The impact of a CEO’s cultural background on corporate risk taking

(1)

Amsterdam Business School

The impact of a CEO’s cultural background on

corporate risk taking

Name: Michiel van der Poel Student number: 10467998 Date: 17-01-2015

Final paper

Word count: 17402

MSc Accountancy & Control, specialization Control

Faculty of Economics and Business, University of Amsterdam First Supervisor: Ir. Drs. A.C.M. de Bakker

(2)

Statement of Originality

This document is written by student Michiel van der Poel 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)

Abstract

In this paper the impact of a CEO’s cultural background on corporate risk taking is researched. It was hypothesized that the cultural background of a CEO does matter, because a CEO has influence on corporate decision making and culture has influence on the personality and behavior of a CEO. The cultural background is quantified by using the score of the CEO’s nationality on Hofstede’s cultural dimensions: Power Distance, Uncertainty Avoidance, Individualism, Masculinity, Long Term Orientation and Indulgence. Risk taking is researched by using the Debt Ratio, Quick Ratio, Interest Coverage Ratio and Corporate Risk Policy. The sample consists of S&P 500 firms between 1998 and 2012. After controlling for various firm and CEO characteristics, no evidence is found that a CEO’s cultural background has impact on corporate risk taking. A clarification for the findings could be that CEOs from S&P 500 firms are not affected by environmental influences such as cultural background, but more from genetic influences such as the level of authority. Moreover the cultural dimension Indulgence was not part of the regressions due to multicollinearity problems.

Keywords: Chief Executive Officer, Corporate decision making, Cultural background, National culture, Risk taking, United States

(4)

1 Introduction

In a study of Mihet (2013) evidence is found that risk taking by foreign firms is best explained by the cultural norms of their country of origin. Is this phenomenon only visible by foreign firms or does the Chief Executive Officer’s (CEO’s) cultural norms of the country of origin also have impact on corporate risk taking? What we already know is that CEOs traits have impact on corporate decision making (e.g. Graham et al. (2013) and Bertrand and Schoar (2003)) and these managerial corporate decisions leads to a level of corporate risk, e.g. financial risk.

Above indicates a certain level of impact of a CEO on corporate risk taking. In this study, the impact of a CEO’s cultural background on corporate risk taking is investigated. The research question is:

What is the impact of a CEO’s cultural background on corporate risk taking?

Prior research has already studied the impact of cultural background on risk taking. For example the study of Mihet (2013), who studied the effects of culture on firm risk taking and analyzed this cross-country and cross-industry. Another study is done by Griffin et al. (2013) who investigated the influence of national culture on corporate risk taking. Characteristic to prior research is that all studies are cross country comparisons. Because a country’s culture and its institutional setting (e.g. law and regulation and economic environment) often go hand in hand, it is hard for cross country studies to distinguish between whether culture or the institutional setting determines the impact on corporate risk taking. Furthermore prior research studied the relationship between cultural background and risk taking on a firm level basis and not specific to the influences of a CEO.

The first contribution of this study is that the research of a CEO’s cultural background is executed in one single country (United States). This excludes all the institutional factors. One of the reasons that the study focuses on US CEOs, is that CEOs in the US have substantially greater effect on firm performance than CEOs in German and Japanese firms (Crossland and Hambrick 2005). This indicates also a greater effect on corporate risk taking. Besides US CEO data is easily accessible via public databases.

The second contribution is that the research is specified to a CEO’s cultural background instead of firm level comparisons. Recent studies stated that CEOs have a great influence on corporate decision making and thus corporate risk taking.

(7)

One of the most prominent psychological theories is used to determine cultural differences: Hofstede’s cultural dimensions (Hofstede, 1980 and 2010). The six cultural dimensions are: Power Distance, Uncertainty Avoidance, Individualism versus Collectivism, Masculinity versus Femininity, Long Term Orientation and Indulgence versus Restraint. The level of corporate risk taking will be determined by the following financial measures: Debt Ratio, Quick Ratio, Interest Coverage Ratio and Corporate Risk Policy.

The plan of the paper is as follows. The next chapter reviews the related literature of the impact of a CEO’s culture on corporate risk taking. It explains the following topics: the dependent variable ‘corporate risk taking’, the influence of a CEO on corporate decision making, the influence of cultural background on personal behavior, the six cultural dimensions of Hofstede (1980, 2010) and prior research which relates culture with corporate risk taking. Next, in chapter 3, the hypotheses development is explained and in chapter 4 an overview is given of the methodology. The data analysis in chapter 5 contains a sample selection procedure, the descriptive statistics and a test on multicollinearity. The results are presented in chapter 6 and finally the concluding paragraphs are presented in chapter 7.

(8)

2 Literature review

The literature review is build up as presented in figure 1. In paragraph 2.1 the definition of corporate risk taking is explained. Next, the influence of a CEO on corporate decision making is discussed. If a CEO does not have influence on corporate decision making, then it is logical that the CEO does not have influence on corporate risk taking. When noted that a CEO has influence on corporate risk taking, then paragraph 2.3 reviews if culture has influence on (CEO) personality and behavior. The six cultural dimensions of Hofstede will be explained in paragraph 2.4 and in paragraph 2.5 are similar studies demonstrated within a different research setting.

Figure 1: Theoretical framework

2.1 Definition of corporate risk taking

It is not easy to give one definition of risk, as several types of risk exist (e.g. corporate risks and personal risks). In the classical decision theory, risk is most commonly conceived as reflecting variation in the distribution of possible outcomes, their likelihoods and their subjective values (March and Shapira, 1987). Risk is embedded in the larger idea of choice as affected by the expected return of an alternative. Mostly risk is seen as a possible loss and not as a possible gain. One of the reasons to take a certain level of risk is to establish a competitive advantage. Ross (2014) found that a risk neutral firm should be risk seeking in developing a competitive

(9)

advantage when competing with other firms. The reason is that the payoff from favorable realizations is more than offsets in the cost of unfavorable realizations. Compensation maximization is another reason to take risk (Coles et al., 2006). This will be explained further in paragraph 4.2.3, where the control variables are defined.

In the risk appetite and tolerance guidance paper of Anderson (2011), it is stated that both risk appetite and risk tolerance are inextricably linked to firm performance. The risk universe (figure 3) is all the risk that the organization might face. Risk tolerance (figure 4) is about what you can allow the organization to deal with, while risk appetite (figure 5) is about the pursuit of risk.

The risk appetite is about how much risk the organization does want to take and how this can be guaranteed. Therefore it is the management’s responsibility to define the risk appetite and take care that the exercise of risk management throughout the organization is consistent with the risk appetite. Different boards, in different circumstances, will take different views on the relative importance of appetite and tolerance (Anderson, 2011).

Diagram 2

Figure 4: Risk Tolerance Figure 5: Risk appetite Figure 2: Risk Scope

(10)

When the organizational risk appetite and risk tolerance are known, the management has to decide how to respond to the identified risks. Management can respond to risk in four ways (Romney and Steinbart, 2012). It can reduce the likelihood and impact of risk by implementing an effective system of internal controls. Another way is to accept the likelihood and impact of the risk. Next, sharing the risk or transfer it to someone else is also an alternative. This can be done by buying insurance, outsourcing an activity, or entering into hedging transactions. At last someone can avoid the risk by not engaging in the activity that produces the risk. This may require the company to sell a division, exit a product line, or not expand as anticipated.

