EQUITY-BASED CEO COMPENSATION, RISK TAKING
AND FIRM FINANCIAL PERFORMANCE
Simona Augulyte (S2291886) University of Groningen Faculty of Economics and Business
Master Study Finance
Supervisor Boris van Oostveen 13 – 06 – 2016
ABSTRACT
This study examines the relationship between risk-taking of a CEO and firm financial performance by applying equity-based compensation of the CEO as an explanatory factor. This approach is relatively new as prior literature focuses more on the relationship between risk-taking and CEO compensation only. The study takes a step forward and examines how the risk-taking of a CEO, triggered by equity-based compensation, affects the financial performance of a firm. By analysing more than 1600 U.S. firms in the period of 1994-2014, I find that vega positively affects risk-taking while delta has a negative effect on it as opposed to the views of the agency theory. The increased risk-taking in turn diminishes the financial performance of a firm.
Keywords: Executive compensation, Equity incentives, Risk taking, Financial performance
2
I. INTRODUCTION
Risk-taking and how it influences firm performancehas long been a popular topic of a scientific literature for both strategic management and corporate finance. Particularly the behaviour of a Chief Executive Officer (CEO) regarding risk is widely analysed by previous literature. However, it is limited to the factors that increase the risk-taking behaviour of a risk-averse CEO. This study takes a step forward and asks a question: whether the risk-taking of a CEO contributes to a positive financial performance of a firm? By using an instrumental variable approach, this study connects a variable that significantly affects the behaviour of a CEO – equity-based compensation – with risk-taking and firm performance.
Figure I. Structure of average CEO compensation in the U.S., 1989-2011
The figure displays the market value of stock options and compensation in 2011 dollars. From 1989 until 1999 the sample comprises of 800 largest U.S. companies, from 2000 onwards the sample equals 500 largest U.S. companies. Sources: Forbes Annual Executive Compensation Reports; U.S. Bureau of Labour Statistics.
It is widely advised to use more equity-based compensation in order to align the incentives of a CEO and the shareholders. Scientific literature generally assumes that CEOs are more risk-averse than the shareholders of the firm as they cannot diversify their risks. Due to their risk aversion, CEOs might not always act at the best interest of the firm by rejecting projects that are relatively risky or taking decisions that are less risky (May, 1995). On the other hand, due to their risk-neutrality, shareholders would lose if only low-risk decisions are made as low risk often leads to low returns. In this case the value of shareholders is not maximized given their risk appetite. Since equity-based compensation connects the performance of a firm directly to the compensation of the manager (Eisenhardt, 1989), it is assumed to effectively solve the incentive alignment problem. As illustrated in Figure I, not only has the compensation of an average CEO of 500 largest U.S. firms increased significantly since 1989, so did the proportion
0 2 4 6 8 10 12 14 16 18 $ MIL
3
of stock options compared to cash compensation. This trend confirms the belief that equity-based compensation helps to align the incentives of a CEO and the shareholders of a firm. However, I argue that an increased risk-taking of a CEO, triggered by equity-based compensation, might not always be beneficial for a firm. On the contrary, for some firms excessive taking can be value-decreasing. For example, Singh (1986) claims that high risk-taking is detrimental to the financial performance of a firm. Nevertheless, previous literature lacks research on the relationship between equity-based compensation and firm performance. Furthermore, past research (Sanders & Hambrick (2007), Shen & Zhang (2013)) fails to fully address the endogeneity issue in the relationship between equity-based compensation, risk-taking and firm performance. Therefore, this study proposes a two-step approach which not only examines this relationship, but also accounts for endogeneity. The following steps are taken in the analysis:
1. Estimate the relationship between equity-based compensation and the risk-taking measures where equity-based compensation measures delta and vega are instruments for risk-taking measures, which are assumed to be endogenous.
2. Use the estimation results of step 1 and estimate how these risk-taking measures relate to the financial performance of a firm.
This study thus applies instrumental variable approach for approximately 1,630 non-regulated firms from the U.S. during a period from 1994 until 2014, creating a solid sample for the analysis.
4
compensation can encourage a risk-taking behaviour which is detrimental to a financial performance of a firm.
This study is contributing to the existing theory in a few aspects. First, it provides with a new way to analyse the relationship between the risk-taking of a CEO, equity-based compensation and financial performance. Agency theory only partially addresses this relationship while focusing solely on risk-taking and equity-based compensation. The results of this study indicate that it is important to fully examine the effects of equity-based compensation, thus investigating the effects of it on the firm performance as well. Second, the results indicate to the companies to be cautious when assigning equity-based compensation to their CEOs. More equity-based compensation can lead to substantial risk-taking of a CEO that would jeopardize the value of a firm.Third, an analysis applied in this study contributes to the existing literature due to a novel way of investigating the relationship between equity-based compensation, risk taking and firm performance. Instrumental variable approach is not new for analyses where the relationship between risk-taking and equity-based compensation is analysed. However, it is an innovative way to analyse the relationship between equity-based compensation and firm performance. The structure of the study is organized as follows. The next section discusses a relevant prior literature while reviewing the factors that affect CEO risk-taking and main arguments of agency theory. The third section elaborates on the methodology of the paper, where the instrumental variable approach is explained in detail. The fourth section discusses the data and descriptive analysis while the fifth section elaborates on the results. The final section presents the conclusions and deliberates on the limitations and suggestions for future research.
II. PREVIOUS LITERATURE
Risk taking behaviour of a CEO and firm performanceThe relationship between CEO risk-taking and firm performance is rather ambiguous as prior research disagrees on the direction of this relationship as well as the causality. On the one hand, poor performance can encourage risk-taking as stated by prospect theory. On the other hand, risk-aversion can follow a poor performance of a firm in order to quickly fix the situation. Furthermore, there is evidence for a reverse relationship as increased risk-taking affects the financial performance as well.
5
which implies higher risk aversion under satisfactory performance while risk taking increases under unacceptable performance. Singh (1986) bases his research on prospect theory which argues that CEO risk aversion depends on the way problems are framed (Wiseman & Gomez-Mejia, 1998). This theory defines the problems as positively framed if acceptable values are expected from the available choices while negatively framed problems lead to an expectation of unacceptable values. Kahneman & Tversky (1979) argue that risk seeking follows the situations where choices include sure losses because the managers are eager to gamble in order to improve a negative situation. On the other hand, when there is something to lose, risk-aversion follows.
However, there is evidence that poor firm performance encourages a CEO to be more risk-averse, contrary to the claims of prospect theory. Wu & Tu (2007) argue that a poor firm performance puts a pressure on the CEO to act immediately and fix the problem. This pressure forces the CEO to look for short-term actions instead of long-term activities, which leads to, for example, a reduction in Research and Development expenses (R&D). The opposite behaviour is observed when the firm performance is good and firm has slack resources. The CEO is not pressured anymore to focus on short-term remedies and thus can concentrate on long-term developments such as investment in R&D. Given contradicting conclusions of prior literature it is not clear what effect does a firm performance have on the risk-taking behaviour of a CEO.
Furthermore, the relation between performance and risk-taking of an executive appears to suffer from causality issues. Wiseman & Gomez-Mejia (1998) argue that the current wealth might only give a point of reference instead of directly prompting preferences for risk. Past successful choices of risky alternatives might have some influence in the risk-taking behaviour too. Bromiley (1991) elaborates on the causality by examining manufacturing companies. The author finds that poor performance of a firm triggers an increase in risk-taking, which in turn leads to a further poor performance. Bromiley (1991) argues that this process can be described as a vicious circle, where a bad performance worsens the situation more and more.
