The impact of executive compensation on
corporate risk-taking in a culturally
diversified environment
Jens Leiker
– s2700484
Abstract: Using data (3,843 obs.) from international firms from 13 countries over a 10-year
period (2007-2016), this research focusses on the relationship between the level of executive compensation and the level of corporate risk-taking. Further, we also investigate the potential moderating effects of national cultures on this specific relationship, using three of Hofstede’s cultural dimensions. Our findings show that there are no significant indications that higher executive compensation is related to higher levels of corporate risk-taking. However, the three moderating variables do show strong evidence of affecting corporate risk-taking on their own. Nevertheless, our outcomes do not show significant signs of national cultures moderating the effect between executive compensation and corporate risk-taking.
Keywords: Executive compensation, Corporate risk-taking, Cultural dimensions
Master Thesis
Study Programme: MSc International Financial Management
Supervisor: Dr. W. Westerman
2
1. Introduction
Research into corporate risk-taking has drastically increased over the past few decades. Many researchers have started looking into excessive corporate risk-taking, often following the financial crisis of 2008 and the many scandals that surfaced into the public eye in the early 2000s. As a result, the concept of corporate risk-taking is extensively researched in order to find determining factors that motivate individuals to take excessive corporate risks. In this research, we focus on one of these determinants, namely executive compensation. One of the purposes behind executive compensation is to reward executives for their performances. This performance is generally linked to the firm’s own performance.
Therefore, executives are often said to be focused on improving the firm’s results in order to increase their own earnings. As a result, executives are often claimed taking unnecessary risks to increase the firm’s performance. These unnecessary risks tend to look positive on the short-term but may turn out negatively for the firm in the long-short-term.
In our research, we have opted to use the two methods to measure corporate risk-taking: Std ROA and Std ROE. The common consensus is that ROA type measurements are better
indicators. Nevertheless, in this research we have decided to implement both measurements as proxies for corporate risk-taking, since it should provide more conclusive results.
3
In order to create our findings, we use a sample that consists of 3,843 observations from 13 countries across the timeframe of 2007-2016. The countries involved in this research are Belgium, Denmark, Finland, France, Germany, Ireland, Luxembourg, Netherlands, Norway, Spain, Sweden, Switzerland and the United Kingdom. Pre-existing research often focuses primarily on the United States, as well as a combination of the United States with certain European countries. We have decided to focus exclusively on European countries, in order to give more insights into this less explored domain.
Our findings show that there are no significant implications that executive compensation is related to higher levels of corporate risk-taking. Therefore, these results do not show definite implications that company policymakers should address the manner in which executive compensation packages are constructed. Furthermore, the findings also imply that all three cultural dimensions scores do not moderate the main relationship. However, our findings do show that all three cultural dimensions scores can have a direct impact on corporate risk-taking, even though these effects will most likely be rather limited.
4
2. Literature review
2.1 Executive compensation
In the past decades, the interest into executive compensation has grown tremendously. For example, for researchers like Hall & Murhpy (2003) the interest was spiked by simply the increased levels of executive compensation over the past decades, whereas the interest of Ayadi et al. (2012) has grown following the financial crises of 2008. The latter authors claim that the correlation between executive compensation and corporate risk-taking has “helped” in causing the financial crises of 2008. Authors like Conyon et al. (2011) already were looking into executive compensation and its effects on corporate risk-taking shortly before the financial crises of 2008, due to the many executive compensation scandals that surfaced into the public eye in the early 2000s.
In general, executive compensation can be divided into five components (Frydman & Jenter, 2010): salary, annual bonus, payouts from long-term incentive plans, restricted option grand and restricted stock grants. In this research we follow the approach used by Fernandes et al. (2013), using total compensation as the measurement for executive compensation. Total compensation is considered as the total sum of yearly salary plus cash and stock bonuses. The whole purpose behind executive compensation, or compensation in general, is to attract new talent, motivate individuals and to reward them for their performances. As a result, the latter two of the three factors play an important role in corporate risk-taking.
5
are often said to focus on improving the results regarding these measurements in order to increase their own earnings. As a result, executives may be taking unnecessary risks to
increase the firm’s performance, which tends to negatively turn out in the long-term. Jensen & Murphy (1990) show that (annual) cash compensation only has a small positive effect on the actual Return of Assets of the firm in the following period(s).
A well-known theory called the agency theory is often introduced by many authors to
reinforce the abovementioned argument. For instance, Jensen and Meckling (1976) define the agency theory as a contract under which one or more persons, the principal(s), employ the other person, the agent, to perform a certain service on their behalf, involving the delegating of authority to the agent. However, if both parties involved in the relationship seek to utilize their own interests, there is good reason to believe that the agent will not always act in the best interest of the principal. This theory can be applied to the relationship between the shareholders or owners of the firms, the principals, and the executives of the firm, the agents. The executives may be taking unnecessary risks to maximize their own interests. As
mentioned above, these decisions might sometimes appear beneficial for the firm. However, often turn out to have a negative impact on the firm in the long-term. Jensen and Meckling (1976) actually show that executives are less risk averse than the shareholders are themselves in regard to making important decisions. However, the level of risk-taking for both the shareholders and the executives could vary between countries. Therefore, this is one of the reasons we also included the analysis of cultural dimensions in our research.
