstract

This study has been practiced for the benefit of the Msc Accountancy & Control, thesis of Matthijs Bredenoord. This paper examines the effect between diversity in boardroom composition and risk-taking behavior in the period of 2008 – 2017. Boardroom diversity is measured by three cultural dimensions from Hofstede: masculinity, uncertainty avoidance and individualism. Risk-taking is measured by R&D intensity, the Altman Z-score, standard deviation return on assets and stock price volatility. It was hypothesized that masculinity and individualism have a positive association with risk-taking behavior and uncertainty avoidance a negative

association. A total of 2024 firm years are studied, derived from the largest 214 non-financial firms listed on the London Stock Exchange. While not all cultural dimensions show consistent effects on the dependent variables, there is evidence that all three cultural characteristics could have influence on risk-taking behavior. Furthermore, firm size can increase or decrease the effect of diversity in board composition and risk-taking behavior.

Keywords: Board diversity, corporate risk-taking, Hofstede cultural dimensions, masculinity, uncertainty avoidance, individualism, nationality diversity, UK, LSE

(6)

1 Introduction

The purpose of this paper is to examine the effect of boardroom diversity on risk-taking behavior. More specifically, this study will investigate the relationship between diversity of the composition of the board of directors and the risk-taking behavior of the firm. Empirical research will formulate an answer to the following research question:

How does diversity in boardroom composition influences the risk-taking behavior for listed non-financial firms in the UK?

In recent years, the focus in boardroom composition is changing towards a more diverse composition of board directors. In 2014, the Dutch minister of health, well-being and sports Jet Bussemaker set a target of 30% women in boardrooms in the Netherlands. The European Parliament, with Viviane Reding leading the way, also tried to get more women in boardrooms to have a better reflection of society in boards of organizations. However, this reflection of the society does not only contain gender diversity. Multiple stakeholders are increasing their focus on diversity in society. In particular, firms are experiencing more influence caused indirectly by diversity and directly from diversity within their organizations. Firms behave differently from other firms, even in the same country, industry or even teams within firms. These differences are initiated by the persons determining the actions on behalf of the firm. Research has pointed out these differences come from various characteristics of firms. For example, influences of gender on firms have been intensively studied. Campbell and Minguez-Vera (2007) studied the effect of gender diversity on firm performance in Spain, which has historically a meager female

participation among board members. Research proved that a more diverse composition is positively associated with firm value. Recently, racial and ethnic minorities influence on firms’ performance has been a topic of research by McKinsey (McKinsey, 2015). They find that in the UK 22% of the nationalities within the UK are present in boardrooms, which makes it one of the countries with the most diverse boardroom compositions and therefore an interesting country to investigate. Whereas the US has only a 3% of the domestic nationalities present in boardrooms. In October 2017, an EY report, led by Sir John Parker, urged firms in the UK to increase the diversity of their boards (The Parker Review Committee, 2017). They state that the current composition of UK boards does not reflect the UK society, nor the international society. UK business minister Margot James added that this is a critical agenda. She requests the largest companies to take the lead in following up the recommendations Sir John Parker has made. The Parker review committee has published the goal to increase the ethnic diversity of UK boards by

(7)

proposing each FTSE 100 board to have at least one director of a minority background by 2021. At the time the report was published, ethnic minority representation in the FTSE 100 was merely around 5%. Both the Parker Review Committee report and the McKinsey report say diversity will deliver advantages for firms and stress that diversity is of great importance. This raises questions about how this diversity will influence a firms' behavior and subsequently its performance.

Characteristics of boards and their relationship to firm behavior remains a fundamental topic in prior corporate governance research. However, the research often focuses on corporate governance mechanisms and paying little attention to the resulting performance (Hermalin & Weisbach, 2001). Literature shows that demographic differences affect the firms' performance. Erhardt, Werbel and Shrader (2003) examined the relationship between demographic diversity on boards of directors with firms’ performance in 1993-1998. The research on boards in the USA showed that demographic diversity on boards and performance, specifically ROI and ROA, are positively associated. This does not inform us about the risk that coincided with the

performance. Demographic diversity says something about the origin of the boardroom

directors, but the cultural characteristics connected to these areas remain unspoken in this study. The cultural background of the board member could be linked to the behavior within the organization. Moreover, diversity in a group can lead to considerable benefit conditional on enhanced integration and communication (Maznevski, 1994). Also, Mihet states that attitudes towards risks are likely to be directed by cultural norms which may encourage or deter risk-taking (2013). She links the organizational behavior to the cultural diversity measured by Hofstede’s cultural dimensions. Hofstede measures a score on six cultural dimensions that describe how culture influences the workplace. This study investigates diversity because society’s focus shifts more and more towards this topic and the influence of diversity has been proven in prior literature.

Difference in the people means different information that directors bring to the table when governing the firm. With every firm, risk is a highly important part of their

decision-making. The financial crisis has increased the attention on firm risks which the board of directors is responsible for. Their monitoring role is exposed to more pressure from the stakeholders. Furthermore, regulations are tightened since the financial crisis in 2008 - 2011. The severe effects are a reminder to assess risks thoroughly. Consequently, the magnitude of the monitoring

(8)

Tightening the regulations resulted in more attention governments, consumers and other stakeholders on the risk-taking behavior of directors.

This paper will build on all existing research on boardroom diversity and risk-taking behavior. However, the specific relationship between cultural diversity in boardrooms and risk-taking behavior in the UK is never studied before. Previous research has merely focused on the demographic diversity of the entire board or the cultural dimensions in the country where the board is active. This paper will contribute by providing evidence of how the different cultural characteristics influence risk-taking behavior. The insights will contribute to the literature for the relationship of diversity on firm behavior. Also, the evidence contributes to knowledge of how to compose boardrooms to achieve a particular behavior and the performance that can be derived from the behavior.

I would especially like to thank prof. dr. V.R. O’Connell for his supervision during the writing of my thesis. His valuable input has been essential in developing this study.

(9)

2 Literature

In the literature section, I will address the structure of the literature review and the key papers that I will use to develop the hypotheses.

2.1 Resource-based theory

Research on corporate directors and boards often focuses on two roles: agency and resource dependence (Hillman et al., 2000). I will elaborate more on the agency role in paragraph 2.3; the resource dependence is explained in this paragraph.

Peoples’ behavior depends highly on the coincidence of their character and the

characteristics of the internal environment a person is subject to. A board of directors is such an internal environment. The behavior of directors is dependent on the characteristics of the other board members and the external environment the board is acting in. The resource-based theory sketches a situation where a firm uses its resources and capabilities to develop its strategy

(Gallego-Álvarez, 2011). This perspective refers to the ability of boards to bring resources to the company (Hillman & Dalziel, 2003). Director’s most valuable resources are the strengths and weaknesses of themselves. Hillman et al. (2000) state that diversity in the composition of the board can provide a diverse set of resources to a company and these resources are linked directly linked to performance. This is in line with the human capital theory (Hillman & Dalziel, 2013), where we distinguish human capital (experience, expertise, reputation) and relational capital (a network of ties to other firms and external contingencies). Resource dependence theory explains how board capital is formed by the characteristics of the directors. Based on this resource dependence theory we can argue that diversity in board composition influences decision-making. Since diversity comes in numerous forms, the directors’ characteristics play a role in determining a firm’s performance.