6
higher returns can be used as a buffer to lower the cost of bankruptcy. Based on this hypothesis, a better performance triggers an executive to increase her risk appetite since the leverage ratio increases. Contrary to Margaritis & Psillaki (2010), Bergen & Bonaccorsi di Patti (2006) find a reverse relationship to be statistically and economically significant. The authors propose that higher leverage is associated with higher firm performance for US banking industry.
It can be concluded, that the relationship between CEO risk-taking and firm performance is rather complex as previous literature does not provide any consensus. Therefore, the next section discusses prior literature on CEO-related factors that influence a particular risk-taking behaviour of a CEO which in turn affects the performance of a firm. The discussion then narrows down to the factor that is central to this analysis – equity-based compensation. Factors, affecting CEO risk-taking behaviour
One of the factors analysed is the power of the CEO. Lewellyn & Muller-Kahle (2012) argue that excessive risk-taking arises from the power of the CEO. Several variables are used to measure the power of the CEO, such as the independence of the board of directors, tenure and CEO duality – whether the CEO was also a chairperson of the board. The authors conclude that more powerful CEOs are able to make decisions they desire personally. Power, according to Lewellyn & Muller-Kahle (2012), affects the risk preferences for a CEO, thus it can lead to either an increase or a decrease in risk-taking. Additionally, Pathan (2009) states that CEO power, measured as an ability to influence board decisions, decreases risk-taking of a bank while presence of a strong board of directors is positively associated with bank risk-taking. However, Haleblian & Finkelstein (1993) claim that dominant CEOs in large firms perform worse in a volatile environment compared to a stable one. It implies that high CEO power can be detrimental for a firm high on risk. Regarding CEO power and firm performance, Daily & Johnson (1997) find that the relationship is positive and causal as not only the CEO power influences firm performance, but also vice versa. Nevertheless, Baliga, et al. (1996) find no evidence that CEO duality affects the firm performance. The authors claim that it is misleading to focus on a single element which can affect the financial performance of a firm which is highly complex. In conclusion, CEO power, being a variable that depends on a definition used, is somewhat ambiguous when explaining the risk-taking behaviour of a CEO.
7
Khan & Vieito (2013) indicate similar findings, as firm risk level in their sample is lower when the CEO is a female. In contrast, Adams & Funk (2012) claim that female directors are more risk loving than their male counterparts. According to these authors, the values and attitudes towards risk for executives differ significantly from the general population. Thus, increased risk-aversion does not necessarily arise from a CEO being a female. Additionally, Maxfield, et al. (2010) find evidence for gender neutrality in risk-aversion when making managerial decisions. The authors claim that female CEOs are not necessarily risk-averse while the main challenge for them is to actually take credit for risk-taking. Hence, inconsistent results regarding the gender of a CEO again indicate that another factor can be more important when determining her risk-taking behaviour.
CEO compensation structure is by far the most analysed factor that influences the risk-taking behaviour of the CEO. Therefore, equity-based compensation is considered as a strong instrumental variable for the risk-taking measures. The main theory behind the analysis of this study – agency theory – provides with theoretical insights supporting the usage of equity-based compensation as a factor explaining CEO risk-taking.
CEO risk-taking and agency theory
Agency theory revolves around two agent-principal problems: 1) how to align their goals and 2) how to verify that an agent, or a manager, is behaving in accordance to what is best for the firm (Eisenhardt, 1989). The second problem arises due to the differences in risk preferences between the agent and the principal. The agent, or the CEO, is perceived to be risk-averse while the shareholders of a firm, being the principal, are risk-neutral. This mismatch in risk preferences can be detrimental to a firm as the agent might act based on her own interests instead of those of the shareholders. May (1995) finds that CEOs consider their own risk when making firm-level decisions. According to the author, when a CEO has a large non-diversifiable share of her wealth in her firm equity, the acquisitions that this firm engages in are often of a diversifying nature, thus reducing personal risk of the CEO.
8
without limiting potential gains. In case a risky project fails, the CEO incurs no gains or losses with respect to stock options while a successful project can generate a substantial wealth to the CEO. As the CEO is averse, equity-based compensation is aimed at decreasing this risk-aversion. Thus, an increased risk-taking is perceived as a positive outcome caused by equity-based compensation since it implies that a CEO invests in a risky but positive Net Present Value (NPV) project.
9
compensation influencing risk-taking of the management team as opposed to the claims of the agency theory. Finally, Eisdorfer, et al. (2013) find that equity-based compensation induces over-investment, especially when the proportion of this type of compensation is large and the leverage of the compensation is lower than that of the firm. In order to reduce agency costs, compensation leverage should be aligned to the leverage of a firm. Thus only providing equity-based compensation does not eliminate agency issues alone.
Equity-based compensation and firm performance
According to agency theory, equity-based compensation should encourage a CEO to take more risk, however, literature falls short of further analysis whether this risk is value-adding. Coles, et al. (2006) analyse the relationship between managerial risk-taking and compensation, but do not evaluate what effect it has on firm performance. Similarly, Armstrong & Vashishtha (2012) find a significant relationship between the sensitivity of stock options to volatility and risk-taking of a manager, but fail to test what are the implications to the value of the firm. The authors only argue that there is a risk-value trade-off without providing any empirical evidence. Shen & Zhang (2013) analyse a sample of 843 cases and find that managers, having high-vega portfolios, are more likely to invest in R&D. However, these firms also exhibit lower operating performance and lower abnormal returns, implying that overinvestment in R&D can trigger a reduction of the firm value. Nevertheless, the authors only partially address endogeneity issue by using excess delta and vega while relying on Fama–French three-factor model as well as the Carhart four-factor model. For this reason the results should be treated with caution as the methodology focuses on zero-investment portfolio approach while not fully addressing the nature of the data. Furthermore, Sanders & Hambrick (2007) claim that when CEOs are having a substantial amount of their wealth in stock options, they take higher bets which lead to either high losses or high gains to the firm. Similarly to Shen & Zhang (2013), their methodology vaguely addresses the endogeneity issue. The authors control for endogeneity by including predicted values of CEO stock options regressed on firm and executive characteristics. However, the authors do not specify which characteristics are used to predict the CEO stock options. Therefore, it is not possible to verify if their model actually solves the endogeneity problem.
10
compensation, is value-adding, we should observe a positive relationship between the risk-taking of a CEO and firm financial performance. However, if this hypothesis is rejected, the empirical results of this study are supporting the view that risk-taking of a CEO, affected by equity-based compensation, is destroying the value of the company. The following section introduces the methodology of this study, which proposes an instrumental variable approach as an accurate method to analyse the relationship mentioned before.
III. METHODOLOGY
The aim of this study is to analyse to what extent the risk-taking of a manager, affected by the equity-based compensation, impacts the financial performance of a firm. More precisely, risk-taking is reflected through managerial decisions, such as investment policy or capital structure. I investigate whether the managerial decisions, being a proxy for risk-taking, increase the financial performance of a firm. The equation below specifies the initial relationship investigated, which will be used as a benchmark for subsequent models:
𝑃𝑒𝑟𝑓𝑜𝑟𝑚𝑎𝑛𝑐𝑒𝑖,𝑡 = 𝛼𝑖 + 𝛽1𝐷𝑒𝑐𝑖𝑠𝑖𝑜𝑛𝑖,𝑡+ 𝛽2𝐶𝑜𝑛𝑡𝑟𝑜𝑙𝑠𝑖,𝑡+ 𝜀𝑖,𝑡 (1)
Where Performance is a proxy for firm financial performance, Decision is a proxy for risk-taking of a CEO, Controls – a proxy for control variables, i is an index for a firm, t – an index for time.
11
Therefore, the instrumental variable analysis provides more accurate results than the fixed effects model.