The risk appetite, risk tolerance and the level of risk taking is influenced by several factors. In this study the impact of a CEO’s cultural background on corporate risk taking is researched. The influence of firm characteristics (firm size and firm industry) and CEO characteristics (age, gender, tenure and compensation) are explained in paragraph 4.2.3. Besides these influences the following part of this paragraph focuses on the influence of institutional factors on corporate risk taking. Hitt et al. (2004) stated that institutions, which include capital markets, law, rules and governance mechanisms in a specific national setting, determine the boundaries of acceptable strategic actions available for organizations. Institutional arrangements have the ability to legitimize or constrain the actions of firms. Dickson et al. (2010) studied the influence of institutional variables on organizational risk taking and proactive firm behaviors. All seven researched institutional variables had a significant impact on the level of risk taking and proactiveness between at least two countries. Three variables were linked to Gross Domestic Product (GDP). The other four were: technology sophistication, economic risk, political risk and legal system. Mihet (2013), who studied the relationship between culture and corporate risk taking in a cross country setting, controlled for institutional factors as well (e.g. religion, corruption and government effectiveness).

Another example that institutional factors influence corporate risk taking is the Sarbanes-Oxley Act (SOX) which was signed into law in 2002 for publicly traded US companies. Bargeron et al. (2010) investigated the influence of SOX on corporate risk taking. They found that several measures of risk-taking decline significantly for US versus non-US firms after SOX. This decline is visible in investment decisions and stock price volatility in US firms, compared with UK and Canadian firms after SOX implementation.

There are several ways to measure the level of risk taking, for example via questions about risky choices in questionnaires. One of the questions that could be asked is about the propensity of risk taking in a casino. Following MacCrimmon and Wehrung (1990), this approach utilizes hypothetical choices, so the understanding and motivation underlying the responses is

(11)

sometimes questionable. Therefore financial measures seem to be the most obvious measure for this study. Prior studies also used financial measures to measure the level of corporate risk taking, e.g. Graham et al. (2010), Griffin et al. (2013) and Mihet (2013). These measures will be explained in more detail in paragraph 4.2.1.

2.2 The influence of a CEO on corporate decision making

Important for this study is whether a CEO has influence on corporate decision making. If this is not the case, then the cultural background of a CEO does not seem to have a great impact on corporate risk taking. Recent studies suggest that CEOs have impact on corporate decision making. For example a recent study of Graham et al. (2013). They studied the behavioral traits of senior executives and also harvested information related to career path, education and demographics. In the research evidence is found that psychological traits such as risk-aversion and optimism links to corporate policies. This influences CEO decisions. For example more risk-tolerant CEOs make more acquisitions and more optimistic CEOs use more short-term debt.

In addition Graham et al. (2013) found evidence for differences in corporate decision making between CEOs and CFOs and among US-based and non-US CEOs. CEOs tend to be more optimistic and risk tolerant than others. Besides US-based CEOs tend to be more optimistic and less risk-averse among non-US CEOs. These differences suggest that US firms behave different in comparison with non-US firms.

Another study that provides evidence that individual managers affect corporate behavior and performance is done by Bertrand and Schoar (2003). They have found that manager’s fixed effects matter for a wide range of corporate decisions. A significant diversity in investments, financial and organizational practices of firms can be explained by the influence of managers. The researchers tracked the individual top managers across different firms over time. This way of research excludes the firm’s fixed effects and time-varying firm characteristics.

2.3 Definition of culture and the influence on personal behavior

Hofstede (2011) defined culture as ‘the collective programming of the mind that distinguishes the members of one group or category of people from others’. It is always a collective phenomenon, but it can be connected to different collectives. Societal, national and gender cultures are much deeper rooted in the human mind than occupational cultures acquired at school or than organizational cultures acquired on the job. These societal, national and gender

(12)

cultures are acquired from children’s earliest youth onwards. Elements of culture are shared standard operating procedures, unstated assumptions, tools, norms, values, habits about sampling the environment, and the like (Triandis and Suh., 2002). These elements of culture can be influenced in different ways, for example by ecology (terrain, flora and fauna) and climate.

Important for this study is whether culture has influence on personal behavior and personality, specific to CEOs. Personality is shaped by genetic and environmental influences. Cultural influences are the most important part of the environmental influences (Triandis and Suh, 2002). In their study, Triandis and Suh (2002) found evidence that ecologies shape cultures and cultures influence the development of personalities. The variation of personality is caused by universal and culture-specific aspects. Some of these culture-specific aspects link to cultural syndromes such as complexity, tightness, individualism and collectivism. The last two are used in the Hofstede model, and will be explained in the next paragraph. Besides, cultural values underlie people’s attitudes and behavior (Sengupta and Sinha, 2005). The cultural values shape individual perceptions and expectations about work and the way of interaction with their colleagues. Furthermore, cultural values provide guidelines for decision making.

Sengupta and Sinha (2005) researched the perceived dimensions of societal and organizational cultures and their impact on managerial work behavior. The findings indicated that the spillover of societal culture on organizational culture influenced the work behavior of managers. Besides Offermann and Hellmann (1997) researched the culture’s consequences for leadership behavior. They researched the relationships between work-related values held by managers and their leadership practices as assessed by subordinates. The four dimensions of Hofstede are used in this study. The extracted data provide evidence that managerial cultural background can relate to what subordinates see in their manager’s leadership behaviors and it provides empirical support for the theoretical work of Hofstede.

In summary, in prior research is found that CEO behavior is influenced by their cultural background. The six dimensions of Hofstede will be explained in the next paragraph.

2.4 Six cultural dimensions of Hofstede

Hofstede’s cultural dimensions (1980, 2010) are a widely used theory to relate culture with nationality. A dimension is an aspect of a culture that can be measured relative to other cultures (Hofstede, 2011). In the 1970s Geert Hofstede got access to a large survey database about values and related sentiments of people in over 50 countries around the world (Hofstede, 1980). The database contained about more than 100,000 questionnaires. All the people worked for the

(13)

multinational corporation IBM. Four dimensions of national culture were initially found. A fifth and a sixth dimension were added during later research in cooperation with other researchers (Hofstede et al., 2010). The six dimensions are: Power Distance (PD), Uncertainty Avoidance (UA), Individualism versus Collectivism (IC), Masculinity versus Femininity (MF), Long Term versus Short Term Orientation (LSO) and Indulgence versus Restraint (IR). In the next sub paragraph, the six cultural dimensions will be described based on several studies of Hofstede.

Power Distance

PD is the extent to which the less powerful members of organizations and institutions accept and expect that power is distributed unequally (Hofstede, 2011). More or less inequality is defined from the people below. Hofstede (2011) stated that all societies are unequal, but some are more unequal than others. In high PD societies it is an emphasis to maintain your current status in the social order (Hofstede, 1980). Besides individuals in low PD cultures are more intent on bettering their position and there is a higher degree of social mobility. Another example is that subordinates in a low PD society are expected to be consulted, while subordinates in high PD societies are told what to do (Hofstede 2011). According to Hofstede et al. (2010), the PD scores tend to be higher for East European, Latin, Asian and African countries and lower for Germanic and English-speaking Western countries.