6
monitoring is the key in these situations, to prevent unnecessary risks from happening. However, there are undoubtedly extensive costs involved in the process of monitoring. Therefore, certain principals have opted for the decision to award agents in order try to align both interests. Nevertheless, this decision to award executives, in order to align their interests with the shareholders, only creates a vicious cycle in which the executives are getting even more opportunities to take unnecessary risks, in order to maximize their own interests. In addition, the upper echelons theory claims that managerial background characteristics of top-level executives also impact decision-making processes, ultimately affecting
organizational outcomes (e.g., Chatterjee et al., 2007; Hambrick et al., 1984). Executive characteristics such as age, education and experience tend to have an impact on their level of risk-taking. Zhang and Rajagopalan (2010) argue that younger executives tend to be less risk averse than their older counterparts. One of the arguments behind their claim, is the fact that older executives have often built up a respectable reputation during their careers, meaning that they are less likely to involve in risky decisions to gain personal earnings, since they fear the risk of damaging their reputation. On top of that, younger executives tend to be more flexible and knowledgeable in regard to new changes in their firm’s environment. Therefore, younger executives tend to make riskier decisions in comparison to older executives, due to the fact that they feel more comfortable in dealing with new innovations and developments. Wally and Baum (1994) focus on the aspect of executive formal education. The authors argue that the amount of formal education or the number of degrees held, acts as a proxy for an executive’s cognitive ability. Formal education helps to acquire and process more complex information and increases the rate of decision-making. However, these aspects could
7
prior career experience, the amount of time spent working in various roles and positions before becoming executives, influences the level of risk taking in achieving personal gains. The authors argue that these prior career experiences affect how executives interpret important information and how they make their decisions. Executives might feel more personally connected to the firm, meaning that personal compensation plays a smaller role in making decisions, resulting in less risky decision-making. All in all, the common consensus is that executive characteristics do play a role in the level of risk-taking. However, the authors do emphasize that levels of risk-taking increases when executives feel that they are given easy opportunities to increase their personal gains, without facing the negative consequences. Following the aforementioned arguments, we believe that an increase in executive compensation increases the level of corporate risk-taking. Therefore, we construct the following hypothesis:
H1: The level of executive compensation is positively related to the level of corporate
risk-taking.
2.2 Cultural dimensions
8
Collectivism), Uncertainty Avoidance, Long-Term Orientation (versus Short-Term) and Restraint (versus Indulgence).
Cultures that score high on Power Distance emphasize high inequality in society in terms of physical and intellectual capacities among people. However, cultures that score low on Power Distance try to minimize inequalities in power and wealth as much as possible. Furthermore, cultures that score high on Individualism emphasize individual freedom and achievement, whereas cultures that score low on Individualism are more emphasized on group cohesion and sharing a common purpose. Cultures that score high on Uncertainty Avoidance feel
uncomfortable with uncertainty and ambiguity. These cultures prefer strict and clear rules and guidelines. On the other hand, cultures that score low on Uncertainty Avoidance emphasize less resistance to change and value innovation. Cultures scoring high on Long-Term
9 2.2.1 Power Distance
The first of three cultural dimensions that we focus on is Power Distance. Shane (1993) argues that individuals in countries with a high score on Power Distance, often have low levels of communications between hierarchies, while there are often also high inequalities between hierarchies as explained by Hofstede (2011). Mihet (2013) reinforces both arguments by emphasizing that individuals in countries with high Power Distance feel less threatened by others in terms of establishing themselves. Therefore, these individuals are less likely to take higher risks. Thompson et al. (2009) contribute to the abovementioned argument, by
suggesting that individuals in low Power Distance cultures are more eager to improve their personal positions and that these countries have considerable higher social mobility, resulting in more opportunity-seeking behavior.
Following the aforementioned arguments, we believe that a high score on Power Distance weakens the relationship between executive compensation and corporate risk-taking. Therefore, we construct the following hypothesis:
H2: A high country-level score on Power Distance weakens the relationship between
executive compensation and corporate risk-taking.
2.2.2 Individualism
10
that, Triandis et al. (1998) argue that individuals in low Individualism countries feel more obliged to follow social duties, expectations, roles and other societal influences.
Following the aforementioned arguments, we believe that a high score on Individualism strengthens the relationship between executive compensation and corporate risk-taking. Therefore, we construct the following hypothesis:
H3: A high country-level score on Individualism strengthens the relationship between
executive compensation and corporate risk-taking.
2.2.3 Uncertainty Avoidance
The final of three cultural dimensions that we focus on is Uncertainty Avoidance. Li and Zahra (2012) emphasize that individuals in low Uncertainty Avoidance countries are more comfortable with uncertainty and unpredictability. On the other hand, individuals in high Uncertainty Avoidance countries are more afraid in situations of uncertainty. Therefore, these individuals will more likely avoid innovative projects or require a higher risk premium. All in all, these individuals are less likely to take high risk decisions. Gomez-Meija and Welbourne (1991) even show that this desire to minimize uncertainty, is reflected in the way they construct executive compensation plans in high uncertainty countries.
Following this reasoning, we believe that a high score on Uncertainty Avoidance weakens the relationship between executive compensation and corporate risk-taking. Therefore, we
construct the following hypothesis:
H4: A high country-level score on Uncertainty Avoidance weakens the relationship
11
3. Methodology
3.1 Dependent variable
The dependent variable in this research is corporate taking. The proxy for corporate risk-taking can be measured via various ways. The two most general and often preferred methods are by applying the Return on Assets (ROA) and Return on equity (ROE) measures. However, the common consensus is that ROA type measurements are better indicators, since these take into account liabilities, which in case of risk is an important aspect to consider. Nevertheless, in this study we have decided to implement both measurements as proxies for corporate risk-taking, since it should provide more conclusive results. We follow the specific measurements created by John et al. (2008), which calculate the degree of corporate risk-taking as the country-adjusted volatility of firm-level earnings, Std ROA. The deviation of the firm’s Net Income/Total Assets as to the country average (for the corresponding year), followed by calculating the standard deviation of this measure for each firm (for each firm with available earnings and total assets for at least 4 years in 2007 to 2016). An identical approach is used to calculate the measurement; Std ROE. The deviation of the firm’s Net Income/Total Equity as to the country average (for the corresponding year), followed by calculating the standard deviation of this measure for each firm (for each firm with available earnings and total assets for at least 4 years in 2007 to 2016). It is important to mention that we have decided to include firms that showed negative values for both Net Income/Total Assets as Net
Income/Total Equity. Our reasoning behind this decision, is the fact that firms can realistically have negative values in certain years, however, this could obviously change in other years, especially since we only calculate earnings and total assets for firms with at least 4 years in 2007 to 2016. Lastly, both variables are winsorized at the one and 99th percentile level in both
12
3.2 Independent variable
The independent variable in this research is the natural logarithm of Executive Compensation (Ln Executive Compensation). Similar to the dependent variable, there are also various methods to measure executive compensation. In this research we follow the approach used by Fernandes et al. (2013), using total compensation as the measurement for executive
compensation. Total compensation is considered as the total sum of yearly salary plus cash and stock bonuses. On top of that, the natural logarithm of total executive compensation is taken, in order to address skewness of the variables and create a more normalized dataset. Lastly, the variable Ln Executive Compensation is winsorized at the one and 99th percentile level in both tails of the distribution in order to remove the effects of outliers. Appendix 1 displays an additional overview of how Ln Executive Compensation has been measured and sourced.