2.2 Upper echelons theory

Following Post and Byron (2015, the upper echelons theory links to the resource-based. According to their study, director's experiences, knowledge, and values form their cognitive frames, which determine how directors seek and interpret information. Strategies emerge from these cognitive frames which means directors' characteristics indirectly shape strategies. Hence, the results from Hambrick et al. (1982), who found that strategic choices are related to the experience and characteristics of a person. Intensity of investments in research and development

(10)

or diversification of strategies, in other words, activities that accompany corporate risk-taking, are in part dependent on the characteristics of the directors (Carpenter et al., 2004).

2.3 Diversity

Any group of people can be described by its diversity. The diversity becomes valuable when it contributes to the group's ability to achieve goals (Maznevski, 1994). In this paper, the definition of diversity is the representation of the different cultures of the director’s country of origin.

Previous research generally relates to two different types of diversity: observable and non-observable. Observable diversity can, for example, be age, gender or race. Non-observable diversity can be education, norms and values, religion or personal characteristics (Maznevski, 1994; Milliken and Martins, 1996; Pelled, 1996; Kilduff et al., 2000). Whereas most research has focused on observable diversity, this paper will focus on the relationship of non-observable diversity, in form of cultural diversity, and firms’ risk-taking behavior. Two observable variables will be taken into account as well; age and gender. Although most board members still are grey males older than 55 years nowadays, as stated earlier in the introduction, the diversification of boards is developing.

Age could affect risk-taking behavior because experience comes with age. Hambrick et al. (1982) claim that young managers are driven to be successful in their jobs. They are eager to keep improving themselves and to keep learning. Whereas older managers are more experienced and are less eager to be successful. Older managers have failed more often which gives them the knowledge to estimate the impact on the firms' reputation.

Another component of diversity is the gender of the person. Men and women have different characteristics which result in different managerial behavior. Jianakoplos and Bernasek (1998) have studied the risk aversion of women and men and concluded that men are likely to take more risk than women. Furthermore, the level of influence of gender varies in multiple aspects. The degree of masculinity or femininity within a country is playing a prominent role in determining the behavior of people (Meier-Pesti & Penz, 2008). The before mentioned studies have shown the differences in behavior caused by diversity. These differences are mainly based on the general population of a country (Croson & Gneezy, 2009; Adams & Funk, 2012). However, Faccio et al. (2014) state that the probability that these differences partly fade for directors in a board because they are not average civilians. To run a firm as a director requires a specific skill set which levels the differences to some extent. On the other hand, Adams and

(11)

Funk (2012) have conducted a survey in Sweden which showed that women and men on boards have different priorities. Women on boards are more ambitious, are eager for achievement and care more about power than ‘typical women’ (Adams & Funk, 2012). Following the same research, females in leadership positions regularly show less risk-averse behavior than males. Board gender diversity does influence the decision-making of a board, but it can be argued that this is not always because of differences between men and women.

Characteristics within a board determine the diversity which reflects on managerial behavior and decision-making. The nature of management skills is such that they are culturally specific: a management technique or philosophy that is appropriate in one national culture is not necessarily appropriate in another (Hofstede, 1984). Hofstede has done extensive research on how culture impacts the workplace. Hofstede defines culture as: "the collective programming of the mind distinguishing the members of one group or category of people from others."

Dimensions of national culture are linked to the decision-making of executives. In prior research, Breuer et al. (2011) show that individualism is linked to overconfidence and over-optimism which has a positive effect on risk-taking. An individualistic society consists of individuals who are focused on taking care of themselves and their immediate families. Its counterpart,

collectivism represents a society in which individuals can expect their family or members of a particular group to look out for them in exchange for loyalty to the ingroup (Hofstede, 1984). Graham et al. (2010) find that CEO's also are affected by their cultural background in their decision-making. CEOs from the US differ significantly from non-US CEOs regarding underlying attitudes in decision making. Moreover, on firm level, Mihet (2013) finds that domestic firms with low uncertainty aversion, high individualism and low tolerance for hierarchical relationships are showing more corporate risk-taking behavior.

Non-observable diversity is difficult to measure. There are few measures to describe a particular culture or diversity. This paper will build on the cultural dimensions of Hofstede. He conducted an extensive study of how values in the workplace are affected by culture. Between 1967 and 1973, Geert Hofstede analyzed an extensive database of value scores collected worldwide within IBM. Hofstede divided culture into six dimensions: power distance index (PDI), individualism versus collectivism (IDV), masculinity versus femininity (MAS), uncertainty avoidance index (UAI), Long term versus short term normative orientation (IND) and

indulgence versus restraint (IND). Each dimension has a score for more than 90 countries. These dimensions represent independent preferences for one state of affairs over another that distinguish countries from each other. However, this can be projected for individuals, as

(12)

Graham, Mihet, Meier-Pesti and Penz and Jianakoplos and Bernasek have done. Culture can affect the institutional and economic development at the macro level, the industrial

diversification and industry concentration can affect at the market level, as well as the corporate and individual level (Mihet, 2013).

At the micro level, culture has been shown to influence the behavior of individuals regarding risk-taking. The research Breuer et al. (2014) proved the link between overconfidence and risk-taking. Tse et al. (1998) observed that home culture could affect the decision-making of executives. More recently Graham et al. (2010) added research on this topic in the U.S. where they proved CEO’s are affected by culture as well. Their survey showed that characteristics derived from the culture of CEO’s have a strong association with uncertainty-aversion.

2.4 Risk-taking

Standard agency theories suggest that the interest of managers and shareholders are not always aligned (Jensen & Meckling, 1976). The separation of ownership creates differences in priorities between shareholders and managers. The appointed board of directors is expected to monitor the managers and prevent them from acting in their self-interest. Their internal governance function is of great importance to increase the shareholder value (Jensen & Meckling, 1976). However, there are incentives to act in their own interest and exhibit risk-taking behavior for their own benefit, although this might not increase the shareholder value. Directors are expected to monitor the organization and steer managers to avoid agency conflicts. Research of John et al. (2008) shows that managers take the risk of losing their jobs into account when they face

corporate risk-taking. Still, Hillman et al. (2003) found that the composition of a board should contain mixed characteristics to perform. There is no proven optimal composition to take advantage of the human capital in the board to minimize the agency problem.

2.4.1 Corporate risk-taking

The most used definition of risk is: “Risk is a measure of the probability and severity of adverse effects” (Lowrance, 1976). Croson and Gneezy (2009) argue that risk-taking is a result of human behavior. The level of risk is dependent on whether the behavior could lead to more than one outcome and whether some of these outcomes are undesirable.

In case of risk-taking in a financial environment, the risk is measured in several ways. Sharpe (1964), Faccio et al. (2011) and Stellingerwerf (2016) researched this field and measured risk by calculating the standard deviation of the return on assets. Claessens et al. (2000) have developed a comprehensive calculation of risk within firms which is often used in prior literature. They

(13)

divided 14 different indicators into four common fields of risk: liquidity risk, solvency risk, financial risk and corporate risk policy. This paper uses indicators from the latter three groups and an external measure. Stock price volatility is an indicator of risk that is frequently used in the literature, mostly in studies related to bank risk-taking. However, Cheng (2008) uses stock price volatility to study the relationship between board size of non-financial firms and the variability of corporate performance.