Uncertainty Avoidance

UA indicates to what extent a culture programs its members to feel either uncomfortable or comfortable in unstructured situations (Hofstede, 2011). Unstructured situations can be defined as unknown, novel, surprising and different from usual. Societies with a high level of UA try to minimize these situations by e.g. strict behavioral codes and laws and rules. This indicates that societies with a high level of UA need clarity and structure, while societies with lower UA are comfortable with ambiguity and chaos. According to Hofstede et al. (2010), UA tend to be higher in East and Central European countries, in Latin countries, in Japan and in German speaking countries, lower in English speaking, Nordic and Chinese culture countries.

Individualism versus Collectivism

Individualism versus Collectivism is the degree to which people in a society are integrated into groups (Hofstede, 2011). Cultures which focus more on individualism are more loose:

(14)

everyone is expected to look after him or herself and his or her family. On the opposite side there are cultures with a high score on collectivism, where people from birth onwards are integrated into strong, cohesive in-groups, often extended families that continue protecting them in exchange for unquestioning loyalty. In a society where individualism dominates, is speaking one’s mind healthy. On the other hand in a society that scores high on collectivism, it is important to maintain harmony. In Hofstede et al. (2010) is stated that individualism is more found in developed and Western countries, while collectivism prevails in less developed and Eastern countries.

Masculinity versus Femininity

Masculinity versus the opposite Femininity refers to the distribution of values between genders which is a fundamental issue for any society (Hofstede, 2011). The IBM studies of Hofstede revealed that women’s values differ less among societies than men’s values. Men’s values vary from very assertive and competitive on the one side, to modest and caring, similar to women’s, on the other hand. The assertive and competitive side is called ‘Masculine’ and the modest and caring side is called ‘Feminine’. There is sympathy for the weak in societies where Femininity prevailed, while masculine cultures have admiration for the strong. Another comparison is that feminine societies have many women in elected political positions and masculine societies have few women in these positions. According to Hofstede et al. (2010) Masculinity is high in Japan, German speaking countries and in some Latin Countries like Italy and Mexico. Masculinity is low in Nordic countries and in The Netherlands.

Long Term versus Short Term Orientation

As the four initial dimensions of Hofstede did not pay attention to economic growth, after years of research Hofstede et al. (2010) succeeded to create a new dimension available for 93 countries. The dimension is called Long Term Orientation (LTO) versus Short Term Orientation (STO). The dimension can be seen as the society’s time horizon, as the LTO take account of future stability. Societies which are more LTO pay more attention to perseverance, thrift and ordering relationships by status (Hofstede, 2011). STO societies have more respect for tradition and pays attention to personal steadiness and stability. Another difference is that a high score on LTO indicates a large savings quote and the necessity that funds are available for future investments, while a STO society finds social spending and consumption more important. According to Hofstede et al. (2010), LTO countries are East Asian countries followed by

(15)

Eastern- and Central Europe. More short- term oriented are the US and Australia and Latin American, African and Muslim countries.

Indulgence versus Restraint

The sixth dimension of Hofstede is Indulgence versus Restraint. This dimension focuses on aspects which are not covered by the other five dimensions (Hofstede, 2011). Indulgence stands for a society that allows a relatively free gratification of basic and natural human desires related to enjoying life and having fun. Restraint stands for a society that controls gratification of needs and regulates it by means of strict social norms. For more indulgence societies, freedom of speech is seen as important, while freedom of speech is not a primary concern in societies where Restraint is prevailed. Equal to the first five dimensions, differences between societies exist. Following Hofstede et al. (2010), indulgence prevails in South and North America, in Western Europe and in parts of Sub-Sahara Africa. Restraint prevails in Eastern Europe, Asia and in the Muslim world.

2.5 Cultural influences on corporate risk taking

As stated earlier, prior research has already investigated the impact of cultural background on corporate risk taking. These studies focused on cross country differences and firm and industry level comparisons. In this paragraph a brief overview is given of the research methods they used and the results they have found.

Griffin et al. (2013) investigated the role of national culture on corporate risk taking. To determine culture they used two values developed by Hofstede (1980); Individualism (versus Collectivism) and Uncertainty Avoidance. The third value they used is Harmony (versus Mastery) developed by Schwartz (1994, 2004). Cultures that scores high on harmony emphasize accepting matters as they are, whereas cultures low on harmony emphasize the importance of assertiveness to advance personal and group interests. They researched 35 countries in the period between 1997 and 2006 and found evidence that there is a positive association between individualism and risk taking, a negative association between Uncertainty Avoidance and risk taking, and a negative association between harmony and risk taking. Furthermore, they demonstrated that culture influences corporate risk taking directly on risky corporate decision making and indirectly through formal institutional factors, which in turn influences risky corporate decision making.

(16)

Institutional factors which are taken into account in this paper are: financial systems and the level of shareholder and creditor protection.

Also Mihet (2013) discovered the impact of culture on corporate risk taking in a direct and indirect way. In this study an international sample is used of 50,000 firms spread across 400 industries and 51 countries. Mihet (2013) used four dimensions of Hofstede (1980) as measure for culture. Evidence is found that corporate risk taking is higher in societies with low Uncertainty Avoidance, low tolerance for hierarchical relationships and in societies which value Individualism over Collectivism. Furthermore evidence is found that countries that scores high on Uncertainty Avoidance and low on Individualism take significantly less risk in industries which are more informationally opaque (e.g. finance, IT, oil refinery and mining). These industries have a higher information uncertainty because of the greater complexity of the operations and the difficulty of assessing and managing risk.

Another study is undertaken by Dickson et al. (2010), who researched the impact of cultural values and institutions on two key dimensions of entrepreneurship: risk taking and proactiveness. The responding firms were independent small- to medium-sized enterprises (SMEs) from six different countries. The researchers used four dimensions of Hofstede (1980) as measure for culture. Uncertainty Avoidance and Power Distance were both found to have a significant and negative influence of risk taking levels within SMEs. A number of institutional factors were also found to significantly impact a firm’s risk taking behavior, such as national economy, national technological policy, the nature of economic and political risk.

(17)

3 Conceptual framework and hypotheses development

3.1 Conceptual framework

The figure below (figure 6) shows the conceptual framework, including the proposed relations. The proposed relations will be explained in more detail in paragraph 3.2.

3.2 Hypotheses development

The six research propositions given in the conceptual framework in figure 6 will be substantiate in this paragraph. All substantiations are supported by prior literature.

Hypothesis 1: Power Distance

Shane (1993) found evidence that managers with a low PD are more innovative and showing more risky behavior. These managers have the aim to improve their current position in the industry by enacting in risky offensive strategies (Dickson et al., 2010).

In the cross country and cross industry study of Mihet (2013) evidence is found that PD is negatively related to firm risk taking. Besides the study of Dickson et al. (2010) found similar results for the small and medium sized companies that PD is negatively related to risk taking. Thus it is expected that the level of CEO Power Distance will be negatively associated with the level of corporate risk taking.