3.3 Moderating variables
13
dimensions are still considered to be one of the best indicators for identifying cultural dimensions among countries. All three moderating variables will be explained below, however, Appendix 1 displays an additional overview of how all three variables have been measured and sourced.
The first of three moderating variables is the score on Power Distance. Cultures that score high on Power Distance emphasize high inequality in society in terms of physical and intellectual capacities among people. To the contrary, cultures that score low on Power
Distance try to minimize inequalities in power and wealth as much as possible. Table 1 shows the Power Distance scores for each country that will be used in this research. It is important to notice that almost all countries display scores lower than 50. Only Spain, Belgium and France show scores higher than 50, with scores of 57, 65 and 68, respectively.
Country Score Country Score
Belgium 65 Netherlands 38
Denmark 18 Norway 31
Finland 33 Spain 57
France 68 Sweden 31
Germany 35 Switzerland 34
Ireland 28 United Kingdom 35
Luxembourg 40
Table 1: Power Distance score per country
Similar to the independent variable, the natural logarithm of the Power Distance score is taken in order to address skewness of the variables and create a more normalized dataset.
Furthermore, also this variable is winsorized at the one and 99th percentile level in both tails of the distribution in order to remove the effects of outliers.
14
research. In contrast to the previous model, all Individualism scores are actually above 50. Most countries display scores around 60-70. However, Spain displays a relatively low score of 51, while the United Kingdom shows a relatively high score of 89.
The natural logarithm is also taken for this variable to create a more normalized dataset. On top of that, also this variable is winsorized at the one and 99th percentile level in both tails of
the distribution in order to remove the effects of outliers.
Country Score Country Score
Belgium 75 Netherlands 80
Denmark 74 Norway 69
Finland 63 Spain 51
France 71 Sweden 71
Germany 67 Switzerland 68
Ireland 70 United Kingdom 89
Luxembourg 60
Table 2: Individualism score per country
The final of three moderating variables is the score on Uncertainty Avoidance. Cultures that score high on Uncertainty Avoidance feel uncomfortable with uncertainty and ambiguity. These cultures prefer strict and clear rules and guidelines. On the other hand, cultures that score low on Uncertainty Avoidance emphasize less resistance to change and value innovation. Table 3 shows the Uncertainty Avoidance scores for each country that will be used in this research. The Uncertainty Avoidance scores actually show the highest dispersion of all the three moderating variables. Countries like Denmark, Sweden and Ireland show relatively low scores of 23, 29 and 35, respectively. On the other hand, countries such as France, Spain and Belgium show scores higher than 85.
Country Score Country Score
Belgium 94 Netherlands 53
Denmark 23 Norway 50
Finland 59 Spain 86
France 86 Sweden 29
Germany 65 Switzerland 58
Ireland 35 United Kingdom 35
Luxembourg 70
15
Similar to the previous moderating variables, the natural logarithm of the Uncertainty Avoidance score has been taken. Lastly, this variable is also winsorized at the one and 99th percentile level in both tails of the distribution in order to remove the effects of outliers.
3.4 Control variables
There are three control variables used in this research: Size, Leverage and Capex. These control variables are implemented in order to improve the results of this research, considering that control variables improve the relationship between the dependent and independent variable, by isolating this correlation. All three control variables will be explained below, however, Appendix 1 displays an additional overview of how all three variables have been measured and sourced.
The first of three control variables is Size. This specific control variable is included in this research following the argument by Fernandes et al. (2013). The authors argue that Size should be included due to the fact that previous research has shown a positive correlation between executive pay and firm size. On top of that, Diaz and Sanchez (2008) emphasize the implementation of Size as a control variable, considering that larger firms are better at monitoring their executives compared to their smaller counterpart, allowing less room for executives to take unnecessary risks. Following abovementioned authors, Size is measured as the natural logarithm of total assets. Lastly, this variable is also winsorized at the one and 99th
percentile level in both tails of the distribution in order to remove the effects of outliers. The second of three control variables is Leverage. This variable is also included based upon the reasoning of Fernandes et al. (2013). The authors argue that leverage increases the riskiness of executive compensation, risk premiums and higher levels of executive
16
Therefore, leverage can be considered as a useful control variable in this research. Leverage is measured as the ratio of Total Debts over Total Assets. Furthermore, this variable is also winsorized at the one and 99th percentile level in both tails of the distribution in order to
remove the effects of outliers.
The final control variable is Capex. Many authors (e.g., Coles et al., 2006; Habib et al., 2015 and Lie et al, 2013) argue that Capex has a positive effect on corporate risk-taking. Therefore, an increase in Capex should also result in an increase of corporate risk-taking. Capex is measured as the ratio of Capital Expenditures over Total assets. Lastly, this variable is also winsorized at the one and 99th percentile level in both tails of the distribution in order to remove the effects of outliers.