2.5 Hypothesis / theory development

In this section, the discussed literature is used to formulate the hypotheses. 2.5.1 Masculinity versus femininity (MAS)

The first dimension analyzed is masculinity (MAS). Hofstede defines it as referring ‘‘to the dominant gender role patterns: the patterns of male assertiveness and female nurturance'' (Hofstede, 2003, p. 280). Wilson and Daly (1985) created a sociological model which points out that gender differences occur due to ‘‘masculine psychology’’. Men are overly represented in traditionally risky businesses since they historically faced more competition than women to obtain the right to mate and climb the social ladder. Today this theory is still used, and we still assume men to have a greater risk appetite. According to Hofstede (1984), cultures with a high masculinity orientation tend to be oriented towards things and money, tend to value

independence, tend to “live to work’’ and to value decisiveness. Cultures with low masculinity orientations "work to live", are serviceable, are oriented towards people and emphasize intuition rather than decision-making on observable fact. Just like Jianakoplos and Bernasek (1998), this indicates that more masculine cultures tend to be less risk-averse than women.

On the other hand, following the research of Berger et al. (2014), female directors are not risk-averse. Their results showed that when female representation on a board grew, risk-taking increased as well. Although this study was among financial firms, Adams and Ragunathan (2017) found the same relationship of women having the same risk appetite as men.

Croson and Gneezy (2009) found three fundamental differences regarding risk, social and competitive preferences between males and females. Women tend to have less appetite for competition, they are more sensitive to social cues and overall are more risk-averse. Based on the aforementioned theories and empirical evidence I have formulated the hypothesis:

H1: Masculinity exhibits a positive association with risk-taking behavior

2.5.2 Uncertainty avoidance index (UAI)

(14)

valuing beliefs and institutions that provide certainty and conformity’’ (Hofstede, 2003). Although uncertainty is not the same as risk, there are similarities. The main difference is that with uncertainty the outcome can be either positive or negative (Knight, 2012). For risk, it is certain that the outcome is adverse. Uncertainty, therefore, consists of the risk of an unfavorable outcome and the chance of a positive outcome. As stated in the theory section a high uncertainty avoidance index, causes less risk-taking behavior (Mihet, 2013). McGrath, MacMillan, and

Scheinberg (1992) conclude that low UAI is associated with less resistance to change, stronger achievement motivation, optimism, stronger ambition for individual advancement and more risk-taking. Thus, based on the aforementioned theories and empirical evidence I have formulated the hypothesis:

H2: Uncertainty avoidance exhibits a negative association with risk-taking behavior

2.5.3 Individualism versus collectivism (IDV)

IDV describes the relationship between the individual and the society. Breuer et al. (2014) show that individualism is linked to overconfidence and over-optimism which has a positive effect on risk-taking. Overconfident people tend to perceive the ability of total control and knowledge, whereas this overconfidence is in fact exaggerated. This overconfidence leads to a higher risk-tolerance for individualistic societies (Pan & Statman, 2009). When individuals are too confident in their plan to succeed, the accuracy of their work can suffer. Their predictions become less precise whereas the individual is sure about his or her qualities (Van der Steen 2004; Grinblatt and Keloharju 2009). In firms, we can project overconfidence on individualistic directors who are free to make risky decisions on their judgment. They are eager to compete against and stand out to others. McGrath, MacMillan, and Scheinberg (1992) state that a stronger ambition for individual advancement is positively related to a high UAI. Based on the aforementioned theories and empirical evidence I have formulated the hypothesis:

(15)

3 Research method

In the research method section, I will elaborate on my empirical approach. The goal of this study is to explore the nature of the relationship between boardroom diversity and risk-taking

behavior. To measure boardroom diversity, I have used a proxy which consists of the scores of three different cultural dimensions from Hofstede (1984). The proxy is further explained in paragraph 3.3.1. The risk-taking behavior measures are explained in paragraph 3.3.3.

3.1 Conceptual model

Below, the conceptual model for this research is displayed. The relationship that is examined is between board diversity and risk-taking behavior. The control variables for this relationship are average age, ratio executives to non-executives, industry information opacity, firm size and proportion of females represented on the board (gender). In the previous paragraph, the hypotheses are developed by the literature. The theories and prior research predict an effect from the independent variable on the dependent variable as can be seen below.

3.2 Regression equation

To test the relationship of the different variables on risk-taking behavior, it is necessary to create a model that takes into account both the dependent and the independent variables and the control variables.

The model is defined as follows:

𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑗𝑗,𝑡𝑡 = 𝛼𝛼0+ 𝛽𝛽1𝑀𝑀𝑀𝑀𝑀𝑀𝑡𝑡+ 𝛽𝛽2𝑈𝑈𝑀𝑀𝑈𝑈𝑡𝑡+ 𝛽𝛽3𝑈𝑈𝐼𝐼𝐼𝐼𝑡𝑡+ 𝛽𝛽4𝑀𝑀𝐴𝐴𝐴𝐴𝑡𝑡+ 𝛽𝛽5𝐹𝐹𝑅𝑅𝐹𝐹𝐹𝐹𝑅𝑅𝑅𝑅𝐹𝐹𝐴𝐴𝑡𝑡

+ 𝛽𝛽6𝑈𝑈𝐼𝐼𝐼𝐼𝐼𝐼𝐹𝐹𝐹𝐹𝐼𝐼𝐼𝐼𝑅𝑅𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝑅𝑅𝐼𝐼𝐼𝐼𝑡𝑡+ 𝛽𝛽7𝐺𝐺𝐴𝐴𝐼𝐼𝐺𝐺𝐴𝐴𝐹𝐹𝑡𝑡+ 𝛽𝛽8𝐸𝐸𝐸𝐸𝐴𝐴𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐴𝐴𝐸𝐸𝐴𝐴𝐼𝐼𝑡𝑡+ λ𝑡𝑡+ 𝜀𝜀𝑗𝑗,𝑡𝑡

Risk-taking behavior is measured by four alternative dependent variables. 𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑗𝑗,𝑡𝑡 stands

for the dependent variables: σROA, Altman Z-score, stock price volatility and research and Independent variable

Board diversity

Dependent variable Risk-taking behavior

Control variables

Age Firm size

Executive/ Non-executive Gender Information opacity

(16)

development intensity at time 𝐼𝐼. Where 𝛼𝛼0 is a constant. The proxies for the independent

variable culture are scores on the cultural dimension dimensions from Hofstede. This will be on three different dimensions: average score of the board on individualism index (𝑈𝑈𝐼𝐼𝐼𝐼𝑡𝑡), the average

score of the board on the uncertainty avoidance index (𝑈𝑈𝑀𝑀𝑈𝑈𝑡𝑡) and the average score of the board

on the masculinity index (𝑀𝑀𝑀𝑀𝑀𝑀𝑡𝑡). The control variables are the average age of the board at 𝐼𝐼

(𝑀𝑀𝐴𝐴𝐴𝐴𝑡𝑡), the percentage of females represented on the board (𝐺𝐺𝐴𝐴𝐼𝐼𝐺𝐺𝐴𝐴𝐹𝐹𝑡𝑡), industry score on the

information opacity index (𝑈𝑈𝐼𝐼𝐼𝐼𝐼𝐼𝐹𝐹𝐹𝐹𝐼𝐼𝐼𝐼𝑅𝑅𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝑅𝑅𝐼𝐼𝐼𝐼𝑡𝑡) and ratio executives / non-executives

(𝐸𝐸𝐸𝐸𝐴𝐴𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐴𝐴𝐸𝐸𝐴𝐴𝐼𝐼𝑡𝑡). Additionally, there are time dummies (λ𝑡𝑡) as controls for time. Lastly,

𝜀𝜀𝑗𝑗,𝑡𝑡 stands for the robust standard errors which are included in all regression models to control

for serial correlation and heteroscedasticity. 3.3 Explanation of variables

3.3.1 Independent variables

Following the recent literature, the cultural framework of Hofstede (1984, 2003), Hofstede et al. (2010) is used to measure culture. Subsequently, the cultural framework is used to proxy for diversity in boardrooms. Geert Hofstede is a Dutch sociologist who created a framework for cultural dimensions theory which describes the effects of a society’s culture on the values of its members and how these values relate to behavior. Initially, Hofstede measured culture on five different dimensions: power distance index (also, PDI), individualism versus collectivism (also, IDV), uncertainty avoidance index (also, UAI), masculinity versus femininity (also, MAS) long-term orientation versus short-long-term orientation (also, LTO). In 2010 a sixth dimension was added: indulgence versus restraint. To create a scope, I have chosen to focus on three dimensions which are widely used in prior literature (Li et al. 2013; Mihet, 2013).