This study argues that equity-based compensation affects the risk-taking behaviour of a CEO, while not affecting the financial performance directly. Therefore I argue that equity-based compensation measures are proper instruments to use in the instrumental variable approach. Previous research supports the lack of a direct relationship between firm financial performance and equity based compensation (Kerr & Bettis (1987), Bennett, et al. (2015). On the other hand, the relationship between risk-taking of a CEO and equity-based compensation is widely analysed. Low (2009), Hagendorff & Vallascas (2011), Armstrong & Vashishtha (2012) find evidence that vega positively affects taking of a CEO. Delta is also found to affect risk-taking significantly (Coles, et al. (2006), Knopf, et al. (2002)), although prior literature fails to agree whether this relationship is positive or negative. To conclude, equity-based compensation can be used as an appropriate instrument to investigate the relationship between risk-taking and firm financial performance.
In the first stage the relationship between equity-based compensation, used as instrumental variable, and the managerial decisions has to be estimated based on the following equation:
𝐷𝑒𝑐𝑖𝑠𝑖𝑜𝑛𝑖,𝑡 = 𝛼𝑖+ 𝛽1𝐼𝑛𝑠𝑡𝑟𝑢𝑚𝑒𝑛𝑡𝑠𝑖,𝑡+ 𝛽2𝐶𝑜𝑛𝑡𝑟𝑜𝑙𝑠𝑖,𝑡 + 𝜀𝑖,𝑡 (2)
Where Decision is a proxy for risk-taking of a CEO, Instruments is a proxy for instrumental variables, Controls – a proxy for control variables, i is an index for a firm, t – an index for time. The fitted values of Decision are then used in the second stage in order to measure how these managerial decisions affect the financial performance of a firm, where Model (1) structure is used.
IV. DATA AND DESCRIPTIVE STATISTICS
Sample selection
12
(R&D) and Capital Expenditures (CAPEX) are gathered from Compustat database. Finally, monthly stock return data and S&P index return data is gathered from the Center for Research in Security Prices (CRSP). Following the prior literature, the sample is restricted to non-regulated industries only. Thus, firms with SIC codes 6000-6799 as well as 4900, which include financial institutions and utility service providers, are removed from the sample. The final sample consists of 1,629 companies from 55 industries, industry distribution can be found in Table A.III in Appendix.
Variable Definitions
Firm financial performanceAs the financial performance of a firm can be defined based on accounting and market data, it is important to reflect the performance from both perspectives in order to account for measurement differences. For this reason return on equity (ROE) is measured as net income divided by book value of equity as well as by the market value of equity. As an additional measure of financial performance, return on assets (ROA) is also included.
Risk-taking measures
In order to measure the risk-taking of a CEO, managerial decisions such as investment and capital structure policy, are taken as a proxy. R&D and CAPEX are chosen as measures for investment policy while leverage level reflects capital structure policy. R&D variable is equal to research and development expenditures scaled by total assets. CAPEX equals capital expenditures less sale of property, plant and equipment scaled by total assets. Leverage is equal to the book values of total liabilities divided by the total assets. The selection of these variables as measures for risk-taking is based on previous literature. For example, Singh (1986) measures risk-taking by a questionnaire in which top managers are asked to rate various decisions. Debt-financing, heavy R&D and reliance on innovation are regarded as decisions increasing risk-taking. Gaver & Gaver (1995) use R&D expense to assets as one of the measures for investment opportunity set. Moreover, Coles, et al. (2006) as well as Sanders & Hambrick (2007) use R&D and CAPEX expenditures as measures for investment policy.
13
R&D expenditures. Coles, et al. (2006) agree that increasing expenditures in intangible assets, such as R&D, implies increased risk. Regarding leverage level, Jensen (1986) argues that increased leverage leads to an increase in such agency costs of debt as bankruptcy costs. Hence, increased level of leverage implies an increase in risk-taking.
Instruments
As the aim of this study is to measure how the risk-taking by CEO, affected by the equity-based compensation, in turn affects the financial performance of a firm, vega and delta are used as instruments. Delta and vega are estimates of the stock-based compensation sensitivity to the changes in the price and volatility of a stock respectively. To measure stock option incentives for the CEO, these variables are calculated based on the methodology used in the prior literature (Core & Guay (1999), Coles, et al. (2006), Armstrong & Vashishtha (2012), Aggarwal & Samwick (2006), Brockman, et al. (2010)). Delta is defined as a dollar change in the risk-neutral value of the CEO’s equity portfolio of stock options for a 1% change in the price of the underlying stock. Similarly, vega is a dollar change in the risk-neutral value of the CEO’s equity portfolio of stock options for a 0.01 change in the annualized standard deviation of the return of the underlying stock. In order to calculate these variables, a widely used Black-Scholes (1973) option-pricing model, adjusted for dividends (Merton, 1973), is implemented and is described in detail in Appendix I.
14
managers to decrease risk-taking. The authors also find that hedging activities of firms positively relate to the sensitivity to stock price or delta, thus providing evidence that delta actually decreases the risk appetite of a manager. Therefore, the expected sign between delta and risk-taking variables is unclear while a strong relationship is expected. To sum up, prior literature implies that both vega and delta are strong candidates for instruments of managerial decisions, which are a proxy for risk-taking of the CEO.
Nonetheless, for vega and delta to be strong instruments, there should be no correlation between these variables and financial performance. Kerr & Bettis (1987) find strong evidence that there is no significant relationship between CEO pay and abnormal returns of a firm as well as overall market movements. Furthermore, Bennett, et al. (2015), when analysing bank holding companies, claim that delta and vega loose explanatory power when inside debt is added to the regression. It implies that inside debt is a crucial explanatory variable of the performance of the company instead of vega and delta. In addition, even if the relationship between vega, delta and performance is statistically significant, the economic significance is rather modest. Jensen & Murphy (1990) find that for every $1,000 increase in shareholder wealth, CEO wealth increases by only $3.25, suggesting that the relationship between financial performance of the firm and equity-based compensation of the CEO is rather weak.
To summarize, both delta and vega are strongly supported by previous literature as proper instruments for risk-taking of a CEO. Additionally, various tests are performed to confirm empirically the validity of these instruments, such as Sanderson & Windmeijer test of excluded instruments, Hansen-J statistic test for overidentification and Kleibergen-Paap test for weak model identification. The outcomes of these tests are reported in the results.
Control variables
To control for CEO-specific characteristics, such variables as gender, age, tenure and cash compensation are included in the model. These variables serve as proxies for CEO risk-aversion.
15
risky investments as retirement age approaches. Similar reasoning applies to tenure; however, longer tenure is also associated with entrenchment and ability to dominate boards of directors (Hill & Phan, 1991). Hence, higher tenure can indicate a higher CEO power to implement decisions that are more beneficial for her rather than the shareholders of the firm.
Cash compensation of a CEO is equal to the sum of the base salary and bonus. In order to minimize the possibility of outliers biasing the results of the analysis, a natural logarithm is taken. Higher cash compensation suggests a better diversification of a CEO as she can invest more outside the firm (Guay, 1999), thus implying a lower risk-aversion. However, entrenchment is also more likely when cash compensation is higher resulting in a higher risk-aversion (Coles, et al., 2006). Therefore, the sign of cash compensation remains ambiguous. To control for firm-specific characteristics, firm size, growth opportunities and risk are included in the analysis as control variables. Firm size is measured by the natural logarithm of the total assets in order to minimize the effect of extreme values. A positive relation with firm financial performance is expected as larger firms can employee more talented and qualified CEOs (Baker, et al., 1988).
