1 | P a g e
Corporate performance and executive compensation: An empirical analysis of risk-taking incentive effect of stock-option compensation on corporate giants.
Bachelor thesis
University of Amsterdam
Supervisor: Aaron Kamm
Name: Sebastian Stercz ID: 10457666
Abstract
According to the theory, executive stock-option compensation (ESO) creates incentives for managers to behave in a way that increases shareholders’ wealth by maximizing market value of a company. Since value of options increases with increasing company’s volatility, it is vastly believed that ESOs tend to motivate managers to increase firm’s volatility which might lead to either value-enhancing or value-destroying effect on the company. In this academic article, the formerly mentioned relationships are examined using three-stage least square methodology in order to cope with simultaneous causality and endogeneity issues. The research revealed that the increasing proportion of ESO indeed increases the risk-taking of CEO. Furthermore, it appeared to be the case that the risk-taking incentive effect of ESO is positively related to corporate performance represented by long-term stock return.
Key words: Executive stock-option compensation, taking effect, corporate performance, risk-aversion, risk-tolerance, agency theory.
3| P a g e
1 Introduction
During the past 3 decades, the equity-based compensation policies have gradually become, by far, the largest component of CEOs’ compensation structure. Many believe that the substitution towards more equity-focused compensation policies was mainly driven by the desire to link the interests of shareholders with those of management (Sanders, G. and Hambrick C.,
D., 2007).
The increasing popularity and rapid
development of equity-based
compensations captured the interest of many academics. Consequently an extensive literature exists, which examines
the link between equity-based
compensation, in particular restricted stocks or stock-options, and the agency theory. Based on the vast diversity of empirical studies conducted, the evidence suggests that the shift in the compensation policies indeed ameliorate the alignment between shareholders’ and managerial interest (Hayes, M., R., Lemmon, M. and Qiu, M., 2012). However, in terms of providing incentives, options significantly differ from stocks. As pointed out by Yermack (1995), the pay-performance incentive effect for option-based compensation is substantially higher than that for stock-based compensation.
As a result option-based compensation may potentially provide top management, apart from beneficial incentives, with harmful incentives leading to excessive risk taking, high leverage or short-term stock value inflation. The case of Kevin Kalkhoven, then-CEO of JDS Uniphase (JDSU), proffers an exclusive example of detrimental effects of erroneous incentives provided by company’s significant option-based compensation policy. The CEO’s opportunistic behavior led to an artificial short-term inflation of company’s stock prices, allowing the CEO to cash out nearly $1.3 billion in option wealth. The latter was followed by the company’s announcement of $38.7 billion restatement sending the stock plunging. (Efendi, J., Files, R., Ouyang, B. and Swanson E., P., 2013). This latter evidence is further supported by academic paper written by Lippert and Porter (1997), in which they argued that options’ value is directly proportional to the firm’s volatility. This phenomenon creates substantial incentives for executives to involve themselves in excessive risk-taking which could hypothetically have a detrimental effect on company’s performance. Similarly, researchers such as Firth et al (2007) and Wu and Tu (2007), focused their studies on particular characteristics of equity-based
compensation and its relation with corporate policies. Others, including Florackis et al. (2009), researched the direct link between equity-based compensations and the effect on firm’s performance. However, these studies did not pay particular attention to the effects of stock-option compensation policies on the risk-taking behavior of top management, while examining the relationship between
stock-option compensation and
performance. Given that the investment risk is a significant component of company’s valuation, it is of utmost importance to integrate the impact of stock-option compensation policies on executives’ risk-taking behavior when estimating the link to corporate performance. (Chen, Y-R. and Ma, Y., 2010) Therefore the main purpose of this paper is to fill the gap in current literature by examining the relationship between option-based compensation and its effects on the company’s performance with respect to CEO risk-taking behavior. The analysis is focused on US companies listed under Russell top 200 index. An extensive dataset of executive compensations as well as relevant financial data is used as an input for a non-linear specification of three stage least squares regression (3SLS). The regression model is appropriately chosen to cope with structural system of simultaneous equation, given the issue of
endogeneity. In total, the analysis is based on 3 simultaneous equations, each containing dependent and independent variables as well as a number of specified control variables composed of adjusted financial data.
Obtained results suggest that there indeed exists positive and significant relation between stock-option grants and risk behavior of a CEO. In other words, increasing the proportion of options in managerial compensation seems to increase CEOs tendency towards higher risk-appetite. Interestingly enough, the analysis also suggests a significantly positive relationship between CEO’s ownership (excluding options) and his risk behavior. As a result, the theoretical concepts suggest that in this specific case the risk-taking incentives of CEO’s dominate the risk-aversion effect. Lastly given the Russell Top 200 index companies, it appeared to be the case that stock-option compensation is likewise significantly and positively related with company’s performance. The remainder of the paper is structured as follows: Section II provides the relevant theoretical concepts regarding the agency theory and its implications as well important findings obtained from previous research. Section III contains the detailed information on the data, empirical methodology, model specification and
5| P a g e
hypothesis formulation. Section IV reveals information about results from hypothesis testing and elaborates on the obtained
additional results. Finally Section V concludes the findings.