Hofstede examined a psychological survey under IBM subsidiaries in 40 different countries between 1967 and 1973. With this data, he computed scores for several cultural dimensions. The framework has been refined since and expanded to almost 100 countries. Each score is between 0 and 100 and reflects the position of each country relative to the other

countries. The number that is taken into account in examining the effect is the average score of all directors on the board for one specific cultural dimension. A firm has one score for each cultural dimension. For example, all the directors’ home countries have a score on the MAS index. These scores are added up and divided by the number of directors.

𝛽𝛽2𝑀𝑀𝑀𝑀𝑀𝑀 (𝑀𝑀𝐴𝐴𝐴𝐴𝐹𝐹𝐼𝐼𝐴𝐴𝐴𝐴 𝑅𝑅𝐼𝐼𝐼𝐼𝐹𝐹𝐴𝐴 𝑀𝑀𝑀𝑀𝑀𝑀 𝑅𝑅𝐼𝐼𝐺𝐺𝐴𝐴𝐸𝐸) =MAS index score country A + B + C …

(17)

The equation above also applies for the cultural dimensions UAI and IDV. To normalize the distributions of the variables and reduce the impact of outliers, all three independent variables are winsorized at the 5% level in both tails. The efficiencies of the estimates of slope and the goodness-of-fit can decrease when cleaning the data. Still, multiple studies which have used Hofstede’s cultural dimensions have used winsorization (Griffin et al., 2017; Shao et al. 2013; Mihet, 2013; Stellingwerf, 2016). Additionally, I have calculated the logarithm of UAI and IDV to normalize the data which resulted in an approximate normally distribution.

3.3.2 Dependent variables Standard deviation ROA

The first proxy for risk-taking behavior is the standard deviation of return on assets (ROA). When firms take riskier decisions, the return on assets tends to deviate more (Mihet, 2013). The ROA particularly measures the degree of a firm’s operational risk-taking based on the volatility of corporate earnings (Faccio et al., 2014; Li et al., 2013; Mihet, 2013). The starting point for this measure is the yearly earnings before interest and taxes (EBIT). Subsequently, EBIT is divided by the total assets which result in the ROA. The standard deviation is calculated, based on the ROA’s of the largest 500 listed firms in the same industry in the UK.

σROA = �∑(𝐸𝐸𝐼𝐼 − 1 𝑖𝑖𝑗𝑗− 𝐸𝐸̅)2

Altman Z-score

Second, the Altman Z-score proxies for risk-taking behavior. The Z-score is a predictor of financial distress of a company (Altman, 2000). The higher the score, the more the firm is able to pay its debts. The variables are classified into five standard ratio categories, including liquidity, profitability, leverage, solvency, and activity. Components of the Z-score are working capital ratio, retained earnings ratio, earnings before interest and taxes ratio, the market value of equity compared to total liabilities ratio and a sales ratio. All ratios, except for the market value of equity, are scaled by the total assets of the firm. The final discriminant function is as follows (Altman, 2000):

𝑍𝑍 = 0.012X1 + 0.014X2 + 0.033X3 + 0.006X4 + 0.999X5

Where X1 = working capital/total assets, X2 = retained earnings/total assets,

(18)

X3 = earnings before interest and taxes/total assets, X4 = market value equity/book value of total liabilities, X5 = sales/total assets, and

Z = overall index. R&D intensity

Research and development expenses are often associated with risk-taking behavior. Total R&D expenditures are taken into account because the capitalized R&D expenditures are incomplete for the sample. There is not made any further distinctions in expensed or capitalized R&D expenditures. Due to the length of the investment and the uncertainty of the amount of costs the probability of returns is relatively little (Cox, 2011). Hence, we can say that R&D expenditures are a measure of corporate risk policy. Brandenburg (1964) said that the nature of research and development investments make project riskier than investments in other functional departments. Therefore, higher R&D expenses are associated with more risk-taking behavior (Kwok et al., 2013). This variable is measured by taking the research and development expenses divided by the operating revenue.

𝑅𝑅&𝐼𝐼 𝑅𝑅𝐼𝐼𝐼𝐼𝐴𝐴𝐼𝐼𝑅𝑅𝑅𝑅𝐼𝐼𝐼𝐼 =𝐼𝐼𝐼𝐼𝐴𝐴𝐹𝐹𝐼𝐼𝐼𝐼𝑅𝑅𝐼𝐼𝐴𝐴 𝐹𝐹𝐴𝐴𝐴𝐴𝐴𝐴𝐼𝐼𝑟𝑟𝐴𝐴 R&D expenses

Volatility

Stock price volatility is an often used proxy for risk-taking behavior. Despite the measure is mostly used for bank risk-taking (Pathan, 2009), Cheng (2008) used this measure to proxy for taking for non-financial firms. He predicted that when stock price volatility increases, risk-taking behavior increases too. Volatility is measured by the stock's average annual price movement to a high and low from a mean price for each year.

3.3.3 Control variables Firm size

Larger firms tend to show less risk-taking behavior because of their opportunities to spread geographically and to diversify (Bruno & Shin, 2014; Nguyen, 2012; Faccio et al., 2014). Additionally, Stellingwerf (2016) found that firm size is negatively associated with risk-taking, specifically with the extent of R&D expenditures. Firm size is measured as the logarithm of a firm’s total assets, which tends to be normally distributed (Bloom & Milkovich, 1998).

(19)

Gender

Furthermore, the gender of the directors will serve as a control variable. Gender is measured as the percentage of female board directors active at the firm. Jianakoplos and Bernasek (1998) say that more masculine cultures are less risk-averse than feminine cultures. Croson and Gneezy (2009) found fundamental differences regarding risk preferences. Namely, women have less risk appetite than men and are less competitive as well.

Age

The average age of the directors will be included as a control variable. The age of a person could influence the management of a firm. The work of Ali et al. (2014) points out that low average age of a board is associated with high market values. It is said that younger directors may take riskier decisions than older directors. Research by Wiersema et al. (1992) shows that younger directors implement significant changes in strategy more easily than older directors.