3.5 Data collection
The main data source is the Compustat Global Vantage Database for the period 2007-2016. This particular sample has been selected based on availability of firm-level data, but also the availability of country-level data for cultural values and key country characteristics. As mentioned previously, we have winsorized all moderating and firm-level variables at the winsorized at the one and 99th percentile level in both tails of the distribution in order to remove the effects of outliers following the reasoning of Li et al. (2013).
17
Furthermore, the United Kingdom offers a relatively higher amount of collected values in comparison to any other country in this sample. However, since it was decided to include a sizeable amount of other European countries, it could be ensured that the values of the United Kingdom account roughly for only 34.93% of this sample. Unfortunately, Portugal could not be included into the of list countries as a result of substantial number of missing values for most years of the timeframe.
The data concerning the moderating variables has been extracted from the Geert Hofstede Dimension Data Matrix, which can be found on his personal website. The data concerning the executive compensation has been extracted from the BoardEx database. It is important to mention that this database includes data regarding a substantial number of professional positions that can exist within a firm. Therefore, the data has been filtered down to the professional titles that characterize CEO positions. This also prevents the potential error of comparing much lower-level professional positions compensations to top tier compensation levels.
3.6 Model specification
18
only test the effect of Ln Executive Compensation on the dependent variable(s), including the control variables. Thereafter, the moderating variables will be subsequently added in models two, three and four, following de order in which we have described the moderating variables throughout this paper: Ln Power Distance, Ln Individualism and Ln Uncertainty Avoidance. In the final three models, we add the three interaction terms in the same order as described above.
19
4. Empirical results
Table 4 shows the distribution of the sample of this research per country. The total amount of observations is 5,400. The United Kingdom is the country with the highest representation of firms at a percentage of 34.93%, followed by Germany with 14.37% and France with a percentage of 11.39%. The least contributing countries are Ireland with 1.85%, followed by Norway with 1.83% and Luxembourg with a percentage of 0.91%. These latter three countries do not present a high percentage of contribution in this European sample. However, as
mentioned above, the incorporation of more countries will benefit the diversity in terms of cultural dimensions scores.
Country N % Country N % Belgium 186 3.44 Netherlands 339 6.28 Denmark 184 3.41 Norway 99 1.83 Finland 224 4.15 Spain 169 3.13 France 615 11.39 Sweden 278 5.15 Germany 776 14.37 Switzerland 495 9.17
Ireland 100 1.85 United Kingdom 1,886 34.93
Luxembourg 49 0.91 Total 5,400 100
Table 4: This table shows the distribution of data for each country.
Table 5 below shows the descriptive statics of the variables used in this research. The descriptive statics show the minimum, maximum, mean, standard deviation and number of observations of each of the variables. As mentioned earlier, all variables are winsorized at the one and 99th percentile level in both tails of the distribution in order to remove the effects of outliers. The minimum and maximum value of the dependent variable Std ROA are
20
The independent variable Ln Executive Compensation displays values varying between 1.00e-06 and 0.107408. Followed by a mean of 0.0023101 and the standard deviation of 0.0039643. The first moderating variable, Power Distance, displays values ranging from 18 to 68 with a mean of 39.80042 and standard deviation of 12.95113. Individualism shows a minimum score of 51 and a maximum score of 89, followed by a mean of 75.45459 and a standard deviation of 10.55218. Uncertainty Avoidance shows values ranging from 23 to 94 with a mean of 53.91855 and a standard deviation of 20.62869 All in all, Uncertainty Avoidance shows the most prominent diversification of distribution, whereas Individualism displays the smallest distribution in values. The minimum and maximum of the control variable Size are 1.375697 and 12.24467, with a mean of 7.545124 and a standard deviation of 2.427477. Furthermore, Leverage shows a minimum and maximum of 0 and 0.8937725, respectively. Leverage has a mean of 0.2112263 and a standard deviation of 0.1459354. Capex displays a minimum of 0.0006173 with a maximum of 0.2041408. Capex has a mean of 0.405644 and a standard deviation of 0.028548. The natural logarithms for the independent and moderating variables are calculated afterwards.
Observations Min Max Mean Std. Dev.
Std ROA 3,483 0.0107487 0.624657 0.12220921 0.1154603
Std ROE 3,483 0.0381916 16.91978 1.260902 2.252484
Executive Compensation
3,483 1.00e-06 0.107408 0.0023101 0.0039643
Power Distance score 3,483 18 68 39.80042 12.95113
Individualism score 3,483 51 89 75.45459 10.55218 Uncertainty Avoidance score 3,483 23 94 53.91855 20.62868 Size 3,483 1.375697 12.24467 7.545124 2.427477 Leverage 3,483 0 0.8937725 0.211263 0.1459354 Capex 3,483 0.0006173 0.2041408 0.0405644 0.028548
Table 5: This table describes the variables used in the analysis. The sample period is from 2007 to 2016. The following variables are described: Std ROA, Std ROE, Executive Compensation, Power Distance score, Individualism score, Uncertainty Avoidance score, Size, Leverage and Capex. All variables, except the
21
Table 6 shows the coefficient correlation between the variables used in this research. The purpose of this table is to show the correlation between the dependent variable Std ROA, the independent variable, moderating variables and control variables. A high value indicates that two or more variables are highly correlated, a term called Multicollinearity. The importance of presenting these correlations is to provide an overview in regard to potential biases in our final results. Luckily, in this table there is no correlation higher than 0.4 (or lower than -0.4), which indicates that there is no sign of multicollinearity. Once again, it reflects that there is no high correlation between two or more explanatory variables.
Std ROA Ln Executive Compensation Ln Power Distance Ln Individual ism Ln Uncertainty Avoidance
Size Leverage Capex
Std ROA 1.0000 Ln Executive Compensation 0.0074 1.0000 Ln Power Distance -0.1623 -0.0065 1.0000 Ln Individualism 0.4153 0.0191 -0.2094 1.0000 Ln Uncertainty Avoidance -0.3444 -0.0018 0.3740 -0.3420 1.0000 Size -0.2169 -0.1194 0.0503 -0.2893 0.1457 1.0000 Leverage -0.0117 -0.0298 -0.0001 -0.0347 -0.0334 0.2584 1.0000 Capex -0.0476 0.0097 -0.0110 -0.0421 0.0524 0.0924 0.1378 1.0000
Table 6: Correlation matrix. This table shows the correlation between the dependent variable Std ROA, independent variable, moderating variables and control variables.