H1: The level of CEO Power Distance will be negatively associated with corporate risk taking

(18)

Hypothesis 2: Uncertainty Avoidance

Ambiguous situations are unknown, novel, surprising and different from usual. Societies with a high level of UA try to minimize these situations. However societies with a low level of UA are comfortable with ambiguity and chaos and do not need clarity and structure. These societies encourage managers to develop a greater willingness to take risks (Hofstede, 1980). As a result of the avoidance of uncertain situations by managers in a high UA society, the level of UA is negatively related to the level of venture capitalist activity (Li and Zara, 2012). The provision of financial capital to early-stage, high-potential startup companies is very uncertain and thus risky.

In the studies discussed in paragraph 2.5 (Dickson et al. (2010), Griffin et al. (2013) and Mihet (2013)) evidence is found that there is a negative relationship between Uncertainty Avoidance and the level of risk taking. Even these studies were undertaken in a different research setting, the same outcomes are expected. Thus it is expected that CEO uncertainty avoidance is negatively associated with the level of corporate risk taking.

H2: The level of CEO Uncertainty Avoidance will be negatively associated with corporate risk taking

Hypothesis 3: Individualism

In cultures which focus more on Individualism, everyone is expected to look after him or herself without the existence of any social control mechanism. In these cultures freedom and autonomy are very important. Managers in cultures high on Individualism over collectivism have a tendency to place a higher value on individual accomplishments than collectivist managers (Hofstede, 1980). This leads to a higher level of risk taking due to hope of larger strategic payoffs as a result of their own effort and leadership (Morris et al., 1993).

In line with previous, Pan and Statman (2009) found that overconfident individuals are more risk tolerant than less overconfident people. This is caused by their ability to control outcomes and overestimating their knowledge.

The studies of Mihet 2013 and Griffin et al. 2013 found significant evidence for the positive relationship between Individualism and corporate risk taking. In the study of Dickson et al. (2010), where this relationship is studied for small and medium sized companies, evidence is found that individualistic and collectivistic cultures have similar orientations toward risk. Although Dickson et al. (2010) did not find a significant relationship, it is expected that CEO Individualism is positively associated with the level of corporate risk taking.

(19)

H3: The level of CEO Individualism will be positively associated with corporate risk taking.

Hypothesis 4: Masculinity

Managers in masculine cultures score highly on McClelland’s (1960) need for achievement (Hofstede, 1980). Individuals with a high need for achievement tend to be ambitious and more willing to engage in risk taking than other managers. Hofstede (1980) also found evidence that managers in feminine cultures more carefully thought out their decisions, while managers in masculine cultures value decisive and immediate actions.

Managers in masculine cultures tend to be more self-confident than managers in feminine cultures (Hofstede, 1980). More confident people are likely to take more (excessive) risks than people who are less confident (Barber and Odean, 2001). Managers in masculine countries have also been shown to place a higher emphasis on showing off (“machismo”), which indicates higher potential risks.

In the studies of Dickson et al. (2010) and Mihet (2013) no significant relationship is found between the Hofstede dimension Masculinity and the level of corporate risk taking. Also in the study of Griffin et al. (2013) where the dimension is used as control variable, no significant relationship is found. Because this study focuses on CEOs specific instead of cross country or industry comparisons and prior research indicates a relationship, it is expected that CEO Masculinity is positively associated with the level of corporate risk taking.

H4: The level of CEO Masculinity will be positively associated with corporate risk taking

Hypothesis 5: Long Term Orientation

Hofstede (2011) stated that more LTO societies have a large savings quote and the necessity that funds are available for future investments. It is unlikely that managers take any risk that lead to uncertainty and the possibility for unavailability of their funds for future investments. CEOs are more overconfident when they lead firms headquartered in countries by a low level of Long Term Orientation (Ferris et al., 2013). Short Term Oriented cultures are capable for more rapid change and the long-term traditions are less of a barrier to innovation. As previously mentioned, overconfident CEOs take more risk than other people. Therefore it is

(20)

expected that societies with a STO take more risk. This results in a negative association between CEO Long Term Orientation and the level of corporate risk taking.

H5: The level of CEO Long Term Orientation will be negatively associated with corporate risk taking

Hypothesis 6: Indulgence

Non-risk cultures presume a determinate, institutional, rule bound and necessarily hierarchical ordering of individual members in regard to their utilitarian interests (Adam et al., 2000). Risk cultures, in contrast, presume not a determinate ordering, but a reflexive disordering. Besides risk cultures lie in non-institutional and anti-institutional societies. Their governing figurations are not rules based, but have a more symbolic character. Restraint stands for a society that controls gratification of needs and regulates it as non-risk cultures do. Indulgence stands for a society that allows a relatively free gratification and is anti-institutional. Thus it is expected that CEO Indulgence is positively associated with the level of corporate risk taking.

(21)

4 Methodology

4.1 Sample

A quantitative (public) database research will be used to answer the research question. The research is focused on US CEOs who have functioned as CEO for a firm listed on the S&P500 in the period of 1998 until 2012. It is not necessary to control for the impact of institutional and cross-country factors on firms, because only US based firms are included in the sample. The CEO nationality needs to be determined in order to link the cultural dimensions of Hofstede (1980, 2010) to a CEO. The nationalities are derived from the public database BoardEx1_{. In the} next paragraph the financial risk measures and control variables will be explained.

4.2 Variables

To test if a CEO’s cultural background has impact on corporate risk taking, several CEO data and financial data is collected. The variables used to research the level of corporate risk taking are explained in paragraph 4.2.1, the measures for cultural background are explained in paragraph 4.2.2 and finally the control variables will be presented in paragraph 4.2.3.

4.2.1 Corporate risk taking

Corporate risk taking can occur in different ways and is dependent of multiple variables. In this study is chosen to use four financial variables to measure the level of corporate risk taking. Every variable stands for one risk category. The categories financial risk, liquidity risk, solvency risk and risky corporate policies are used. The way of categorizing risk measures is similar to the study of Claessens (2000), where no less than fifteen firm-specific variables to asses a firm’s riskiness are divided into seven different groups. The financial measures will be derived from the Compustat database, in which financial data of S&P 500 firms is collected over years.

The first category is financial risk. Graham et al. (2010) stated that when a firm uses more debt, the firm produces more risk and higher expected returns. Debt is risky when the firm is not able to repay the amount of debt. According to Claessens et al. (2000), high financial leverage

1_{. Dr. M. van der Poel, who was Assistant Professor at the Finance Department of Rotterdam School of}

(22)

and associated large interest payments will reduce the ability of a firm to deal with financial shocks. The traits of managers can influence the Debt Ratio, as optimistic and/or overconfident managers choose higher debt levels and issue new debt more often (Hackbarth, 2008). This is caused by growth and/or risk perception biases. In the studies Claessens et al. (2000), Hackbarth (2008) and MacCrimmon and Wehrung (1990), the Debt Ratio is used as a risk measure.