To account for growth opportunities, Tobin’s Q is used as a proxy. Tobin’s Q is equal to the market value of the company divided by total asset value or the replacement value. High-growth firms are thought to have higher Tobin’s Q. This ratio is commonly used as a proxy for growth opportunities (Martin (1996), Opler & Titman (1993)). The relation between growth opportunities and financial performance is rather unclear since prior research often uses Tobin’s Q as a firm performance measure (Aggarwal & Samwick (2006), Mehran (1995)). Moreover, there is no consensus in the previous literature whether high growth leads to higher profitability. Gartner (1997) claims that, due to limited managerial capabilities, firm growth can be value-decreasing. MacMillan & Day (1987) state that fast growth can lead to high profitability given the fact that new firms successfully generate profits when aggresively and rapidly entering the market. Hence, the relationship between firm growth opportunities and financial performance remains rather ambigious.
16
of S&P 500 index is used as the return of the market portfolio. McGuire, et al. (1988) argue that risk and financial performance of a firm defers per industry. Furthermore, Amit & Wernerfelt (1990) state that reduced firm risk is value-creating as it enables firms to operate more efficiently and increase cash flows. Therefore, a negative relationship between firm risk and financial performance is expected.
A full list of the definitions of the variables and their sources can be found in Table A.II. in the Appendix.
Descriptive statistics and correlations
Table I represents the main findings of the descriptive statistics of the variables used in the empirical analysis. In order to avoid outliers influencing the results, vega, delta, cash compensation, ROA, ROE (book) and ROE (market) are winsorized at the 1st and 99th percentile. We can see that delta is much larger in value compared with vega, thus implying that the value of the equity portfolio of a CEO in this sample fluctuates much more due to changes in the price rather than volatility of the underlying stock. Moreover, both vega and delta are highly skewed, therefore, a natural logarithm transformation is used in the analysis to scale down the magnitude of the values of both variables.
Additionally, pairwise correlations are estimated between the variables and reported in Table II in order to analyse possible relationships between variables before the regression analysis. Not surprisingly there is a strong positive correlation between vega and delta equal to 0.82. There is a negative correlation between both equity-based compensation measures and CAPEX expenditures. It suggests that increasing sensitivity of the CEO stock options due to a change in the price and volatility of an underlying stock leads to a reduced expenditure in CAPEX or vice versa. It would imply that increased option portfolio sensitivity to price and volatility of the stock leads to an increased risk-taking since decreasing CAPEX expenditures imply increasing risk-taking. On the other hand, R&D expenditures also correlate negatively with both vega and delta while being a proxy for increased risk-taking. However, leverage correlates positively with both equity-based compensation measures, thus implying that risk-taking and equity-based compensation have a positive association.
17
Particularly R&D expenditures exhibit a relatively high negative correlation (-0.33) with ROA as well as with market value of ROE (-0.23). On the other hand, CAPEX correlates positively with firm financial performance measures, thus adding more support to the hypothesis that riskier decisions might be not value-adding.
Please note that the regression analysis is not likely to suffer from multicollinearity problems as there are no perfect correlations.
Table I. Descriptive statistics for a sample of U.S. firms, 1994-2014
The table reports descriptive statistics of a sample of 1629 non-regulated firms with variables grouped according to CEO
characteristics, Risk-taking measures, Financial performance measures and Firm characteristics, where data is retrieved from
Execucomp database for CEO compensation and from Compustat and CRSP databases for firm-level characteristics. Delta is a natural logarithm of a dollar change in the risk-neutral value of the CEO’s equity portfolio of stock options for a 1% increase in the price of the underlying stock. Vega is a natural logarithm of a dollar change in the risk-neutral value of the CEO’s equity portfolio of stock options for a 0.01 increase in the standard deviation of the return of the underlying stock. Cash compensation equals the natural logarithm of the sum of salary and bonus of a CEO. Tenure is equal to a number of years a CEO held her office rounded to the nearest whole number, Gender is an indicator variable equal to one if a CEO is a female and zero otherwise, Age is measured in years and indicates the age of a CEO. R&D is research and development expenditures scaled by total assets. CAPEX equals net capital expenditures scaled by total assets. Leverage is the book value of total liabilities scaled by total assets. ROA is net income scaled by total assets. ROE (book) is equal to net income scaled by the book value of equity,
ROE (market) is equal to net income scaled to the market value of equity. Size equals the natural logarithm of the total assets
of the firm. Tobin’s Q is a measure for firm growth opportunities. Systematic risk is equal to the slope of a Market Model Regression where the monthly return of a company’s stock is regressed against the monthly return of S&P 500 index, Total
risk is the annualized standard deviation of monthly stock returns over past 60 months.
Mean Std. dev. Min Max Obs.
CEO characteristics Delta (ln) 5.00 1.40 1.38 8.17 13307 Vega (ln) -0.72 1.71 -5.93 2.70 13307 Cash compensation (ln) 1201.78 989.47 190.62 6143.24 13307 Tenure (years) 6.96 6.83 0.00 52.00 12673 Gender 0.02 0.14 0.00 1.00 13307 Age (years) 55.31 7.14 29.00 86.00 12955 Risk-taking measures R&D 0.04 0.07 0.00 2.09 13307 CAPEX 0.06 0.06 -0.35 1.51 13307 Leverage 0.21 0.19 0.00 3.68 13307
Financial performance measures
18
Table II. Pairwise correlation matrix for a sample of the U.S. firms, 1994-2014
The table reports pairwise correlations for the variables used in the empirical analysis, where data is retrieved from Execucomp database for CEO compensation and from Compustat database for firm-level characteristics. Delta is a natural logarithm of a dollar change in the risk-neutral value of the CEO’s equity portfolio of stock options for a 1% increase in the price of the underlying stock. Vega is a natural logarithm of a dollar change in the risk-neutral value of the CEO’s equity portfolio of stock options for a 0.01 increase in the standard deviation of the return of the underlying stock. Cash compensation equals the natural logarithm of the sum of salary and bonus of a CEO. Tenure is equal to a number of years a CEO held her office rounded to the nearest whole number, Gender is an indicator variable equal to one if a CEO is a female and zero otherwise, Age is measured in years and indicates the age of a CEO. R&D is research and development expenditures scaled by total assets. CAPEX equals net capital expenditures scaled by total assets. Leverage is the book value of total liabilities scaled by total assets. ROA is net income scaled by total assets. ROE (book) is equal to net income scaled by the book value of equity, ROE (market) is equal to net income scaled to the market value of equity. Size equals the natural logarithm of the total assets of the firm. Tobin’s Q is a measure for firm growth opportunities. Systematic risk is equal to the slope of a Market Model Regression where the monthly return of a company’s stock is regressed against the monthly return of S&P 500 index, Total risk is the annualized standard deviation of monthly stock returns over past 60 months.
19
V. EMPIRICAL RESULTS
Table III shows the results of the OLS estimations for all dependent variables. Columns 1, 2 and 3 report the results where ROA is a dependent variable. In columns 4, 5 and 6 the dependent variable is book value of ROE while in columns 7, 8 and 9 it is market value of ROE. For the sake of consistency, this structure holds for the whole analysis. The standard errors are robust to heteroscedasticity.
We can see from Table III that all risk-taking measures are statistically significant at 1%. CAPEX is positively affecting the financial performance, whereas both R&D and leverage affect the firm performance negatively. These results suggest that risk-increasing decisions are jeopardizing the value of the firm, thus providing with a support for the findings of Singh (1986).
In order to determine whether the models suffer from omitted variables bias, firm- and time-fixed effects are used. The results of this analysis are shown in Table IV. The analysis reveals that signs of all risk-taking measures are consistent regardless of the dependent variable while also being statistically significant at 1%. Both R&D and leverage negatively affect the financial performance of a firm while the relationship between CAPEX and financial performance is positive. Furthermore, the values of coefficients for risk-taking measures are reasonable from an economic perspective. For example, an increase of R&D variable by 1 unit leads to a decrease in ROA variable by 47.09 percentage points. Similar decrease in both market and book ROE is observed. The impact of leverage on financial performance is smaller; however, it is still relatively economically significant. Consequently, the results suggest that an increase in risk-taking decreases the financial performance of a firm.