The next correlation matrix, table 7, shows the correlation between the dependent variable Std ROE, the independent variable, moderating variables and control variables. Logically, most of the correlations are equivalent to the correlations in table 6. In this table there is also no sign of multicollinearity. Nevertheless, it is important to notice that the correlation between this dependent variable and most other variables, is lower than in the previous situation.
22 Std ROE Ln Executive Compensation Ln Power Distance Ln Individual ism Ln Uncertainty Avoidance
Size Leverage Capex
Std ROE 1.0000 Ln Executive Compensation 0.0099 1.0000 Ln Power Distance -0.1104 -0.0065 1.0000 Ln Individualism 0.1815 0.0191 -0.2094 1.0000 Ln Uncertainty Avoidance -0.1668 -0.0018 0.3740 -0.3420 1.0000 Size -0.1112 -0.1194 0.0503 -0.2893 0.1457 1.0000 Leverage -0.0483 -0.0298 -0.0001 -0.0347 -0.0334 0.2584 1.0000 Capex 0.0095 0.0097 -0.0110 -0.0421 0.0524 0.0924 0.1378 1.0000
Table 7: Correlation matrix. This table shows the correlation between the dependent variable Std ROE, independent variables, moderating variable and control variables.
In the first model of table 8, we exclusively analyze the effect of executive compensation on corporate risk-taking, using the measurement of Std ROA. The coefficient of Ln Executive Compensation is negative at -0.00153 and insignificant, indicating that there is no evidence that the level of executive compensation on its own, increases the level of corporate risk-taking. Therefore, there is no evidence to support hypothesis 1. However, all three control variables are significant at the 1% level. Both Leverage and Capex display positive coefficients, essentially confirming our initial expectations. When in fact, the variable Size turns out to have a negative coefficient, a contrasting result in comparison to what we expected initially. The adjusted R-squared for model 1 is 0.151. This percentage can be considered reasonable, given the restricted sample size of our research.
risk-23
taking has shown nearly no difference in results compared to model 1. However, the adjusted-R squared has increased compared to the previous model.
Model three of table 8 includes the effect of the natural logarithm score of Individualism (Ln Individualism) on corporate risk-taking. The coefficient of Individualism is positive at 0.296 and significant at the 1% level, indicating that there is strong evidence that an increased score on Individualism increases the level of corporate risk-taking. The variable Ln Power Distance also remained significant at the 1% level in model three. The adjusted-R squared has also increased in model three.
24
(1) (2) (3) (4)
Std ROA Std ROA Std ROA Std ROA
Ln Executive Compensation -0.00153 (0.00134) -0.00153 (0.00133) -0.00106 (0.00124) -0.000937 (0.00124) Ln Power Distance -0.0596*** (0.00614) -0.0315*** (0.00586) -0.0174 (0.0113) Ln Individualism 0.296*** (0.0126) 0.220*** (0.0196) Ln Uncertainty Avoidance -0.0546*** (0.0108) Size -1.0109*** (0.00781) -0.0105*** (0.000773) -0.00547*** (0.000754) -0.00569*** (0.000752) Leverage 0.0409*** (0.013) 0.0391*** (0.0128) 0.0268*** (0.0120) 0.0196*** (0.0120) Capex 0.136*** (0.0644) 0.131** (0.0636) 0.103*** (0.0595) 0.0783*** (0.0595) _cons 0.191*** (0.0107) 0.405*** (0.0245) -1.009*** (0.0642) -0.641*** (0.0968)
Year Fixed Effects Yes Yes Yes Yes
N Adjusted R² 3,843 0.151 3,843 0.173 3,843 0.191 3,843 0.191
Table 8: This table shows the regressions of models 1-4 of corporate risk-taking (Std ROA). The symbols ***, **, * denote statistical significance at the 1%, 5%, and 10% levels, respectively.
25
In model six, we look at the interaction effect of Ln Executive Compensation and the score on Individualism. The coefficient of the interaction term Ln Executive compensation X Ln Individualism is -0.00755, yet also insignificant. This means that there is no evidence that an increased score on Individualism, has an increased effect on the relationship between executive compensation and corporate risk-taking. Therefore, there is also not enough evidence to support hypothesis 3.
26
(5) (6) (7)
Std ROA Std ROA Std ROA
Ln Executive Compensation -0.00338 (0.0158_ -0.0387 (0.0447) -0.0827 (0.0678) Ln Power Distance -0.00921 (0.0321) -0.00435 (0.0326) -0.0467 (0.0590) Ln Executive Compensation X Ln Power Distance -0.00119 (0.00436) -0.00194 (0.00444) 0.00413 (0.00832) Ln Individualism 0.220*** (0.0196) 0.167** (0.0657) 0.101 (0.100) Ln Executive Compensation X Ln Individualism -0.00755 (0.00894) -0.0169 (0.0140) Ln Uncertainty Avoidance -0.0547*** (0.0108) -0.0550*** (0.0108) -0.101* (0.0549) Ln Executive Compensation X Ln Uncertainty Avoidance -0.00664 (0.00770) Size -0.00569*** (0.000753) -0.00574*** (0.000755) -0.00571*** (0.000755) Leverage 0.0195*** (0.0120) 0.0193*** (0.0120) 0.0191*** (0.0120) Capex 0.0776*** (0.0595) 0.0789*** (0.0595) 0.0775*** (0.0595) _cons -0.610*** (0.147) -0.362 (0.329) -0.0530 (0.486) Year Fixed Effects
N Adjusted R² Yes 3,843 0.196 Yes 3,843 0.196 Yes 3,843 0.196
27
The concepts of the next two tables, table 10 and 11, are relatively comparable to the two previous two tables. However, the fundamental and essential difference, is the fact that we now measure the effects on corporate risk-taking using the measurement of Std ROE. In the first model of table 10, we exclusively examine the effect of executive compensation on corporate risk-taking. The coefficient of Ln Executive Compensation is negative at -0.00642 and insignificant, indicating that there is no evidence that the level of executive compensation on its own, increases the level of corporate risk-taking. Therefore, there is also no evidence to support hypothesis 1 in this situation. All three control variables are also significant at the 1%. Both Leverage and Capex display positive coefficients, confirming our initial expectations once again. Size has a negative coefficient, emphasizing similar results as to the previous tables. However, it reiterates the contrasting effects in comparison to what we expected initially. The adjusted R-squared for model 1 is 0.113. As expected, the model shows a lower level of explanatory power in comparison to the previous measurement.