Information opacity

Information opacity tells us something about the amount of information in a particular industry which is available for the market. When an industry is more opaque, it is more difficult to make considered decision about a firm. Morgan (2002) states that the high scale of divergence between bond raters implies that financial, insurance, oil and gas, mining, and IT firms are inherently more opaque than other types of firms. Uncertainty is present because it is hard to monitor and determine their assets and liabilities. The measure that is used in this study is the inverse of the private information available, the informational opacity. For the proxy, this measure uses an indicator of "relative firm-specific stock return variation" created by Durnev et al. (2004) and Rajan and Zingales (1998), which provides a score for stock price informativeness at the industry level. Table 1 displays the opacity index created by Durnev et al. (2004) and Rajan and Zingales (1998). They created this index by collecting high-frequency firm-level data. Stock prices become less informative when there is little firm-specific data available. Industry and market-related factors can explain most of the volatility in stock price returns. Data below is calculated based on stock price informativeness of firms in the United States from 2000 until 2005. The exact

(20)

Industry Opacity index

Wholesale Trade 5.550

Mining 5.477

Transportation, Communications, Electric, Gas and

Sanitary service 5.123

Retail Trade 5.111

Services 5.019

Agriculture, Forestry, and Fishing 4.986

Construction 4.931

Manufacturing 4.756

Table 1: opacity index per industry

Executive and non-executive directors

The number of executives compared to the number of non-executives within the board is used as ratio exec/non-exec. Herewith we can observe if this ratio has any influence on risk-taking behavior. Where the management is in the hands of the executive directors, the non-executive directors are responsible for the supervision on the management. Together they form the board of directors of a firm. Below the equation for calculating the proportion of executives in the board of directors.

(21)

4 Data

4.1 General

I have collected my research data from DataStream. Here I can collect the financial data of the top 250 listed firms in the United Kingdom. The personal board member data is available through Orbis. This database contains data about nationalities, age, board functions and the industry. I have chosen the UK because the top listed firms have considerable diverse boards compared to other leading economies.

DataStream

The financial data is retrieved from Datastream, available at the University of Amsterdam. Datastream is a financial database for company data, share prices and macroeconomic data. The data covers 175 countries and goes back 50 years. To obtain a sample that is large enough, the 250 largest listed firms in the UK will be analyzed.

Hofstede insights

As mentioned in the research method paragraph, Hofstede’s cultural dimension scores are used to proxy for boardroom diversity. Scores for almost 100 countries can be found and retrieved from the website: https://www.hofstede-insights.com/.

Orbis

Orbis is a database available at the University of Amsterdam and contains global company data for listed and non-listed firms. Listed firms are described more extensively than non-listed firms. The following characteristics are retrieved via Orbis: appointment date, resignation date, age, gender, and nationality.

4.2 Sample

For this paper, the data consists of three types: financial data, cultural dimension data and the characteristics of the directors. UK’s 250 largest listed firms are selected for the sample. The period which will be studied covers the years 2008-2017. As the period contains ten years, the initial sample size consisted of 2500 firm years. Due to comparability problems, the risk-taking behavior measures do not permit financial firms to be included in the sample. Financial firms have such different balance sheets that they distort the risk-taking measures and become incomparable. Table 2 presents the sample selection process.

(22)

# Observations Description

2500 Number of firms times number of years - 340 Financial services industry excluded

- 136 Cultural dimension scores not available for countries 2024 Total number of observations

Table 2: Sample selection process

In the table below the different industries which are represented are shown. Out of 214 firms, manufacturing firms are represented the most, followed by services and retail trade firms. As seen below Transportation, Communications, Electric, Gas and Sanitary services come next followed by mining, construction and wholesale trade. Table 3 below presents how the firms are distributed over the different industries.

Industry Frequency Percentage

Mining 17 8%

Construction 13 6%

Manufacturing 58 27%

Transportation, Communications, Electric, Gas and Sanitary services 29 14%

Wholesale Trade 9 4%

Retail Trade 42 20%

Services 46 21%

214 100%

Table 3: Industry distribution

4.3 Descriptive statistics 4.3.1 Core statistic values

An overview of the descriptive statistics is displayed in table 4 below. As can be seen in table 4, the number of R&D intensity observations is lower than the other dependent variables. R&D intensity data is less available in databases because this is a detailed variable and not all firms have significant R&D expenses. The data for the other variables is approximately equally available. The average Altman Z-score is 3,17. When the Altman Z-score becomes lower, a firm shows more risk-taking behavior. The minimum score is -2,5 which means that a firm is highly insolvent and cannot pay its debts. Scores above 3 indicate firms are not likely to go bankrupt. With a maximum score of 48,3, the difference between the minimum and maximum score is relatively high. This can be explained by the strong balance sheet of firms. Some firms are highly solvent and consequently can pay their debts back multiple times.

For the dependent variable R&D intensity, the average number is 3,04. This means the firms of my sample expense 3,04% of their operating revenue on research and development. The higher this percentage, the more risktaking behavior a company exhibits. The lowest score is

(23)

-0,1, and the maximum score is 39,34. Industries as wholesale trade, retail trade and services have relatively low research and development expenditures. Whereas the transportation,

communications, electric, gas and sanitary services industries, the manufacturing industry and the mining industry have relatively high research and development expenditures (Mihet, 2013). The standard deviation of ROA has an average of 4,34, where a higher standard deviation means more risk-taking behavior. The lowest score of 0,01 means the ROA of a firm barely changed compared to the previous year. The maximum score of 59,72% indicates a highly volatile ROA. The mean for volatility is 26,7, this indicates that the maximum change in stock price in a year is 26,7 on average. Similar to R&D and standard deviation ROA, a higher number means more risk-taking behavior.

All independent variables have the same amount of observations, i.e., 1777. Firm years are solely included when the values for cultural dimensions are available. This resulted in 1777 different values for masculinity, uncertainty avoidance index and individualism. To normalize the data, the logarithm is computed for UAI and IDV. Normally scores on the cultural dimension vary from 0 to 100.

Dependent N Average Standard

deviation Low High

Altman 1.718 3,17 3,10 -2,50 48,3 R&D 591 3,04 5,46 -0,10 39,3 σ ROA 1.657 4,34 4,68 0,01 59,7 Volatility 1.577 26,7 8,61 10,8 60,2 Independent MAS 1.777 63,52 3,70 53 66 UAI 1.777 3,65 0,12 3,56 3,93 IDV 1.777 4,45 0,06 4,25 4,49 Control Information opacity 1.777 5,04 0,23 4,76 5,55 Age 1.705 55,38 3,69 39 67 Gender 1.768 0,17 0,12 0,0 0,67 Firm size 1.768 9,37 0,68 7,54 11,58 Exec/Non-exec 1.607 0,36 0,16 0,15 0,73

Table 4: Core statistic values. The Altman Z-score is a predictor for financial distress where a lower number means more likely

to go bankrupt. σROA is standard deviation of return on assets. The standard deviation is calculated with the ROA of the firms in the industry in the concerning year. Volatility stands for the stock's average annual price movement to a high and low from a mean price for each year. MAS, UAI, and IDV are Hofstede's cultural dimensions. MAS is the score on the masculinity index, UAI is uncertainty avoidance index and IDV is the individualism score. All three variables are winsorized and UAI and IDV are displayed as the logarithm of their score. Information opacity measures how opaque an industry is. Age is the average age of the board. Gender is the % of females in the board. Firm size is the logarithm of total assets. Exec/Non-exec is the % executives on the board.