20
Table III. Firm performance and risk-taking for a sample of the U.S. firms, 1994-2014
The table reports the results of OLS estimation for a sample of 1629 non-regulated firms, where data is retrieved from Execucomp database for CEO compensation and from Compustat and CRSP databases for firm-level characteristics. Columns 1, 2 and 3 report the results where ROA is a dependent variable, while in columns 4, 5 and 6 (7, 8 and 9) the dependent variable is book value of ROE (market value of ROE). ROA is net income scaled by total assets. ROE (book) is equal to net income scaled by the book value of equity, while ROE (market) is equal to net income scaled to the market value of equity. R&D is research and development expenditures scaled by total assets. CAPEX equals net capital expenditures scaled by total assets. Leverage is the book value of total liabilities scaled by total assets. Cash compensation equals the natural logarithm of the sum of salary and bonus of a CEO. Tenure is equal to a number of years a CEO held her office rounded to the nearest whole number, Gender is an indicator variable equal to one if a CEO is a female and zero otherwise, Age is measured in years and indicates the age of a CEO. Size equals the natural
logarithm of the total assets. Tobin’s Q is a measure for firm growth opportunities. Systematic risk is equal to the slope of a Market Model Regression where the monthly return of a company’s stock is regressed against the monthly return of S&P 500 index, Total risk is the annualized standard deviation of monthly stock returns over past 60 months. The robust standard errors are displayed in parentheses under the coefficients. ***, **, and * denote significance at the 0.01, 0.05, and 0.10 levels, respectively.
ROA ROE (book) ROE (market)
21
Table IV. Firm performance and risk-taking for a sample of the U.S. firms, 1994-2014
The table reports the results of firm- and time-fixed effects estimation for a sample of 1629 non-regulated firms, where data is retrieved from Execucomp database for CEO compensation and from Compustat and CRSP databases for firm-level characteristics. Columns 1, 2 and 3 report the results where ROA is a dependent variable, while in columns 4, 5 and 6 (7, 8 and 9) the dependent variable is book value of ROE (market value of ROE). ROA is net income scaled by total assets. ROE (book) is equal to net income scaled by the book value of equity, while ROE (market) is equal to net income scaled to the market value of equity. R&D is research and development expenditures scaled by total assets. CAPEX equals net capital expenditures scaled by total assets.
Leverage is the book value of total liabilities scaled by total assets. Cash compensation equals the natural logarithm of the sum of salary and bonus of a CEO. Tenure is equal to a number of years
a CEO held her office rounded to the nearest whole number, Gender is an indicator variable equal to one if a CEO is a female and zero otherwise, Age is measured in years and indicates the age of a CEO. Size equals the natural logarithm of the total assets. Tobin’s Q is a measure for firm growth opportunities. Systematic risk is equal to the slope of a Market Model Regression where the monthly return of a company’s stock is regressed against the monthly return of S&P 500 index, Total risk is the annualized standard deviation of monthly stock returns over past 60 months. The robust standard errors are displayed in parentheses under the coefficients. ***, **, and * denote significance at the 0.01, 0.05, and 0.10 levels, respectively.
ROA ROE (book) ROE (market)
[1] [2] [3] [4] [5] [6] [7] [8] [9] R&D -47.090*** -46.867*** -41.161*** (1.873) (7.642) (3.084) CAPEX 22.362*** 25.914*** 23.033*** (1.756) (7.021) (2.845) Leverage -13.672*** -9.106*** -16.648*** (0.646) (2.618) (1.054) Cash compensation (ln) 2.663*** 2.640*** 2.220*** 5.256*** 5.241*** 4.940*** 2.604*** 2.591*** 2.086*** (0.186) (0.190) (0.188) (0.758) (0.759) (0.763) (0.306) (0.308) (0.306) Tenure (years) 0.031 0.028 0.044** 0.070 0.065 0.079 0.001 -0.003 0.016 (0.019) (0.019) (0.019) (0.078) (0.078) (0.078) (0.031) (0.032) (0.031) Gender -0.940 -0.871 -0.782 -3.883 -3.825 -3.733 -1.817 -1.767 -1.668 (0.706) (0.721) (0.712) (2.880) (2.883) (2.884) (1.162) (1.168) (1.158) Age (years) -0.051*** -0.047*** -0.046*** 0.061 0.066 0.065 -0.060** -0.056* -0.055* (0.017) (0.018) (0.018) (0.071) (0.071) (0.071) (0.029) (0.029) (0.029) Size (ln) -0.473*** 0.597*** 0.766*** -1.871*** -0.790 -0.729 0.713** 1.666*** 1.893*** (0.175) (0.174) (0.172) (0.713) (0.697) (0.698) (0.288) (0.282) (0.281) Tobin’s Q 0.668*** 0.669*** 0.588*** 1.146*** 1.143*** 1.101*** 0.394*** 0.392*** 0.288*** (0.039) (0.040) (0.040) (0.160) (0.160) (0.161) (0.065) (0.065) (0.065) Systematic risk -0.621*** -0.505*** -0.552*** -1.580*** -1.459*** -1.509*** -0.673*** -0.565** -0.614*** (0.132) (0.135) (0.133) (0.538) (0.539) (0.539) (0.217) (0.218) (0.216) Total risk -2.994*** -2.921*** -2.722*** -4.924*** -4.855*** -4.712*** -4.369*** -4.311*** -4.072*** (0.338) (0.345) (0.341) (1.379) (1.381) (1.381) (0.557) (0.560) (0.556) Constant -1.756 -12.670 -7.002 -7.645 -18.963 -13.811 -8.481 -18.466 -12.066 (1.803) (1.825) (1.793) (7.355) (7.297) (7.266) (2.968) (2.957) (2.919)
F-statistic (H0: no fixed effects) 4.370*** 4.910*** 4.990*** 2.070*** 2.110*** 2.120*** 2.290*** 2.410*** 2.430***
F-statistic 66.560*** 48.500*** 59.480*** 11.770*** 10.920*** 10.870*** 28.180*** 24.070*** 30.790***
Within R2 0.153 0.116 0.139 0.031 0.029 0.029 0.071 0.061 0.077
22
Instrumental variable approach
As the fixed effects model only controls for time-invariant characteristics and does not address the endogeneity issues, instrumental variable approach is applied to the analysis while using time- and firm-fixed effects. Moreover, this approach reduces the possibility of spurious results and isolates the effect of risk-taking measures on financial performance. Equity-based compensation measures vega and delta are instruments for risk-taking variables which are treated as endogenous variables. Natural logarithm values of vega and delta are used for the regressions as both variables are highly skewed.
Table V demonstrates the first-stage results (Model 2) from two-stage estimation of firm performance with R&D being an endogenous variable in columns 1, 4 and 7, while in columns 2, 5 and 8 (3, 6 and 9) the endogenous variable is CAPEX (Leverage). In the R&D equation both delta and vega are statistically significant at 1%. Vega is positively affecting R&D with coefficient being equal to 0.003. Although the coefficient of delta has a similar magnitude, the sign of it is opposite, implying that increasing option portfolio sensitivity to stock price reduces investments in R&D. These results are partially contradicting the predictions of agency theory which states that stock options should increase risk-taking of a CEO. In this case both coefficients of delta and vega should be positive (Coles, et al. (2006), Armstrong & Vashishtha (2012)). Very similar results are observed in equations where leverage is the endogenous variable as increasing option portfolio sensitivity to stock price decreases the risk-taking of CEO. It has to be noted that vega is not statistically significant in these equations while the coefficient of it has a positive sign. The reason for opposite signs of the coefficients for delta and vega can be explained by the convex payoff structure of options. Carpenter (2000) argues that increased volatility of a firm’s option also increases the value of the stock option portfolio of the CEO based on the convexity of an option. This relationship in turn should lead to a greater risk loving. However, sensitivity to stock price illustrates a direct relationship between option value and stock movements whereby the wealth of the CEO directly depends on the stock price of her firm. According to Carpenter (2000), this connection increases an incentive to avoid risk instead of encouraging it. This explanation implies that the risk-taking behaviour of a CEO is more complex than agency theory suggests.