Another risk which firms faces is running out of liquidity. The Quick Ratio will be used to capture the corporation’s ability to turn assets and earnings into liquidity quickly (Claessens et al., 2000). The Quick Ratio is defined as the ratio of current assets net of inventory (cash, other working capital and trade receivables) to current liabilities (short-term debt and trades payables). The Quick Ratio is preferred above the current ratio, because it compares only the most liquid short-term assets to all short-term liabilities.

Claessens et al. (2000) used the Interest Coverage Ratio as solvency measure. The Interest Coverage Ratio measures credit risk. The higher cash flows are relative to interest payments for debt, the less likely it is that the firm is not able to pay for debt services. The Interest Coverage Ratio can be computed by dividing the EBIT by interest expenses.

The latest measure is R&D investments, which is commonly used as a measure of risky corporate policies (e.g. Griffin et al., 2013 and Coles et al. (2006)). R&D investments are risky because they have a low probability of success and the benefits are uncertain. The Corporate Risk Policy will be computed as follows: R&D expenses divided by total assets.

4.2.2 Cultural background

As stated earlier, the six dimensions of Hofstede (1980, 2010) are used to research the cultural background of a CEO. These dimensions are explained in paragraph 2.4. The country scores are relative (societies are compared to other societies). These relative scores have been proven to be quite stable over decades. Appendix I gives an overview of the scores per dimension as researched by Hofstede.

4.2.3 Control variables

Based on previous research findings it is expected that certain firm and CEO characteristics have influence on corporate risk taking. In order to control for those effects, the control variables are described in this paragraph.

(23)

Firm characteristics

Griffin et al. (2013) found evidence that larger firm size weakens the effect of culture on corporate risk taking. They think that this effect is caused by highly controlled management systems, including better corporate governance practices. Besides Hope (2003) shows that larger firms are associated with higher quality disclosure, mainly due to higher agency costs. The measure to calculate Firm Size is the natural logarithm of sales as used in the studies of Hope (2003) and Coles et al. (2006).

Some industries are riskier than others. For example commodity industries (biofuels, metal products and non-ferrous metals) are riskier and more opaque than other industries. Production methods are more complicated and R&D expenditures are higher (Mihet 2013). Where Griffin et al. (2013) and Kanagaretnam et al. (2014) researched specific industries to control for industrial differences, Mihet (2013) performed a cross industrial and cross country analysis to take the industrial differences into account. As mentioned earlier, in this study evidence is found that countries that scores high on Uncertainty Avoidance and low on Individualism take significantly less risk in industries which are more informationally opaque (e.g. finance, IT, oil refinery and mining). The SIC code will be used to distinguish industries in this study. Because of the wide range of SIC codes, the industry classification of Fama & French is used. Fama & French established 12 industry portfolios based on SIC codes, which are presented in table 1. Dummy variables will be used to distinguish between industries. The portfolio data is taken from the Data Library of Kenneth French. The SIC codes can be found in appendix II and will be derived from the Compustat database.

# Description Content

1. Consumer non durables Food, tobacco, textiles, apparel, leather, toys 2. Consumer durables Cars, TV’s, furniture, household appliances

3. Manufacturing Machinery, trucks, planes, furniture, paper, printing

4. Energy Oil, gas and coal extraction and products

5. Chemicals Chemicals and allied products

6. Business equipment Computers, software, electronic equipment

7. Telecom Telephone and television transmission

8. Utilities Utilities

9. Shops Wholesale, retail and some services (laundries, repair shops)

10. Health Healthcare, medical equipment and drugs

11. Money Finance

12. Other Mines, construction, transfers, hotels, bus services, entertainment

(24)

CEO characteristics

The existing literature suggests that males tend to be more overconfident than females, for example Barber and Odean (2001). Overconfident people are likely to take more (excessive) risks than people who are less confident. In the study of Graham et al. (2013) they used CEO Gender as control variable. To distinguish between male and female, a dummy variable is created (0 for female and 1 for male). The CEO Gender is derived from the Compustat database.

Serfling (2014) studied the relationship between CEO Age and the riskiness of corporate policies. Serfling found that older CEOs reduce firm risk through less risky investment policies. Especially older CEOs invest less in R&D, make more diversifying acquisitions, manage firms with more diversified operations and maintain a lower leverage.

Some US-based studies demonstrates that equity based pay encourages managerial risk taking. Coles et al. (2006) studied the relationship between managerial incentives and risk taking. They investigated the CEO wealth to stock return volatility (vega). They found evidence that a higher sensitivity of CEO wealth to stock volatility implements riskier policy choices. For example more investment in R&D and less investment in PPE. Another study that confirms these findings is done by Graham et al. (2010). They found that risk-averse CEOs are significantly more likely to be compensated by salary and less likely to be compensated with performance-related packages. The CEO Compensation is measured as the salary (standard pay) divided by total pay.

Prior research suggests that the CEO Tenure is affecting the level of risk taking. CEO tenure is the number of years that the CEO has been in current position. In the paper of Berger et al. (1997) is stated that when a CEO has a long tenure in the office, that the leverage is lower. This means that the CEO is more risk-averse as leverage is one of the risk measures. Tenure will be measured as the number of years in current (CEO) position.

4.2.4 Summary regression variables

Table 2 presents a summary of all the regression variables as explained in prior paragraphs. The cultural background is called Panel A, Corporate risk taking is called Panel B, Firm characteristics are called Panel C and CEO characteristics are called Panel D. Besides the formula and data source are given.

(25)

Panel Formula Source Panel A: Cultural background

Power Distance Relative scores per country Hofstede (1980)

Uncertainty Avoidance Relative scores per country Hofstede (1980)

Individualism Relative scores per country Hofstede (1980)

Masculinity Relative scores per country Hofstede (1980)

Long Term Orientation Relative scores per country Hofstede (2010)

Indulgence Relative scores per country Hofstede (2010)

Panel B: Corporate risk taking

Debt Ratio Total Liabilities / Total Assets Compustat

Quick Ratio (Current Assets - Inventory) / Current

Liabilities Compustat

Interest Coverage Ratio EBIT / Interest Expenses Compustat

Corporate Risk Policy R&D Expenses / Total Assets Compustat

Panel C: Firm characteristics

Firm Size Ln (sales) Compustat

Firm Industry Dummy variables (#11) Compustat

Panel D: CEO characteristics

CEO Gender Dummy variable Execucomp

CEO Age Age Execucomp

CEO Compensation Salary (standard pay) / Total compensation Execucomp

CEO Tenure Number of years in current (CEO) position Execucomp

4.3 Model

To estimate the impact of a CEO’s cultural background on corporate risk taking, the linear model below is used. The model represents the hypotheses as discussed in chapter 3.