The control variable information opacity is calculated according Rajan and Zingales’ (1998) paper. The score varies from 4,76 to 5,55 where a low score means low opacity in the

(24)

industry and subsequently less risk-taking behavior. Vice versa, a high opacity is linked to riskier behavior for a firm. Control variable age is the average age of the directors in a firm year. The sample’s youngest board in a firm year is 39 on average and the oldest 67 years on average. Gender stands for the proportion of females present a board of directors. There are boards without women and the highest representation women is two-thirds of women. Firm size is measured by its logarithm. The average firm size is £9,9 billion of total assets. The last control variable, is the proportion of executives to non-executives on a board. The lowest proportion is 16% executives compared to non-executives and the highest 73%. The sample’s average

proportion on the board is 36% executives compared to non-executives. 4.3.2 Correlation matrix

The correlation matrix is displayed below in table 5. Before showing the regression results, we first examine the correlations between the independent variables. These correlations are used to understand the strength and direction of the relationship. All three independent variables show mixed relationships with the risk-taking behavior, some in contrast with the hypotheses. We can observe that the highest, the correlation value is -67 percent. This is just within the range of -70 and 70 percent correlation to not have any multicollinearity problems following Brooks (2008). However, paragraph 4.3.3 further elaborates on the possible multicollinearity in this data.

(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (1) Altman 1,00 (2) R&D 0,09 1,00 (3) σ ROA 0,16 0,13 1,00 (4) Volatility 0,05 -0,15 0,06 1,00 (5) MAS 0,11 0,00 -0,05 0,18 1,00 (6) UAI -0,10 -0,01 0,04 -0,07 -0,65 1,00 (7) IDV 0,10 0,12 -0,05 -0,07 0,65 -0,67 1,00 (8) Information opacity 0,06 -0,17 0,07 0,24 -0,06 0,07 -0,16 1,00 (9) Age -0,07 0,22 0,02 -0,11 -0,04 0,03 -0,04 0,01 1,00 (10) Gender -0,03 0,07 -0,02 -0,13 0,03 0,01 0,08 -0,08 0,18 1,00 (11) Firm size -0,29 -0,06 -0,05 -0,33 -0,38 0,46 -0,37 0,05 0,23 0,11 1,00 (12) Exec/Non-exec -0,02 0,01 0,02 -0,07 -0,15 0,19 -0,22 0,01 0,11 0,31 0,10 1,00

Table 5: pairwise correlation matrix. The Altman Z-score is a predictor for financial distress where a lower number means more likely to go

bankrupt. σROA is standard deviation of return on assets. The standard deviation is calculated with the ROA of the firms in the industry in the concerning year. Volatility stands for the stock's average annual price movement to a high and low from a mean price for each year. MAS, UAI and IDV are Hofstede’s cultural dimensions. MAS is score on the masculinity index, UAI is uncertainty avoidance index and IDV is the individualism score. All three variables are winsorized and UAI and IDV are displayed as the logarithm of their score. Information opacity measures how opaque an industry is. Age is the average age of the board. Gender is the % of females in the board. Firm size is the logarithm of total assets. Exec/Non-exec is the % executives on the board.

(25)

4.3.3 Multicollinearity

When performing a regression analysis, the relationship between the independent (predictor) variables is of importance. It is not desirable to have a strong relationship between independent variables. A strong relationship between the independent variables creates difficulty in estimating the regression coefficients (Field et al., 2013). An often used method to test the multicollinearity is a variance inflation factors (VIF) test. Table 6 below shows the VIF for every relationship. Every VIF is below 3, whereas the acceptable cut-off is 10 (Field et al., 2013). Therefore, there is no concern for multicollinearity in the data.

VIF Altman R&D ROA Volatility

UIA 2.41 2.41 2.32 2.32 IDV 2.27 2.34 2.23 2.30 MAS 2.23 2.11 2.23 2.17 Firm size 1.39 1.57 1.40 1.37 Gender 1.21 1.31 1.21 1.23 Exec/Non-Exec 1.21 1.28 1.20 1.21 Age 1.11 1.25 1.11 1.11 Information opacity 1.02 1.07 1.02 1.02 Mean VIF 1.54 1.59 1.53 1.53

(26)

5 Results

Building upon the previous section where the correlation matrix is shown, the table in paragraph 5.1 presents the results of the OLS regression for the R&D, Altman Z-score, standard deviation ROA and volatility models. I will elaborate on the regression results in paragraph 5.2. Paragraph 5.3 contains the results of a sensitivity test.

5.1 Main results

Table 7: Table significance: * = p<0.05, ** = p<0.01, *** = p<0.001. β = beta P = p-value. OLS regression results The Altman Z-score is a predictor for financial distress where a lower number means more likely to go bankrupt. σROA is standard deviation of return on assets. The standard deviation is calculated with the ROA of the firms in the industry in the concerning year. Volatility stands for the stock's average annual price movement to a high and low from a mean price for each year. MAS, UAI and IDV are Hofstede’s cultural dimensions. MAS is score on the masculinity index, UAI is uncertainty avoidance index and IDV is the individualism score. All three variables are winsorized and UAI and IDV are displayed as the logarithm of their score. Information opacity measures how opaque an industry is. Age is the average age of the board. Gender is the % of females in the board. Firm size is the logarithm of total assets. Exec/Non-exec is the % executives on the board.

The relationship between risk-taking and board diversity is studied by carrying out four separate regressions which resulted in four different models. Table 7 shows the results for the four different models. Of the first model, the dependent variable R&D measures the intensity of the research and development expenditures compared to the operating revenue. When a firm's Z-score has a higher R&D intensity, it shows more risk-taking behavior. In model 2, we can see the Altman Z-score as dependent variable. This is a predictor for financial distress. When a firm comes closer to 0, the firm is less able to pay its debts. Therefore, a higher score in the second model leads to less risk-taking behavior. The third model contains dependent variable standard deviation ROA. When firms show more deviations in their returns on assets, they tend to take more risks. Therefore, a higher standard deviation ROA means more risk-taking behavior. Lastly,

(1) (2) (3) (4)

R&D Altman σ ROA Volatility

Prediction β P β P β P β P MAS + -0,214 0,009 ** 0,025 0,237 0,042 0,568 0,246 0,002 ** UAI - 8,893 0,055 6,019 0,000 *** 8,645 0,021 * 6,249 0,260 IDV + 49,39 0,000 *** 5,245 0,201 -5,812 0,557 -75,37 0,000 *** Information opacity + -4,893 0,000 *** 0,901 0,012 * 1,803 0,004 ** 8,718 0,000 *** Age - 0,445 0,000 *** 0,019 0,314 0,033 0,422 * -0,038 0,502 Gender - -2,0 0,323 -0,518 0,585 0,713 0,544 -2,442 0,239 Firm size - -0,220 0,547 -1,544 0,000 *** -0,771 0,000 *** -4,547 0,000 *** Exec / Non-exec - -0,162 0,925 0,807 0,254 1,405 0,183 -0,475 0,754 Constant -91,0 0,000 *** -10,13 0,300 -3,52 0,835 149,1 0,000 *** Adj. R-sq 0,111 0,101 0,019 0,211 F-statistic 3,71 0,000 *** 24,73 0,000 *** 2,32 0,001 ** 27,62 0,000 ***

(27)

the fourth model has volatility as dependent variable. More stock price volatility means more risk which is in line with more risk-taking behavior.

5.2 Discussion of results

Based on the results displayed in table 7, the goodness of fit, significance and test results of the hypotheses are discussed in the following paragraph.