23
is no significant relationship between CAPEX and equity-based compensation measures as can be seen from columns 2, 5 and 8. Moreover, vega is significant only when R&D is an endogenous variable. It has to be noted that, despite the statistical significance in most of the estimations, the magnitude of the coefficients of vega and delta is relatively low, thus the economic significance of these variables is rather weak. For example, when leverage is an endogenous variable, a 1% increase in the sensitivity to the stock price would lead to approximately a 0.00015 unit decrease in leverage level. Therefore, the influence of equity-based compensation on risk-taking measures when controlling for firm- and time-effects is relatively moderate.
24
Table V. First-stage regressions: risk-taking and equity-based compensation for a sample of the U.S. firms, 1994-2014
The table reports the results from the first stage instrumental variables estimations with firm- and time-fixed effects. The regression is illustrated by Model 2 for a sample of 1629 non-regulated firms, where data is retrieved from Execucomp database for CEO compensation and from Compustat and CRSP databases for firm-level characteristics. Columns 1, 2 and 3 report the results where ROA is a dependent variable in the second-stage, while in columns 4, 5 and 6 (7, 8 and 9) the dependent variable in the second-stage is book value of ROE (market value of ROE). ROA is net income scaled by total assets. ROE (book) is equal to net income scaled by the book value of equity, while ROE (market) is equal to net income scaled to the market value of equity. R&D is research and development expenditures scaled by total assets. CAPEX equals net capital expenditures scaled by total assets. Leverage is the book value of total liabilities scaled by total assets.
Delta is the natural logarithm of a dollar change in the risk-neutral value of the CEO’s equity portfolio of stock options for a 1% change in the price of the underlying stock option. Vega is the
natural logarithm of a dollar change in the risk-neutral value of the CEO’s equity portfolio of stock options for a 0.01 change in the standard deviation of the return of the underlying stock option.
Cash compensation equals the natural logarithm of the sum of salary and bonus of a CEO. Tenure is equal to a number of years a CEO held her office rounded to the nearest whole number, Gender
is an indicator variable equal to one if the CEO is a female and zero otherwise, Age is measured in years and indicates the age of a CEO. Size equals the natural logarithm of the total assets of the firm. Tobin’s Q is a measure for firm growth opportunities. Systematic risk is equal to the slope of a Market Model Regression where the monthly return of a company’s stock is regressed against the monthly return of S&P 500 index, Total risk is the annualized standard deviation of monthly stock returns over past 60 months. The robust standard errors are displayed in parentheses under the coefficients. ***, **, and * denote significance at the 0.01, 0.05, and 0.10 levels, respectively.
R&D CAPEX Leverage R&D CAPEX Leverage R&D CAPEX Leverage
[1] [2] [3] [4] [5] [6] [7] [8] [9] Delta (ln) -0.004*** 0.002 -0.015*** -0.004*** 0.002 -0.015*** -0.004*** 0.002 -0.015*** (0.001) (0.001) (0.003) (0.001) (0.001) (0.003) (0.001) (0.001) (0.003) Vega (ln) 0.003*** 0.001 0.002 0.003*** 0.001 0.002 0.003*** 0.001 0.002 (0.001) (0.001) (0.003) (0.001) (0.001) (0.003) (0.001) (0.001) (0.003) Cash compensation (ln) 0.002* -0.003 -0.023*** 0.002* -0.003 -0.023*** 0.002* -0.003 -0.024*** (0.001) (0.002) (0.003) (0.001) (0.002) (0.003) (0.001) (0.002) (0.003) Tenure (years) 0.000 0.000 0.001*** 0.000 0.000 0.001*** 0.000 0.000 0.001*** (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) Gender -0.003 0.003 -0.002 -0.003 0.003 -0.002 -0.003 0.004 -0.001 (0.003) (0.003) (0.013) (0.003) (0.003) (0.013) (0.003) (0.003) (0.013) Age (years) 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) Size (ln) -0.020*** -0.006*** 0.026*** -0.020*** -0.006*** 0.026*** -0.020*** -0.006*** 0.026*** (0.002) (0.002) (0.004) (0.002) (0.002) (0.004) (0.002) (0.002) (0.004) Tobin’s Q 0.000 0.001* -0.006*** 0.000 0.001* -0.006*** 0.000 0.001* -0.006*** (0.000) (0.000) (0.002) (0.000) (0.000) (0.002) (0.000) (0.000) (0.002) Systematic risk -0.002 -0.002** 0.000 -0.002 -0.002** 0.000 -0.002 -0.002** 0.000 (0.001) (0.001) (0.003) (0.001) (0.001) (0.003) (0.001) (0.001) (0.003) Total risk 0.003 0.003 0.015 0.003 0.003 0.015 0.004 0.003 0.016 (0.003) (0.003) (0.010) (0.003) (0.003) (0.010) (0.003) (0.003) (0.010) Sanderson-Windmeijer F statistic 11.730 6.120 20.050 11.730 6.110 20.030 11.870 6.150 19.470
Cragg-Donald Wald F statistic 13.390 9.360 38.720 13.400 9.340 38.690 13.610 9.480 36.820
Anderson-Rubin Wald test 105.190 105.190 105.190 28.930 28.930 28.930 95.870 95.870 95.870
25
Table VI. Second-stage regressions: risk-taking on firm performance for a sample of the U.S. firms, 1994-2014
The table reports the results from the second stage instrumental variables estimations with firm- and time-fixed effects. The regression is illustrated by Model 1 for a sample of 1629 non-regulated firms, where data is retrieved from Execucomp database for CEO compensation and from Compustat and CRSP databases for firm-level characteristics. Columns 1, 2 and 3 report the results where ROA is a dependent variable in the second-stage, while in columns 4, 5 and 6 (7, 8 and 9) the dependent variable in the second-stage is book value of ROE (market value of ROE). ROA is net income scaled by total assets. ROE (book) is equal to net income scaled by the book value of equity, while ROE (market) is equal to net income scaled to the market value of equity. R&D is research and development expenditures scaled by total assets. CAPEX equals net capital expenditures scaled by total assets. Leverage is the book value of total liabilities scaled by total assets.
Cash compensation equals the natural logarithm of the sum of salary and bonus of a CEO. Tenure is equal to a number of years a CEO held her office rounded to the nearest whole number, Gender
is an indicator variable equal to one if a CEO is a female and zero otherwise, Age is measured in years and indicates the age of a CEO. Size equals the natural logarithm of the total assets. Tobin’s
Q is a measure for firm growth opportunities. Systematic risk is equal to the slope of a Market Model Regression where the monthly return of a company’s stock is regressed against the monthly
return of S&P 500 index, Total risk is the annualized standard deviation of monthly stock returns over past 60 months. The robust standard errors are displayed in parentheses under the coefficients. ***, **, and * denote significance at the 0.01, 0.05, and 0.10 levels, respectively.