𝒀_{𝟏,𝟐,𝟑,𝟒 𝒊,𝒋,𝒕} = 𝛽0 + 𝛽1 (𝑃𝑜𝑤𝑒𝑟 𝐷𝑖𝑠𝑡𝑎𝑛𝑐𝑒)j + 𝛽2 (𝑈𝑛𝑐𝑒𝑟𝑡𝑎𝑖𝑛𝑡𝑦 𝐴𝑣𝑜𝑖𝑑𝑎𝑛𝑐𝑒)j + 𝛽3 (𝐼𝑛𝑑𝑖𝑣𝑖𝑑𝑢𝑎𝑙𝑖𝑠𝑚)j + 𝛽4 (𝑀𝑎𝑠𝑐𝑢𝑙𝑖𝑛𝑖𝑡𝑦)j + 𝛽5 (𝐿𝑜𝑛𝑔 𝑇𝑒𝑟𝑚 𝑂𝑟𝑖𝑒𝑛𝑡𝑎𝑡𝑖𝑜𝑛)j + 𝛽6 (𝐼𝑛𝑑𝑢𝑙𝑔𝑒𝑛𝑐𝑒)j + 𝛽7 (𝐹𝑖𝑟𝑚 𝑆𝑖𝑧𝑒)i,t + ∑ 𝛽7+𝑘 11 𝑘=1 (𝐷𝑢𝑚𝑚𝑦 𝐹𝑖𝑟𝑚 𝐼𝑛𝑑𝑢𝑠𝑡𝑟𝑦)i + 𝛽19 (𝐶𝐸𝑂 𝐺𝑒𝑛𝑑𝑒𝑟)j + 𝛽20 (𝐶𝐸𝑂 𝐴𝑔𝑒)j,t + 𝛽21 (𝐶𝐸𝑂 𝐶𝑜𝑚𝑝𝑒𝑛𝑠𝑎𝑡𝑖𝑜𝑛)j,t + 𝛽22 (𝐶𝐸𝑂 𝑇𝑒𝑛𝑢𝑟𝑒)j,t + 𝜀𝑖,𝑗,𝑡

The character ‘i’ represents the firm, ‘j’ represents the CEO and ‘t’ represents the fiscal year. 𝑌1

to 𝑌4 represent the dependent variables Debt Ratio, Quick Ratio, Interest Coverage Ratio and

Corporate Risk Policy respectively. 𝛽1 to 𝛽6 estimate the effect of the six cultural background

dimensions on corporate risk taking of a CEO. 𝛽7 to 𝛽22 estimate the effect of the control

variables explained in paragraph 4.2.3. The epsilon (𝜀) represents the random error component.

(26)

4.4 Regression method

The regression method ‘panel data’ is applied to run the four regressions. Panel data can be used when the data includes observations of the same firms over many years. As the sample consists of S&P 500 firms over a period of fifteen years (1998-2012), this method is considered as the best for this research. Several panel data techniques exist. In this research is chosen for fixed effects regressions, because these are common in finance related studies. According to Torres-Reyna’s (2007) manual for STATA, fixed effects is only interesting when analyzing the impact of variables that vary over time. When using fixed effects, it is assumed that something within the individual firm may impact or bias the dependent variable. Another assumption is that the time-invariant characteristics are unique to the individual firm and is not correlated with other individual characteristics. To control for the possible existence of heteroscedasticity, clustered standard errors are applied in all regressions. Heteroscedasticity exist when random variables in the model includes sub-populations that have a different variability. Autocorrelation or serial correlation can give statistical problems in panels with long time series. Autocorrelation means a relationship between a given variable and itself over various time intervals. It is assumed that the use of panel data regressions excludes this problem.

(27)

5 Data analysis

5.1 Sample selection procedure

As discussed in paragraph 4.1, the sample consists of CEOs who have functioned as CEO for a firm listed on the S&P500 in the period of 1998 until 2012. Unfortunately not all observations and data necessary to run the statistics are available in public databases. Table 3 illustrates the sample selection procedure which is executed to remove erroneous and missing sample data. The note under table 3 gives a full explanation of the selection procedure.

CEO-Firm-Years obs Description Lost obs

7,500 (1) Expected number of observations -

6,159 (2) Observations provided in nationality database 1,341

6,143 (3) CEOs not available in BoardEx database 16

5,829 (4) Nationalities not available in BoardEx database 314

5,822 (5) Nationalities not in Hofstede (2010) model 7

5,795 (6) Incomplete scores in Hofstede (2010) model 27

5,590 (7) Influence of a CEO on corporate risk taking (< 4 month) 205

5,409 (8) CEO Compensation variable is missing 181

5,409 2,091

Table 3: Sample selection procedure

Note: (1) The starting point is 15 years of observation and 500 firms. Thus the expected CEO-Firm-Year

observations are 7,500. (2) The provided database with CEO nationalities consists of 6,159 CEO-Firm-Years. One of the reasons for the lost observations is that the data was extracted mid 2012, while the fiscal year of most companies was not ended yet. For 2012 only 25 CEO-Firm-Years are available, instead of the expected 500. (3) Besides not all CEO nationalities are available in the provided database. The provided CEO nationalities are retrieved from the BoardEx database, which is a database that contains data from publicly listed companies about their board members. (4) Not all nationalities are found in this database. About 543 CEO-Firm-Years are missing. Some nationalities are found via internet research, mainly through the Notable Names Database (NNDB). Finally 7 CEOs, which represents 16 CEO-Firm-Years, are not found in the BoardEx database and other resources. (5) Besides for 314 CEO-Firm-Years the CEO nationality is not available in the BoardEx database and not found in other resources.

Also in the independent and control variables are erroneous and missing data discovered. The Hofstede model (2010) consists of 110 nationalities. (5) For 3 CEOs, which represents 7 CEO-Firm-Years, the nationality is not included in the model. These are Lebanon, Kenyan and Netherlands Antillean nationalities.

(6) For 27 CEO-Firm-Years not all scores of the six dimensions has been researched by Hofstede. For

example Egypt does not have scores on Long Term Orientation and Indulgence. Only nationalities with all scores on the cultural dimensions are part of this study. (7) Next, some CEOs have a tenure of zero. This

(28)

means that the CEO is at this position less than one year. In this study it is assumed that a CEO has influence on corporate risk taking when he or she is at a CEO position for four month or more. Observations are removed when the executive was shorter than four month at a CEO position. As a consequence, 205 CEO-Firm-Years has been removed. (8) The latest step in the sample selection procedure is the removal of CEO-Firm-Years from which the CEO Compensation variable is missing. This is the reason why 181 additional CEO-Firm-Years have been removed.

The total data reduction consists of 2,091 CEO-Firm-Years. This is about 28% of the expected amount of observations (7,500). The remaining 5,409 observations will be used to run the statistical regressions.

5.2 Descriptive statistics

Table 4 presents the descriptive statistics for the six cultural dimensions of Hofstede (2010), panel A. Besides, the number of CEO-Firm-Years and CEO-Firm combinations are presented per country for the countries which are part of the sample. Not surprisingly, most observations are CEOs from the United States. No less than 4,970 observations (91.9%) are from the United States and the remainder 439 observations (8.1%) are from other countries. Canada (85), Great Britain (80) and India (75) represent the most observations after the United States.

The scores on Hofstede’s cultural dimensions differ a lot from country to country. This is visible in the range of minimum and maximum scores. The mean scores lie close to the scores of the United States, as this country presents more than 90% of all observations. For the median, the scores are exactly the same as the United States. Prominent is the score on Uncertainty Avoidance of Greece (112) which is above 100. The initial scores of the research of Hofstede (1980) are in a range of 0-100. As a consequence that more countries were added in later years, the scores have been recalculated. Another prominent score is the score on Indulgence of Pakistan (0). This means that no doubt exist that the people in this society are very restricted in their lives.