5.2.1 Model 1: Association with R&D intensity Goodness of fit

In table 7, the association of boardroom diversity with R&D intensity is measured. The adjusted R-squared of model 1 is 0,111, which indicates that 11,1% of the association with R&D intensity is explained by the model. A significant relationship can be observed between R&D intensity and the independent variables with an F-statistic of 3,71 with a significance level of p<0,01. This implies that one or more variables are significant for the outcome of R&D intensity and the overall fit for the model is good (Field et al., 2013).

Sign of β’s Independent variables

For the first the model, the hypothesized association of masculinity and risk-taking behavior is positive. According to Hofstede (2003), cultures with a high masculine orientation tend to be oriented towards things and money, tend to value independence and to value decisiveness. Jianakoplos and Bernasek (1998) state that masculine cultures tend to be less risk-averse than feminine cultures. The coefficient of -0,214 shows that an increase in the masculinity score is associated with risk-taking behavior of a firm regarding R&D intensity. This significant association shows the contrary of what is predicted in the hypothesis. Consequently, the first hypothesis is rejected.

As stated in the theory section a low uncertainty avoidance index can lead to less risk-taking behavior (Mihet, 2013). McGrath, MacMillan, and Scheinberg (1992) conclude that a low uncertainty avoidance index is associated with less resistance to change, stronger achievement motivation, optimism, stronger ambition for individual advancement and more risk-taking. The coefficient is 8,893, which means that an increase in the uncertainty avoidance index, can lead to an increase in R&D intensity as well. Similar to masculinity, this is contrary to the hypothesis which rejects hypothesis two for the first model.

(28)

Individualism describes the relationship between the individual and the society. Breuer et al. (2014) show that individualism is linked to overconfidence and over-optimism which has a positive effect on risk-taking. McGrath, MacMillan, and Scheinberg (1992) state that a stronger ambition for individual advancement is positively related to a higher UAI. The hypothesized positive association between individualism and risk-taking behavior is confirmed by the positive coefficient of 49,39 and significance on a 0,1% level. Hypothesis two is not rejected.

Control variables

Following Morgan (2008), uncertainty comes from certain assets which are particularly difficult to monitor or price and creates more opacity in an industry. It was expected that industries which are more opaque and have a higher opacity score, that firms would show more risk-taking behavior (Huang, 2008). The coefficient is -4,893, which indicates that when the information opacity increases by 1, the R&D intensity decreases with -4,893. This is not in line with the literature, which predicts more risk-taking behavior when industries become more opaque. The association between information opacity and R&D intensity is significant on a 0,1% level Concerning the control variable age, the research of Ali et al. (2014) points out that a young average age of a board is associated with high market values and that young directors take riskier decisions. Research by Wiersema et al. (1992) shows that younger directors undergo major changes in strategy more easily than older directors. The coefficient is 0,445 and confirms the hypothesis that boards with a younger average age take on more risk. When the average age drops by 1 year, the R&D intensity increases by 0,445. The relationship shows statistical

significance on a 0,1% level. The next control variable is gender. Following Croson and Gneezy (2009), men and women have fundamental differences regarding risk preferences, social

preferences, and competitive preferences. Their surveys showed that women are more risk-averse, more sensitive to social cues, and women are less competitive than men in general. The coefficient is -2, which implies that if there are more women on the board, the R&D intensity will increase. This is in contrast with the expectations derived from the literature. Larger firms usually show less risk-taking behavior (Stellingwerf, 2016). The confirming coefficient of -0,220 indicates that if the firm size increases, the R&D intensity decreases. While there is not much research on the association between risk-taking behavior and the ratio of executives and non-executives, O’Connell and Cramer (2010) have found a significant relationship between this ratio of executives and non-executives and firm performance. They have found that when there are more non-executive directors on a board, the ROA increases. Although this is not directly linked with the standard deviation of ROA, an increase of ROA can

(29)

lead to an increase in standard deviation ROA. However, the insignificant result of my regression shows the contrary. It rejects the hypothesis that an increase in executives decreases risk-taking behavior.

5.2.2 Model 2: Association with the Altman Z-score Goodness of fit

In table 7, the association of boardroom diversity with the Altman Z-score is measured. The adjusted R-squared of the second model is 0,101, which indicates a weak explanatory power. The model explains 10,1% of the association on R&D intensity. A significant relationship can be observed between the Altman Z-score and the independent variables with an F-statistic of 24,73 and p<0.001. This implies that one or more variables are significant for the outcome of Altman Z-score and the overall fit for the model is good (Field et al., 2013).

Sign of β’s Independent variables

For the second model, masculinity shows a coefficient of 0,025, this is in line with the hypothesis. However, there is no statistical significance. This rejects hypothesis two for the second model. The coefficient for uncertainty avoidance index is 6,019, which means an increase in UAI leads to an increase in Altman Z-score. That a higher score on the UAI leads to less risk-taking behavior is in line with the hypothesis (Mihet, 2013). Furthermore, this relationship shows a significance of 0,1%. Hypothesis two is not rejected for the second model. For individualism, the coefficient is 5,245, which means that an increase in the IDV score leads to less risk-taking behavior. This is not in line with the prediction, while there is also no statistical significance. Therefore, hypothesis three is rejected for the second model.

Control variables

A more opaque industry tends to lead to more risk-taking behavior. Contrary to the literature, the regression results show that an increase in opacity leads to less risk-taking behavior. The association between information opacity and standard deviation ROA is significant on a 5% level. For the predictor variable, age, an increase also leads to less risk-taking behavior. We can observe that this prediction has been confirmed in the results. However, there is no statistical significance. The coefficient for gender is -0,518. This indicates that when there are more women on the board, there is less risk-taking behavior. On the contrary, the research of Jianakoplos and Bernasek (1998) shows that men are likely to take more risk than women. The results show no statistical significance. For the variable, firm size, the coefficient is -1,54, which indicates that

(30)

when the firm size grows, risk-taking behavior declines. This is not in line with the predictions from Stellingwerf (2016) who predicts that larger firms take fewer risks. The relationship is significant with p<0,01. We can observe a coefficient of 0,807 for the ratio of executive and non-executive board members. This implies that when the number of non-executive directors on a board increases, the risk-taking behavior declines. This result is in line with the prediction derived from the literature.

5.2.3 Model 3: Association with standard deviation ROA Goodness of fit

In table 7, the association of boardroom diversity with standard deviation ROA is measured. The adjusted R-squared of model 3 is 0,019, which indicates that the model explains 1,9% of the influence on R&D intensity. A significant relationship can be observed between the standard deviation ROA and the independent variables with an F-statistic of 2,32 and p<0.01. This implies that one or more variables in the model are significant.

Sign of β’s

Independent variables

For masculinity, the coefficient is 0,042. This indicates slightly more risk-taking behavior when the masculinity score increases and is in line with the hypothesis. Though there is no significance in the relationship between masculinity and standard deviation ROA which rejects the first hypothesis for model three. The prediction for the uncertainty avoidance is that when this score increases, the risk-taking behavior decreases. The coefficient of 8,645 suggests otherwise and is in contrast with the hypothesis. Both for masculinity and uncertainty avoidance there is no significant relationship with risk-taking behavior. Consequently, the hypothesis for the relationship between individualism and standard deviation ROA is rejected. The result for individualism is also in contrast with my prediction. The negative coefficient implies that when there is more individualism, the risk-taking behavior decreases. Moreover, there is no statistical significance which rejects the third hypothesis.