ROA ROE (book) ROE (market)
[1] [2] [3] [4] [5] [6] [7] [8] [9] R&D -618.650*** -1185.377*** -943.795*** (131.132) (310.696) (208.216) CAPEX 700.558*** 1421.111*** 1254.374*** (198.455) (444.726) (363.210) Leverage -153.354*** -305.920*** -273.873*** (22.388) (60.527) (46.865) Cash compensation (ln) 3.445*** 3.900*** -1.646** 6.812*** 7.832*** -3.276 3.824*** 4.881** -5.055*** (0.589) (1.090) (0.791) (1.375) (2.397) (2.056) (0.930) (2.004) (1.528) Tenure (years) -0.001 -0.159 0.145*** 0.007 -0.319 0.295** -0.049 -0.340* 0.211** (0.087) (0.099) (0.053) (0.169) (0.222) (0.135) (0.135) (0.180) (0.100) Gender -2.563 -2.838 -0.533 -7.114 -7.873 -3.205 -4.391 -5.343 -1.170 (1.654) (1.970) (2.001) (4.927) (5.619) (4.947) (2.947) (3.853) (3.689) Age (years) -0.067 0.044 -0.011 0.029 0.253 0.140 -0.084 0.110 0.002 (0.091) (0.076) (0.049) (0.178) (0.172) (0.128) (0.141) (0.137) (0.090) Size (ln) -12.200*** 3.745** 3.546*** -25.228*** 5.680 5.177** -17.839*** 7.387** 7.051*** (2.478) (1.762) (0.819) (6.044) (4.034) (2.157) (3.966) (3.213) (1.499) Tobin’s Q 0.363 -0.051 -0.487* 0.537 -0.338 -1.184* -0.081 -0.914 -1.696 (0.323) (0.249) (0.258) (0.611) (0.563) (0.656) (0.353) (0.605) (0.783) Systematic risk -1.550** 0.674 -0.635 -3.434** 0.971 -1.684 -2.154** 1.566 -0.750 (0.646) (0.652) (0.457) (1.449) (1.519) (1.154) (1.013) (1.187) (0.800) Total risk -4.121** -3.535** -0.887 -7.173* -6.106* -0.807 -6.099** -5.455* -0.704 (1.771) (1.569) (1.380) (4.015) (3.609) (3.561) (2.878) (2.870) (2.352) F-statistic 6.130 3.300 6.440 4.090 2.650 4.050 3.240 1.550 3.200
Cragg-Donald Wald F statistic 13.395 9.362 38.721 13.395 9.343 38.690 13.607 9.485 36.822
Kleibergen-Paap Wald rk F statistic 11.730 6.121 20.045 11.734 6.110 20.033 11.875 6.146 19.469
Hansen J statistic 11.581 6.840 3.323 9.168 5.050 1.446 17.530 4.472 1.001
26
Robustness check
In order to investigate whether past risk-taking decisions influence the current financial performance more than the current decisions, the instrumental variable estimation is performed using lagged values of delta, vega and risk-taking measures. The results are reported in Table VII. Compared with Table V, the results imply the same conclusion – when sensitivity to stock price increases, CEO risk-taking decreases. A reverse relationship holds for option sensitivity to volatility, which induces a CEO to take riskier decisions. When R&D and leverage are endogenous variables, both delta and vega are statistically significant at 1%. The statistical and economic significance of both instrumental variables has improved compared to results with not lagged variables. It suggests that past stock option sensitivity to stock price and volatility has a stronger influence on the risk-taking of a CEO than the current sensitivity.
27
Table VII. First-stage regressions: lagged risk-taking and equity-based compensation for a sample of the U.S. firms, 1994-2014
The table reports the results from the first stage instrumental variables estimations with firm- and time-fixed effects. The regression is illustrated by Model 2 for a sample of 1629 non-regulated firms, where data is retrieved from Execucomp database for CEO compensation and from Compustat and CRSP databases for firm-level characteristics. Columns 1, 2 and 3 report the results where ROA is a dependent variable in the second-stage, while in columns 4, 5 and 6 (7, 8 and 9) the dependent variable in the second-stage is book value of ROE (market value of ROE). ROA is net income scaled by total assets. ROE (book) is equal to net income scaled by the book value of equity, while ROE (market) is equal to net income scaled to the market value of equity. R&D is lagged research and development expenditures scaled by total assets. CAPEX equals lagged net capital expenditures scaled by total assets. Leverage is the lagged book value of total liabilities scaled by total assets. Delta is a lagged natural logarithm of a dollar change in the risk-neutral value of the CEO’s equity portfolio of stock options for a 1% change in the price of the underlying stock option. Vega is a lagged natural logarithm of a dollar change in the risk-neutral value of the CEO’s equity portfolio of stock options for a 0.01 change in the standard deviation of the return of the underlying stock option. Cash compensation equals the natural logarithm of the sum of salary and bonus of a CEO. Tenure is equal to a number of years a CEO held her office rounded to the nearest whole number, Gender is an indicator variable equal to one if the CEO is a female and zero otherwise, Age is measured in years and indicates the age of a CEO. Size equals the natural logarithm of the total assets. Tobin’s Q is a measure for firm growth opportunities. Systematic risk is equal to the slope of a Market Model Regression where the monthly return of a company’s stock is regressed against the monthly return of S&P 500 index, Total risk is the annualized standard deviation of monthly stock returns over past 60 months. The robust standard errors are displayed in parentheses under the coefficients. ***, **, and * denote significance at the 0.01, 0.05, and 0.10 levels, respectively.
R&D CAPEX Leverage R&D CAPEX Leverage R&D CAPEX Leverage
[1] [2] [3] [4] [5] [6] [7] [8] [9] Delta (ln) -0.006*** 0.003*** -0.022*** -0.006*** 0.003*** -0.022*** -0.006*** 0.003*** -0.022*** (0.001) (0.001) (0.003) (0.001) (0.001) (0.003) (0.001) (0.001) (0.003) Vega (ln) 0.003*** 0.000 0.007*** 0.003*** 0.000 0.007*** 0.003*** 0.000 0.007*** (0.001) (0.001) (0.003) (0.001) (0.001) (0.003) (0.001) (0.001) (0.003) Cash compensation (ln) 0.001 -0.008*** -0.008** 0.001 -0.008*** -0.008** 0.001 -0.008*** -0.009** (0.001) (0.002) (0.003) (0.001) (0.002) (0.003) (0.001) (0.002) (0.003) Tenure (years) 0.000 0.000 0.001*** 0.000 0.000 0.001*** 0.000 0.000 0.001*** (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) Gender -0.003 0.005* -0.015 -0.003 0.005* -0.015 -0.003 0.005* -0.015 (0.002) (0.003) (0.014) (0.002) (0.003) (0.014) (0.002) (0.003) (0.014) Age (years) 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) Size (ln) -0.016*** -0.003 0.009* -0.016*** -0.003 0.009* -0.016*** -0.003 0.008* (0.003) (0.003) (0.005) (0.003) (0.003) (0.005) (0.003) (0.003) (0.005) Tobin’s Q 0.002** 0.000 -0.007*** 0.002** 0.000 -0.007*** 0.002** 0.000 -0.007*** (0.001) (0.000) (0.002) (0.001) (0.000) (0.002) (0.001) (0.000) (0.002) Systematic risk -0.005** -0.003** 0.002 -0.005** -0.003** 0.002 -0.005** -0.003** 0.002 (0.002) (0.001) (0.003) (0.002) (0.001) (0.003) (0.002) (0.001) (0.003) Total risk 0.011* 0.002 0.026*** 0.011* 0.002 0.026*** 0.010* 0.002 0.026*** (0.006) (0.005) (0.009) (0.006) (0.005) (0.009) (0.006) (0.005) (0.009) Sanderson-Windmeijer F statistic 10.950 6.930 36.530 10.960 6.930 36.530 10.630 6.870 36.890
Cragg-Donald Wald F statistic 22.380 11.210 55.300 22.400 11.200 55.300 21.650 11.130 55.790
Anderson-Rubin Wald test 37.350 37.350 37.350 17.850 17.850 17.850 22.660 22.660 22.660
28
Table VIII. Second-stage regressions: lagged risk-taking on firm performance for a sample of the U.S. firms, 1994-2014
The table reports the results from the second stage instrumental variables estimations with firm- and time-fixed effects. The regression is illustrated by Model 1 for a sample of 1629 non-regulated firms, where data is retrieved from Execucomp database for CEO compensation and from Compustat and CRSP databases for firm-level characteristics. Columns 1, 2 and 3 report the results where ROA is a dependent variable in the second-stage, while in columns 4, 5 and 6 (7, 8 and 9) the dependent variable in the second-stage is book value of ROE (market value of ROE). ROA is net income scaled by total assets. ROE (book) is equal to net income scaled by the book value of equity, ROE (market) is equal to net income scaled to the market value of equity. R&D is lagged research and development expenditures scaled by total assets. CAPEX equals lagged net capital expenditures scaled by total assets. Leverage is the lagged book value of total liabilities scaled by total assets. Cash compensation equals the natural logarithm of the sum of salary and bonus of a CEO. Tenure is equal to a number of years a CEO held her office rounded to the nearest whole number, Gender is an indicator variable equal to one if the CEO is a female and zero otherwise, Age is measured in years and indicates the age of a CEO. Size equals the natural logarithm of the total assets of the firm. Tobin’s Q is a measure for firm growth opportunities. Systematic risk is equal to the slope of a Market Model Regression where the monthly return of a company’s stock is regressed against the monthly return of S&P 500 index, Total risk is the annualized standard deviation of monthly stock returns over past 60 months. The robust standard errors are displayed in parentheses under the coefficients. ***, **, and * denote significance at the 0.01, 0.05, and 0.10 levels, respectively.