In the second model of table 10, the variable Ln Power Distance is included. The coefficient of Ln Power Distance is -0.813. On top of that, the coefficient is significant at the 1% level. Therefore, there is strong evidence that an increased score on Power Distance decreases the level of corporate risk-taking. The adjusted-R squared has slightly increased in this specific model.
28
The final model of table 10, includes the effect of the score on Uncertainty Avoidance (Ln Uncertainty Avoidance) on corporate risk-taking. The coefficient of Ln Uncertainty Avoidance is negative at -0.225. However, the coefficient is not significant. Therefore, this coefficient is different compared to similar coefficient of the previous measurement. This coefficient shows that there is no evidence that an increased score on Uncertainty Avoidance decreases the level of corporate risk-taking. The coefficient of Power Distance has become insignificant again. However, the coefficient of Individualism has remained significant at the 1% level.
(1) (2) (3) (4)
Std ROE Std ROE Std ROE Std ROE
Ln Executive Compensation -0.00642 (0.0267) -0.00654 (0.0266) -0.00275 (0.0263) -0.00226 (0.0263) Ln Power Distance -0.813*** (0.123) -0.590*** (0.124) -0.389 (0.240) Ln Individualism 2.346*** (0.267) 2.031*** (0.417) Ln Uncertainty Avoidance -0.225 (0.229) Size -0.0999*** (0.0155) -0.0947*** (0.0155) -0.0547*** (0.0160) -0.0555*** (0.0160) Leverage 0.367*** (0.258) 0.391*** (0.257) 0.488*** (0.254) 0.518*** (0.256) Capex 1.794*** (1.280) 1.861*** (1.273) 2.082*** (1.261) 2.183*** (1.265) _cons 1.975*** (0.214) 4.895*** (0.490) -6.297*** (1.361) -4.779** (2.058)
Year Fixed Effects Yes Yes Yes Yes
N Adjusted R² 3,843 0.113 3,843 0.124 3,843 0.144 3,843 0.144
29
In table 11 of this research, we examine the moderating effects of the three cultural dimensions scores on the relationship between executive compensation and corporate risk-taking, using Std ROE as the measurement for corporate risk-taking. In the first model of table 9, we look at the interaction effect of Ln Executive Compensation and Ln Power Distance. The coefficient of the interaction term Ln Executive compensation X Ln Power Distance is negative at -0.0192. However, the coefficient is once again insignificant. Therefore, there is no evidence that an increased score on Power Distance has a significant effect on the relationship between executive compensation and corporate risk-taking. As a result, we repeatedly do not have enough evidence to support hypothesis 2.
In the following model, we look at the interaction effect of Ln Executive Compensation and the natural logarithm of the score on Individualism. The coefficient of the interaction term Ln Executive Compensation X Ln Individualism is -0.103, however, it is insignificant. Therefore, this means that there is no evidence that an increased score on Individualism has an effect on the relationship between executive compensation and corporate risk-taking. Once again, there is also not enough evidence to support hypothesis 3.
30
(5) (6) (7)
Std ROE Std ROE Std ROE
Ln Executive Compensation -0.0673 (0.337) -0.550 (0,961) -0.795 (1.443) Ln Power Distance -0.521 (0.683) -0.588 (0.694) -0.352 (1.255) Ln Executive Compensation X Ln Power Distance -0.0192 (0.0927) -0.0293 (0.0945) 0.00448 (0.177) Ln Individualism 2.029*** (0.417) 1.305 (1.398) 0.941 (2.133) Ln Executive Compensation X Ln Individualism -0.103 (0.190) -0.155 (0.298) Ln Uncertainty Avoidance -0.226 (0.229) -0.230 (0.229) -0.489 (1.169) Ln Executive Compensation X Ln Uncertainty Avoidance -0.0370 (0.164) Size -0.0556*** (0.0160) -0.0563*** (0.0161) -0.0561*** (0.0161) Leverage 0.519*** (0.256) 0.521*** (0.256) 0.522*** (0.256) Capex 2.183*** (1.265) 2.166*** (1.265) 2.166*** (1.265) _cons -4.290 (3.136) -0.896 (6.994) 0.823 (10.34) Year Fixed Effects
N Adjusted R² Yes 3,843 0.144 Yes 3,843 0.144 Yes 3,843 0.144
31
5. Conclusion
5.1 General conclusion
Interest into executive compensation has drastically increased over the past couple of decades. The reasoning behind this growth of attention can be traced back to various reasons. Hall & Murphy (2003) mention that their interest was spiked by the fact that they noticed
continuously increasing executive compensation levels over de past decades. Other
researchers have started looking into executive compensation levels due the financial crises of 2008, as to find whether the aspect of increasing executive compensation has helped causing this specific crisis (Ayadi et al., 2012). Lastly, several researchers (e.g., Conyon et al., 2011) started looking into executive compensation levels following the many executive
compensation scandals that surfaced in the early 2000s.