Control variables

For information opacity, the coefficient is 1,803, which indicates more risk-taking behavior with an increase in information opacity. This is in line with the hypothesis. Age shows a small positive coefficient where in fact the literature predicts the contrary. The relationship is significant with p<0,05. For the variable, gender, the prediction is contradicted with a coefficient of 0,713.

(31)

The variable firm size shows a significant association with risk-taking behavior. The results imply that larger firms show less risk-taking behavior and significant with p<0,001. While there is not much research on the association of risk-taking behavior and the ratio of executives and non-executives, O’Connell and Cramer (2010) have found a significant relationship between this ratio and firm performance. They have found that when there are more non-executive directors on a board, the ROA increases. Although this is not directly linked with the standard deviation of ROA, an increase in ROA does mean an increase in the standard deviation of ROA. However, the insignificant result of my regression shows the contrary. My results reject the hypothesis that an increase in the number of executives decreases risk-taking behavior. At the same time, it does not reject that an increase in executives, increases risk-taking behavior.

5.2.4 Model 4: Association with volatility Goodness of fit

In table 7, the association of boardroom diversity with volatility can be derived. The adjusted R-squared of the model is 21,1, meaning that 21,1% of the dependent variable is explained by the variables included in the model. A significant relationship can be observed between volatility and the independent variables with an F-statistic of 27,62 and p<0.001. This implies that one or more variables are significant for the outcome of volatility and the overall fit for the model is good (Field et al., 2013).

Sign of β’s

Independent variables

We can observe a coefficient of 0,246 for masculinity. Similar to model 2 and 3, this is in line with the predicted direction. A more masculine board composition can lead to more risk-taking behavior. The association between masculinity shows statistical significance with p<0,01. Therefore, hypothesis 1 for model 1 is not rejected. The uncertainty avoidance index results indicate that risk-taking behavior increases when the UAI score increases. This is in contrast with the hypothesis. Furthermore, there is no significant association between uncertainty avoidance index and volatility. The second hypothesis is rejected for the fourth model. Whereas the association between individualism and volatility is significant at the p<0,001, the results imply that an increase in the individualism score on a board leads to a decrease in risk-taking behavior. Which is not in line with the hypothesis. We can reject the third hypothesis.

(32)

Control variables

For the control variable information opacity, the coefficient is 8,718, which indicates that an increase in the information opacity increases the volatility. This is in line with the prediction and is significant on a 0,1% level. The association between the control variables, age, and volatility, is in line with the prediction. A board with a higher average age shows less risk-taking behavior. The coefficient of 0,038 indicates a negative relationship. For the gender, the coefficient is -2,442, which indicates that more women on a board lead to less risk-taking behavior. However, the association is not significant. Firm size’s coefficient -4,547 shows a negative relationship with risk-taking behavior. A result which is in line with the literature. Moreover, this relationship is significant at p<0,001 level. The relationship between the ratio executives to non-executives on a board and volatility has a coefficient of -0,475. This indicates that when there are more

executives, the risk-taking behavior declines. The literature predicts the contrary. The association shows no statistical significance with p>0,05.

5.2.5 Summary of results

H Independent variable Dependent variable Prediction Direction Result

1 MAS 1. R&D intensity + - Rejected

1 MAS 2. Altman Z-score - + Rejected

1 MAS 3. σ ROA + + Rejected

1 MAS 4. Volatility + + Not Rejected

2 UAI 1. R&D intensity - + Rejected

2 UAI 2. Altman Z-score + + Not Rejected

3 UAI 3. σ ROA - + Rejected

2 UAI 4. Volatility - + Rejected

3 IDV 1. R&D intensity + + Not Rejected

3 IDV 2. Altman Z-score - + Rejected

3 IDV 3. σ ROA + - Rejected

3 IDV 4. Volatility + - Rejected

Table 8: Summary of results

As can be seen in table 8 above, none of the hypotheses show consistent test results for the different models. Results for masculinity shows opposite directions for two of the four models. The beta for masculinity and σROA shows the right direction but is not significant. The

association between masculinity and volatility is positive, which is hypothesized and shows statistical significance with p<0,01. The results for UAI show that there is merely one hypothesis not rejected. Model two, with the Altman Z-score, shows statistical significance with P<0,001. A higher score on the uncertainty avoidance index leads to a higher Altman Z-score, which means that it is less likely that a firm will go bankrupt.

(33)

For the independent variable, individualism, the only hypothesis which is not rejected, is in the first model. The results show a significance of P<0,001 for the association between individualism and R&D intensity. The remaining three relationships show opposite directions than predicted by the literature.

5.3 Sensitivity test

To extend the primary analysis, I have conducted a sensitivity test. Herewith we can explore the hypotheses to a greater extent. The control variable firm size has most statistical significance in the original sample. For every model except the first. The sample is split at the average value of firm size. The off for firm size is at the average of £ 9,9 billion of total assets. Thus, the cut-off divides the bottom part in 1098 firms with total assets below average and 392 firms above average.

Large firms usually behave less risky because of their opportunities to spread

geographically and to diversify (Bruno & Shin, 2014; Nguyen, 2012; Faccio et al., 2014). These firms generally rely more on advanced and controlling management systems including better corporate governance mechanisms. Hence, we expect cultural values to have less influence on larger firms (Li et al. 2013). The results of the sensitivity test can be observed in table 9 below.

Table 9: Table significance: * = p<0.05, ** = p<0.01,*** = p<0.001 Regression results sensitivity test: Above, a sensitivity

checks showed for each hypothesis in each model. The sample is divided based on the control variable and cut off at the average of £9,9 billion of total assets in ‘low ‘and ‘high’ part. Subsequently, two new regressions are executed and the new coefficients are compared to the coefficient of the entire sample.

Hypothesis Independent variable Dependent variable Coefficient low firm size Coefficient high firm size Original result 1 MAS R&D -0,175 * 0,439 -0,214 ** 1 MAS Altman 0,008 0,080 0,025 1 MAS σROA 0,010 -0,074 0,042 1 MAS Volatility 0,362 *** -1,129 *** 0,246 ** 2 UAI R&D 5,869 3,297 8,839 2 UAI Altman 5,869 ** -0,701 6,019 *** 2 UAI σROA 6,833 7,842 8,645 * 2 UAI Volatility 16,06 ** -36,6 * 6,249 3 IDV R&D 25,69 *** 69,0 49,39 *** 3 IDV Altman 4,776 -0,366 5,245 * 3 IDV σROA -8,767 88,8 * -5,812 ** 3 IDV Volatility -49,01 *** -91,7 -75,37 ***

Referenties

GERELATEERDE DOCUMENTEN

Compared with the impacts of CEO inside debt to total ratio on risk-taking policies, I also find that CEO equity-linked to total ratio has a negative influence on firm

Therefore, the research question covered in this paper is as follows: Does a firm’s home country culture have a moderating effect on the relationship between board gender

However one must note that risk taking in firms is not only determined by individual factors or characteristics as organizational and national level factors

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

So there is found some evidence that board gender diversity will increase or decrease the performance of the firm, that internationalization has a positive effect on

In this study I find significant results for the uncertainty avoidance variable which implies that uncertainty avoidance affects the relationship between board size and

Surprisingly, the proportion of women on the board of directors and the percentage of firms that have at least one female board member is higher for one-tier boards than

Moreover, the moderating effect of country-level investor protection and bank-level corporate governance on the relationship between gender diversity and a bank’s risk taking is