(29)

Country

CEO-Firm-Year obs CEO-Firm obs PD UA IND MAS LTO IR

Arabic Countries 25 2 80 68 38 53 23 34 Australia 24 5 36 51 90 61 21 71 Belgium 12 2 65 94 75 54 82 57 Canada 85 26 39 48 80 52 36 68 China 3 2 80 30 20 66 87 24 Croatia 2 1 73 80 33 40 58 33 France 12 4 68 86 71 43 63 48 Great Britain 80 19 35 35 89 66 51 69 Germany 5 1 35 65 67 66 83 40 Greece 1 1 60 112 35 57 45 50 India 75 13 77 40 48 56 51 26 Iran 5 1 58 59 41 43 14 40 Ireland 19 4 28 35 70 68 24 65 Malta 1 1 56 96 59 47 47 66 Morocco 10 1 70 68 46 53 14 25 Netherlands 26 4 38 53 80 14 67 68 Pakistan 10 3 55 70 14 50 50 0 Singapore 6 1 74 8 20 48 72 46 Spain 9 3 57 86 51 42 48 44 Taiwan 26 2 58 69 17 45 93 49 Turkey 3 1 66 85 37 45 46 49 United States 4,970 986 40 46 91 62 26 68 Grand Total 5,409 1,083

Mean Score (weighted) 40.96 46.40 88.84 61.30 27.77 66.75

Median Score 40 46 91 62 26 68

Standard Deviation 6.09 5.08 9.90 4.22 8.16 6.92

Min Score 28 8 14 14 14 0

Max Score 80 112 91 68 93 71

Table 4: Descriptive statistics panel A, inclusive count of CEO-Firm-Year and CEO-Firm observations per country

Table 5 represents the descriptive statistics of the dependent variable corporate risk taking and control variables Firm characteristics and CEO characteristics. The values are derived from public databases and are considered as correct. A sense check is performed and afterwards some assumptions and adjustments are made to make the values valuable for this study. These assumptions and adjustments are explained in detail in the note below table 5.

Panel B presents four measures for corporate risk taking. The Debt Ratio is winsorized at 1% as it contains outliers in the initial sample. This means that the highest and lowest 1% of the sample values are set to the value of the 1st_{percentile and 99}th_{percentile respectively. The ratio} after winsorizing has a minimum value of 0.12 and a maximum of 1.24. The reason for values above 1 is that the firm has a negative equity. The mean of the Debt Ratio in the study of

(30)

N valid N missing Mean Median SD Min Max

Panel B: Corporate risk taking

(1) Debt Ratio 5,402 7 0.57 0.57 0.21 0.12 1.24

(2) Quick Ratio 5,219 190 1.51 1.12 1.66 0.13 54.90

(3) Interest Coverage Ratio 5,028 381 17.49 6.99 28.69 -0.63 119.40

(4) Corporate Risk Policy 3,547 1,862 0.04 0.02 0.06 0.00 0.87

Panel C: Firm characteristics

(5) Firm Size 5,409 0 8.58 8.51 1.34 0.95 13.01

Panel D: CEO characteristics

(6) CEO Gender 5,409 0 0.98 1.00 0.13 0.00 1.00

(7) CEO Age 5,409 0 55.84 56.00 6.72 34.00 84.00

(8) CEO Compensation 5,409 0 0.27 0.20 0.23 0.00 1.00

(9) CEO Tenure 5,409 0 6.39 5.00 6.17 0.00 42.00

Table 5: Descriptive statistics panel B, C (excl. Firm Industry) and D

Note: (1) The Debt Ratio is calculated as the Total Liabilities divided by Total Assets. The underlying

variables are retrieved from the Compustat database. The Compustat database codes are LT and AT respectively. For seven observations, at least one of the two variables is missing in the database. For some observations the Total Liabilities exceed the Total Assets. This is possible when the equity of the firm is negative. The initial data derived from Compustat contained Debt Ratio’s up to 2.39. This is the reason for the decision to winsorize the Debt Ratio at 1% to eliminate these outliers. The percentage of winsorizing is kept as low as possible to maintain the most initial data, while eliminating the outliers. (2) The Quick Ratio is calculated as the Current Assets minus Inventory divided by Current Liabilities. The underlying variables are retrieved from the Compustat database and have the following database codes respectively: ACT, INVT and LCT. For 190 observations at least one of the three values is missing. (3) The Interest Coverage Ratio is calculated as the Earnings Before Interest and Tax (EBIT) divided by the Interest Expense. The underlying variables are retrieved from the Compustat database. The Compustat database codes are EBIT and XINT respectively. As the value for interest could be low and value for EBIT high, the range of the Interest Coverage Ratio is wide. The minimum value is minus 6,415 and maximum plus 97,931 before winsorizing. Therefore the decision is made to winsorize the Interest Coverage Ratio at 5% to eliminate these outliers. The percentage of winsorizing is kept as low as possible to maintain the most initial data, while eliminating the outliers. For 381 observations, at least one of the two values is missing. (4) The Corporate Risk Policy measure is calculated as the Research and Development Expense divided by Total Assets. The values are derived from the Compustat database and have XRD and AT as database codes. For 1,862 observations, no R&D values are available in the Compustat database. (5) The Firm Size of the company is measured as the natural logarithm of Sales. The values are derived from the Compustat database and have SALE as database code. (6) CEO Gender is derived from the Execucomp database and is named GENDER. (7) CEO Age is derived from Execucomp and has database code AGE. The database is missing the age of 245 CEOs who are part in this study. These missing data is derived manually from the BoardEx database. (8) The CEO Compensation is calculated as Salary divided by Total Compensation. The data is retrieved from Execucomp and have the database codes SALARY and TDC2 respectively. As it is not certain whether higher variable pay leads to more corporate risk taking or taking more corporate risk leads to more compensation due to better results, causal uncertainty must taken into account. To mitigate this causal uncertainty, the Salary/Total compensation ratio of previous year (t-1) is

(31)

used. (9) CEO Tenure is calculated with Date Became CEO and Date Left As CEO, which are derived from the Execucomp database. For 69 observations manually research in the BoardEx database was necessary, because the data is not available in the Execucomp database.

Graham et al. (2010) is 0.27. A clarification for the difference is that the sample was different. The research sample of Graham et al. (2010) consists of 10,000 CEOs who subscribed on the CEO magazine in the US. The higher the Debt Ratio, the higher corporate risk taking is considered. The range of the Quick Ratio considers differences in liquidity risk between firms. The mean has a value of 1.51, as this means that the current assets (minus inventory) is about one and half times bigger than the current liabilities. The minimum value of 0.13 means that the firm faces liquidity risk, while the maximum value of 54.90 suggests that there is no liquidity risk. In summary, the higher the Quick Ratio, the lower corporate risk taking is considered. The Interest Coverage Ratio is winsorized at 5%, because this measure contained outliers in the initial sample when the firm had little interest expense. This means that the highest and lowest 5% of the values are set to the value of the 5th_{percentile and 95}th_{percentile respectively. The mean of} 17.49 shows that the average firm has an EBIT which is 17.49 times larger than the interest expense. The higher the Interest Coverage Ratio, the lower corporate risk taking is considered. The mean of the Corporate Risk Policy is 0.04. This can be seen as very low. As this is measure is the R&D expenditure divided by Total Assets, this amount could be high when the Total Assets are high. The mean of 0.04 is the same as the mean of the study of Coles et al. (2006), where the relationship between managerial incentives and risk taking is researched. The higher the value of Corporate Risk Policy, the higher corporate risk taking is considered.