32
For those reasons, we have decided to look into this relationship between executive
compensation and corporate risk-taking. However, we also believed the particular culture of the country that executives operate in, could also have an impact on the decision-making process by executives in order improve their level of compensation, ultimately impacting the level of corporate risk-taking. Therefore, this research has also examined the aspect of national cultures. Existing research has rather neglected to look into these important circumstances. Therefore, we have not only enhanced existing research by including this aspect of national cultures, but it could also potentially be another motivation for future research to look into cultural dimensions.
Two proxies for corporate risk-taking were implemented in creating our results, Std ROA and Std ROE, since it would create more conclusive results. Previous researchers (Jensen et al., 1976; John et al., 2008) have demonstrated significant results that the level of executive compensation positively influences the level of corporate risk-taking. However, our results do not show definite implications that we can agree to this claim. Essentially, all fourteen models do not only show insignificant coefficients regarding this relationship, but also demonstrate a negative relation. All in all, our results would basically prove a contrary relationship if the results would be significant. Therefore, our own exclusive results do not emphasize motives for company policymakers to address the manner in which executive compensation packages are constructed.
As explained previously, besides the relationship between executive compensation and corporate risk-taking, we also studied the potential moderating effect of national cultures on the relationship between executive compensation and corporate risk-taking. In this research we have examined three of the six Hofstede (2011) dimensions; Power Distance,
33
identifying cultural dimensions among countries. Furthermore, the decision to focus on these three dimensions exclusively, was based on the limited existing research that has been done in regard to executive compensation, cultural dimensions and corporate risk-taking. These studies found the highest effect caused by these three dimensions on both executive compensation and corporate risk-taking.
Existing research (Shane, 1993; Mihet, 2013; Thompson et al., 2009) argues that individuals in low Power Distance cultures are more eager to improve their personal positions, resulting in more opportunity-seeking behavior. Therefore, we expected that countries with a high country-level score on Power Distance would weaken the relationship between executive compensation and corporate risk-taking. However, our findings imply that the level of Power Distance in a country does not moderate the main relationship between executive
compensation and corporate risk-taking. Nevertheless, the results do align with the above-mentioned authors in the fact that a high country-level score on Power Distance does have a negative, direct impact on corporate risk-taking.
Furthermore, earlier research (Kreiser et al., 2010; Triandis et al., 1998) emphasizes that executives in high level Individualism countries have more autonomy to make their own decisions and feel less obliged to follow social duties. Therefore, these executives are more likely to take riskier decisions for their personal interests. For that reason, we expected that countries with a high country-level score on Individualism would strengthen the main relationship. Nonetheless, our findings also do not imply that the level of Individualism in a country moderates the main relationship. On the contrary, our findings do display significant results that Individualism positively and directly impacts corporate risk-taking.
Lastly, previous research (Li et al., 2012; Gomez-Meija et al., 1991) points out that
34
often reflected in the way firms in these countries construct executive compensation plans. As a result, we expected that countries with a high country-level score on Uncertainty Avoidance would weaken the relationship between executive compensation and corporate risk-taking. Nonetheless, similar to the previous two cultural dimensions, our outcomes imply that the level of Uncertainty Avoidance in a country does not moderate the main relationship between executive compensation and corporate risk-taking. On the other hand, our findings do once again display significant results in the fact that Uncertainty Avoidance negatively and directly impacts corporate risk-taking.
5.2 Recommendations
We have potential recommendations for future research. First of all, this research focuses exclusively on (Western) European countries. Future research could further improve implications by including more countries from across the globe, for instance Asian or East-European countries. On top of that, including additional countries might also improve the subsequent limitation of data availability. The data in regard to executive compensation was relativity limited, which moderately restricted our sample size. The addition of extra countries will improve the potential sample size, in turn providing more favorable circumstances for better results.
Another limitation of this research is the proxy for corporate risk-taking being used. In this research we follow the approach used by John et al. (2008), using Std ROA and Std ROE, as measurements for corporate risk-taking. However, future research could implement a different or third measurement for corporate risk-taking in order to potentially improve the statistical results.
35
(2013), using total compensation as the measurement for executive compensation. However, in the past other researchers have also used the equity-based compensation proxy as
measurement for executive compensation, which takes into account stock and option
compensation. Future research could implement this particular proxy or perhaps even another proxy in order to potentially improve the statistical results.
Moreover, it is important to emphasize that our research could have issues regarding reversed causality, meaning that the dependent variables affect the independent variable, instead of the other way around as we hypothesis in our research. Potentially, simultaneous causality might also even be a potential risk. Future research could potentially avoid this risk by following the advice of Leszczensky and Wolbring (2019), by using a cross-lagged panel model with fixed effects, a model in which one variable is measured at an earlier point in time and is examined in contrast to another variable later in time.
Lastly, future research could include additional or less common cultural dimension
36
References
Ayadi, R., Arbak, E., & Pieter De Groen, W., 2012. Executive compensation and risk taking in European banking. Research Handbook on International Banking and Governance. Alchian, A. A. & Demsetz, H., 1972. Production, information costs, and economic organization-. American Economic Review, 62(5) 777-795.
Bebchuk, L. & Spamann, H., 2010. Regulating Banker’s Pay. Georgetown Law Journal, 98, 247-287.
Chatterjee, A., & Hambrick, D. C., 2007. It’s all about me: Narcissistic chief executive officers and their effects on company strategy and performance. Administrative Science Quarterly, 52, 351–386.
Coles, J. L., Daniel, N. D. & Naveen, L., 2006, Managerial incentives and risk-taking, Journal of Financial Economics, 79, 431–468.
Conyon, M., Fernandez, N., Ferreira, M., Pedro, M. & Murphy, K.J., 2011. The Executive Compensation Controversy: A Transatlantic Analysis.
Diaz, M. A., & Sanchez, R., 2008. Firm size and productivity in Spain: A stochastic frontier analysis. Small Business Economics, 30, 315-323.