ROA ROE (book) ROE (market)
[1] [2] [3] [4] [5] [6] [7] [8] [9] R&D -323.000*** -776.402*** -344.629*** (77.863) (193.874) (99.174) CAPEX 244.604*** 754.239*** 148.499* (79.955) (256.696) (83.381) Leverage -65.871*** -175.371*** -57.990*** (10.819) (37.462) (14.803) Cash compensation (ln) 2.500*** 4.334*** 1.676*** 4.457*** 10.204*** 2.298* 2.445*** 3.474*** 1.666*** (0.518) (0.877) (0.341) (1.414) (2.732) (1.229) (0.643) (0.827) (0.491) Tenure (years) -0.035 -0.046 0.040 -0.057 -0.121 0.132 -0.093 -0.080 -0.020 (0.058) (0.048) (0.030) (0.150) (0.156) (0.107) (0.071) (0.049) (0.042) Gender -1.364 -1.858 -1.483 -4.117 -6.097 -4.607 -2.798 -2.790 -2.777 (1.175) (1.263) (1.266) (4.731) (5.207) (4.592) (2.199) (2.147) (2.208) Age (years) -0.037 -0.054 0.004 0.019 -0.033 0.128 -0.039 -0.049 -0.002 (0.060) (0.041) (0.032) (0.159) (0.141) (0.117) (0.073) (0.044) (0.045) Size (ln) -5.041*** 0.857 0.526 -15.221*** -0.810 -1.836 -4.157** 1.904** 1.691*** (1.650) (0.876) (0.422) (4.035) (2.630) (1.400) (1.975) (0.764) (0.570) Tobin’s Q 1.482*** 0.862*** 0.357** 2.796*** 1.269*** -0.052 1.052*** 0.419*** -0.039 (0.416) (0.213) (0.166) (0.923) (0.414) (0.470) (0.382) (0.133) (0.155) Systematic risk -1.927** 0.336 -0.253 -4.957** 0.957 -0.904 -2.002** 0.085 -0.244 (0.816) (0.399) (0.268) (2.032) (1.397) (0.993) (0.964) (0.452) (0.387) Total risk -1.143 -3.823*** -2.051** -1.467 -8.284** -3.333 -2.285 -4.801*** -3.372*** (1.849) (1.117) (0.840) (4.708) (3.937) (3.113) (2.288) (1.193) (1.154) F-statistic 9.610 8.580 13.230 4.070 3.660 5.240 5.260 7.050 8.070
Cragg-Donald Wald F statistic 22.384 11.205 55.297 22.397 11.201 55.305 21.655 11.131 55.794
Kleibergen-Paap Wald rk F statistic 10.955 6.926 36.526 10.960 6.927 36.526 10.635 6.870 36.889
Hansen J statistic 5.505 26.613 26.192 0.596 8.525 5.445 11.573 37.072 28.326
29
results of the second stage as R&D positively and significantly affects the financial performance while CAPEX and leverage negatively affect the performance. These contradictions imply that there are other factors that influence the risk-taking of the CEO. It could also mean that the level of risk between the decisions such as R&D and leverage is different and thus cannot be treated in the same way. Finally, the contradictions in the estimations without fixed effects confirm that within-firm factors influence the relationship between risk-taking and financial performance. Therefore, it can be concluded that the inclusion of fixed effects in the instrumental variable analysis provides more consistent results and there is a omitted variable bias which has to be accounted for.
As a final robustness check, the regressions are performed with raw values of delta and vega. The results are provided in Tables A.VI. and A.VII. in the Appendix. The effect of delta on risk-taking is not economically significant as the coefficients for delta are equal to 0.000 in all cases. Furthermore, delta is only statistically significant when leverage is the endogenous variable. In case of vega, the coefficients are approximately twice as small as their counterparts transformed with a natural logarithm, thus also implying a decreased economic significance. These findings suggest that natural logarithm is a proper transformation given the skewness of the variables. Similar findings regarding the association between risk-taking and performance are found as leverage has a negative relationship with the measures of the firm financial performance while CAPEX affects the financial performance positively.
VI. CONCLUSION
The purpose of this study is to investigate the relationship between equity-based compensation, risk-taking of a CEO and firm financial performance. Motivation for analysis of this topic stems from both theoretical and empirical perspectives. Agency theory argues that stock options should solve the agency problem of a mismatch between risk preferences of the CEO and the shareholders of a company. Stock options should induce a risk-averse CEO to take more risk, which in turn should benefit shareholders that are risk-neutral. Coles, et al. (2006), Armstrong & Vashishtha (2012) investigate this issue and find that the option sensitivity to stock volatility measured by vega encourages a CEO to take more risk. However, prior research falls short on investigating further whether the risk-taking, encouraged by equity-based compensation, is increasing firm financial performance.
30
additional insights about the effect of equity-based compensation on risk-taking and firm performance. Equity-based compensation is used as an instrumental variable to address the endogeneity issue and investigate to what extent risk-taking of a CEO is affected by stock options. Contrary to the findings of Armstrong & Vashishtha (2012), I find that vega positively affects the measures of risk-taking while delta affects them negatively. This outcome is corresponding to the insights provided by Carpenter (2000) who states that stock option convexity determines the opposite effects of vega and delta on risk-taking. While the value of an option increases when stock volatility increases, it induces a CEO to be less risk-averse. On the other hand, a direct relationship between the stock price and the value of the option does not encourage a CEO to be more risk-loving. That being said, this study illustrates a complex relationship between stock options and risk-taking behaviour of a CEO. Stock options do not necessarily encourage a CEO to become less risk-averse as opposed to what agency theory claims. If the sensitivity of the option to stock price is higher than sensitivity to stock volatility, stock options may even induce a CEO to increase risk-aversion.
31
compensation in stock options is more popular there rather than in Europe. However, analysing the trends in Europe or other continents can add further understandings to the analysis. Finally, this analysis focuses only on stock options as a factor determining the risk-taking of a CEO. There might be other important variables that can significantly influence the risk-taking behaviour of a CEO, especially given a low economic significance of vega and delta.