The Firm Size is calculated as the natural log of sales. The minimum value 0.95 represents sales of $2.6 million and the maximum of 13.01 represents the Wal Mart store with sales of $445 billion. The mean of 8.58 is similar to the study of Coles et al. (2006), which has an average value of 8.25.

A dummy variable for CEO Gender is used to determine between female (0) and male (1). The males are over represented with a total of 5,316 (98.28%) observations, while the study consists of 93 (1.72%) female observations. The mean CEO Age is 55.84, with a minimum of 34 and a maximum of 84. Serfling (2014), who researched the relationship between CEO Age and the riskiness of corporate policies, found a mean of 55.22 which is close to the mean of this study. The CEO Compensation is calculated as the Salary divided by Total Compensation. The minimum score is zero as this means that the Total Compensation consist of variable pay only. Contrary the maximum score is 1, as this means that there is no variable part in the Total Compensation. The mean of 0.26 means that 26% of the Total Compensation consists of salary

(32)

(standard pay). The mean CEO Tenure of 6.39 means that the CEO is at that position for an average of 6.39 years. In the study of Serfling (2014) and Coles et al. (2006), the mean of CEO tenure is 7.36 and 7 respectively. The difference could be explained by the difference in tenure of S&P 500 and S&P 1500 firms.

The distribution of observations over the twelve Fama & French industry portfolios is presented in table 6. The most observations are found in the Business equipment industry. Utilities is not represented at all. The 15 observations for the industry Money are added by category Other for statistical reasons. In summary, 10 Fama & French industries are left over to run the regressions.

# F&F industry Frequency Percent

1. Consumer non durables 544 10.06

2. Consumer durables 189 3.49 3. Manufacturing 805 14.88 4. Energy 435 8.04 5. Chemicals 236 4.36 6. Business equipment 1,147 21.21 7. Telecom 221 4.09 8. Utilities 0 0.00 9. Shops 688 12.72 10. Health 527 9.74 11. Money 15 0.28 12. Other 602 11.13 Grand Total 5,409 100.00

Table 6: Frequency table Fama & French industry portfolios

Table 7 shows a correlation matrix of the cultural dimensions of Hofstede (2010) used in this study. The correlation coefficient measures the linear association between independent variables. The correlation coefficient has a range between -1 (negative correlation) and +1 (positive correlation). A correlation coefficient close to 0 means that little or no correlation exists between the variables. All the six cultural dimension have a significant correlation with each other at a 1% significance. Especially Individualism and Power Distance (-0.8212), Indulgence and Power Distance (-0.9015) and Indulgence and Individualism (0.8846) are highly correlated. The highly significant correlation coefficients for independent variables may cause multicollinearity problems. A multicollinearity test is performed in the next paragraph. A correlation matrix of the variables of panel B, C and D are shown in appendix III. This correlation matrix excludes the Firm Industry variable. The variables presented in appendix III show little correlation, which indicates no existence of multicollinearity.

(33)

Hofstede dimensions PD UA IND MAS LTO IR

Power Distance (PD) 1

Uncertainty Avoidance (UA) 0.374*** 1

Sig. [0.000]

Individualism (IND) -0.821*** -0.406*** 1

Sig. [0.000] [0.000]

Masculinity (MAS) -0.383*** -0.470*** 0.522*** 1

Sig. [0.000] [0.000] [0.000]

Long term orientation (LTO) 0.475*** 0.310*** -0.674*** -0.583*** 1

Sig. [0.000] [0.000] [0.000] [0.000]

Indulgence (IR) -0.902*** -0.326*** 0.885*** 0.368*** -0.497*** 1

Sig. [0.000] [0.000] [0.000] [0.000] [0.000]

Table 7: Correlation matrix of Hofstede’s cultural dimensions (* significant at 10%, ** significant at 5%, *** significant at 1%)

5.3 Multicollinearity test

In this paragraph a multicollinearity test is performed. Multicollinearity indicates a (strong) presence of correlation among the independent variables. This could lead to interpretation problems. For example, when an independent variable goes up, not only the dependent variable goes up, but also another independent variable. This effect is caused by the increase of standard errors of the coefficient. To detect multicollinerarity, the Variance Inflation Factor (VIF) is used. As the existing literature use different rules of thumb for the VIF (O’Brien, 2007), in this paper it is assumed that when an independent variable has a VIF between 0 and 5, there is no indication for collinearity. A VIF between 5 and 10 means that there is collinearity, but not too severe. Values above 10 indicate serious collinearity.

The results for the multicollinearity test can be found in table 8. The results are presented from the highest VIF to the lowest VIF per independent variable per model. There is a strong indication for multicollinearity of three dimensions of Hofstede: Indulgence (>9), Individualism (>7.5) and Power Distance (>5). All other independent variables are below 5, which indicate no collinearity. The high multicollinearity between Indulgence, Individualism and Power Distance was expected as these independent variables are highly correlated with each other (table 7). Following O’Brien (2007), one of the remedies to reduce multicollinearity is to remove one of the independent variables. After removing the independent variable with the highest VIF (Indulgence), the VIF of all independent variables left over become lower than 5. This indicates

(34)

that there is no multicollinearity in one of the four models. Therefore in the next chapter, the complete analysis with all Hofstede’s cultural dimensions is performed without Indulgence.

VIF with dimension Indulgence VIF without dimension Indulgence

Variable Model

1 Model 2 Model 3 Model 4 Model 1 Model 2 Model 3 Model 4

IR 8.99 9.17 8.75 9.44 - - - - IND 7.51 7.59 7.91 8.03 4.68 4.65 4.60 4.34 PD 5.66 5.69 5.01 5.32 3.25 3.25 3.39 2.96 LTO 2.26 2.25 2.04 2.58 2.20 2.19 2.03 2.46 MAS 1.83 1.82 1.86 1.79 1.80 1.80 1.80 1.79 UA 1.39 1.42 1.39 1.61 1.38 1.40 1.39 1.54 CEO Age 1.21 1.21 1.19 1.22 1.21 1.21 1.18 1.20 CEO Tenure 1.21 1.21 1.19 1.21 1.20 1.20 1.18 1.20 Firm Size 1.16 1.17 1.16 1.18 1.16 1.17 1.16 1.18 CEO Comp. 1.10 1.11 1.11 1.10 1.10 1.10 1.11 1.10 CEO Gender 1.01 1.01 1.01 1.02 1.01 1.01 1.01 1.02 Mean VIF 3.03 3.06 2.97 3.14 1.90 1.90 1.89 1.88

The impact of a CEO’s cultural background on corporate risk taking