Fernandes, N., Ferreira, M. A., Matos, P., & Murphy, K. J. (2013). Are U.S. CEOs Paid More? New International Evidence. The Review of Financial Studies, 26(2), 323-367. Frydman, C., & Jenter, D., 2010. CEO Compensation. Annual Review of Financial Economics, 2, 75-102.
37
Habib, A. & Hasan, M., 2015. Firm life cycle, corporate risk-taking and investor sentiment. Accounting and Business Research, 57(2), 465-497.
Hall, E. T., & Hall, M. R., 1990. Understanding cultural differences. Yarmouth, Me; Intercultural Press.
Hall, B. J., & Murphy, K. J., 2003. The trouble with stock options. National Bureau of Economic Research.
Hambrick, D.C., & Mason, P.A., 1984. Upper echelons: The organization as a reflection of its top managers. Academy of Management Review, 9, 193–206.
Hofstede, G. (2001). Culture’s consequences: Comparing values, behaviours, institutions, and organizations across nations. Thousand Oaks, CA: Sage Publications.
Hofstede, G. (2011). Dimensionalizing Cultures: The Hofstede Model in Context. Online Readings in Psychology and Culture, 2(1), 1-26.
Jensen, M. C., & Meckling, W. H., 1976. Theory of the firm: Managerial behaviour, agency costs and ownership structure. Journal of Financial Economics, 3(4), 305- 360.
Jensen, M.C., & Murphy, K., 1990. Performance Pay and Top-Management Incentives. Journal of Political Economy, 98(2), 225-264.
John, K, Litov, L., & Yeung, B., 2008. Corporate governance and risk-taking. Journal of Finance, 63, 1679–1728.
Kreiser, P, Marino, L.D., Dickson, P. & Mark W.K., 2010. Cultural influences on
38
Leszczensky, L. & Wolbring, T., 2019. How to Deal With Reverse Causality Using Panel Data? Recommendations for Researchers Based on a Simulation Study. Sociological Methods & Research, 1-29.
Lerner, J., & Wulf, J., 2007. Innovation and incentives: Evidence from corporate R&D. The Review of Economics and Statistics, 89(4), 634-644.
Li, Y. & Zahra, S.A., 2012. Formal institutions, culture, and venture capital activity: a cross-country analysis. Journal of Business Ventures, 27, 95–111.
Licht, A.N., Goldschmidt, C. & Schwartz, S.H., 2005. Culture, law, and corporate governance. International Review of Law and Economics. 25, 229–255.
Mihet, R., 2013. Effects of culture on firm risk-taking: A cross-country and cross industry analysis. Journal of Cultural Economics, 37(1), 109-151.
Sedikides, C., Gaertner, L., & Toguchi, Y., 2003. Pancultural self-enhancement. Journal of Personality and Social Psychology, 84, 60-79.
Shane, S., 1993. Cultural influences on national rates of innovation. Journal of Business Venturing, 8(1), 59-73.
Thompson, J.D., Scott, W.R., & Zald, M.N., 2009. Organizations in action: social science bases of administrative theory. Transaction Publishers, New Jersey, USA.
Triandis, H. C., Chen, X. P., & Chan, D. K. S., 1998. Scenarios for the measurement of collectivism and individualism. Journal of Cross-Cultural Psychology, 29, 275–289.
39
Yamaguch, S., Gelfand, M., Ohashi, M. M., & Zemba, Y., 2005. The cultural psychology of control: illusion of personal versus collective control in the U.S. and Japan. Journal of Cross-Cultural Psychology, 36, 750–761.
40
Appendices
Main Variables Definition Source
Dependent variables Std ROA (Corporate risk-taking proxy)
Std ROE (Corporate risk-taking proxy)
The deviation of the firm’s Net Income/Total Assets as to the country average (for the corresponding year), followed by calculating the standard deviation of this measure for each firm (for each firm with available earnings and total assets for at least 4 years in 2007 to 2016). Measured in millions of U.S. dollar ($). Winsorized at the one and 99th percentile level in both tails of the distribution.
The deviation of the firm’s Net Income/Total Equity as to the country average (for the corresponding year), followed by calculating the standard deviation of this measure for each firm (for each firm with available earnings and total assets for at least 4 years in 2007 to 2016). Measured in millions of U.S. dollar ($). Winsorized at the one and 99th percentile level in both tails of the distribution. Compustat Global Vantage Database Compustat Global Vantage Database Independent variable Ln Executive Compensation Moderating variables Ln Power Distance Ln Individualism
The natural logarithm of Executive
Compensation. Executive Compensation is the total sum of yearly salary plus cash and stock bonuses. Measured in millions of U.S. dollar ($). Winsorized at the one and 99th percentile level in both tails of the distribution.
The natural logarithm of the Power Distance Score. A country’s cultural dimension score is measured upon a scale from 0-100. The higher the score, the higher level of Power Distance in that country. Winsorized at the one and 99th percentile level in both tails of the distribution.
The natural logarithm of the Individualism Score. A country’s cultural dimension score is measured upon a scale from 0-100. The higher the score, the higher level of
41 Ln Uncertainty Avoidance Control variables Size Capex Leverage
The natural logarithm of the Uncertainty Avoidance Score. A country’s cultural dimension score is measured upon a scale from 0-100. The higher the score, the higher level of Uncertainty Avoidance in that country. Winsorized at the one and 99th percentile level in both tails of the distribution.
The natural logarithm of Total Assets. Measured in millions of U.S. dollar ($). Winsorized at the one and 99th percentile level in both tails of the distribution.
The ratio of Capital Expenditures over Total Assets. Measured in millions of U.S. dollar ($). Winsorized at the one and 99th percentile level in both tails of the distribution.
The ratio of Total Debts over Total Assets. Measured in millions of U.S. dollar ($). Winsorized at the one and 99th percentile level in both tails of the distribution.
Geert Hofstede Dimension Data Matrix Compustat Global Vantage Database Compustat Global Vantage Database Compustat Global Vantage Database