2 Theoretical framework
Essential features of Agency theory, stock–option compensation and risk-taking.
In the 1960’s and early 1970’s, economists such as Arrow, described the risk sharing problem as one that arises due to mismatch in risk attitude of cooperating parties. The concept later developed into well-known agency theory, which apart from risk sharing, comments on the potential implication of incongruity between managements’ and shareholders’ interests (agency problem) (Jensen C., M. and Murphy J., K., 1990). In particular, the theory suggests that the adequate choice of compensation policy, e.g. stock option compensation, significantly contributes to the mitigation of the agency problem by aligning the interests of managers with those of the owners of the corporations. Wiseman and Gomez (1998) showed that lacking the appropriate incentive is closely related to a specific risk-averse and timid behavior of CEOs, who prefer low-risk projects yielding insufficiently small payoffs. Agency theorists such as Jensen and Murphy (1990) proposed that one way of making CEOs to overcome their
risk-aversion and thus to rank the investment alternatives as stakeholders do, is to shift the compensation towards stock options. Given the underlying principle of options, the CEOs should in theory strive to increase the stock price, thus maximizing the shareholders’ value, which is tightly associated with the magnitude of monetary payoff that the CEO eventually obtains. As mentioned by Sanders et al (2007) the value of an option is determined by six factors: risk free interest rate (r), the underlying stock price (S), the exercise price of the option (X), time remaining till expiration of an option(T-t), firm’s dividend rate payment (d) and finally underlying volatility of stock (σ). Risk free interest rate r is exogenously determined by global markets, thus being out of the reach for any corporate executive. Similarly, exercise option price (X) and expiration time (T-t) are both specific features of options and cannot be influenced by executives.
On the other hand, security’s stock price (S) is within the reach of CEOs. Since the compensation options are call-options, their value is directly proportionate to the underlying stock price (S) of the security. In other words, increasing the stock price
in turn increases the value of options. This phenomenon is one of the primary reasons why stockholders approve stock-options compensations. Furthermore, executives can easily manipulate with two other main factor which determine the option value, the firm’s dividend rate payment (d) as well as underlying volatility of equity (σ).
Dividends
Firstly, given that managers often decide about the dividend payments, which in essence decrease the value of options (by decreasing the stock price S), the CEOs incentive is to drastically decrease the dividends to shareholders or alternatively replace the dividends by massive share repurchases (Jolls Ch., 1998). Similarly, Hall and Murphy (2002) argued that stock-option compensation motivates executives by creating a direct link between firm performance and executive wealth, thereby providing incentives for executives to firmly favor actions that increase share price and avoid actions that decrease share price, such as dividends.
However, as there is almost none empirical or theoretical reason to believe that such an
adjustment of the firm’s financial policy has a significant effect on the underlying value of the company, the dividend policy may not be of great concern to stakeholders (Jolls, Ch., 1998).
Firm’s Volatility
Secondly and more importantly, top managers can influence the volatility of equity (σ). Given that options’ value is increasing in volatility, executives can increase the firm’s volatility in two basic ways. One possible alternative suggests that executives might increase the number of high-risk projects which in turn increases risk at the left-hand side of firm’s balance sheet at market values. The relevant example is mentioned by Sanders et al (2007) who demonstrated the excessive risk-taking of CEO by different scenarios in probability payoff matrix. Each scenario in figure 1 represents CEO’s decision with a certain level of underlying risk. In general, a decision that could potentially lead to a wide range of possible outcomes is considered to be riskier than one that can lead to a narrow, tightly bounded array of possible results.
7 | P a g e
Figure 1 depicts different possible investment scenarios for a given CEO. Percentage gain or loss is given in the top row, followed by estimated probabilities of occurrence in the cells below. The overall riskiness of scenario increases from A through E. Note that in case of A and B scenarios, the range of possible values is exactly the same (narrow range of possible values), thus the only difference is the magnitude of investment. Scenario C includes moderate range of possible values and D and E include even more extreme outcomes with a possibility of big gains or losses.
On one hand, according to Low (2009), the highly undiversified, risk-averse managers who obtain casual compensation (salary, bonus or stocks) tend to turn down a project with risk-increasing, positive NPV, unless they are further compensated for taking on an additional systematic or firm specific risk. Therefore, looking back at figure 1, absent the appropriate compensation, risk-averse CEO prefers A over B (or B over C) simply because the possible outcomes are substantially less extreme, (in case of B over C) thus bears less risk. However because shareholders are in most cases widely diversified and risk-neutral. they demand managers to be more aggressive and favor the investment with the highest expected value possible. (Scenario C in Figure 1.) Therefore these former actions (scenario A) are highly
undesirable from the shareholders’ point of value-maximization principle.
Stock-option compensation presents a potential solution to the risk-aversion problem by allowing CEO to participate in limitless upside gains, while at the same time providing a floor to avoid significant loses. Given that value of an option monotonically increases in volatility (σ), the risk-averse executives have strong incentives to engage themselves in a more risky project yielding potentially greater payoff to both shareholders as well as CEO assuming that the investment works out well (Strong preference of scenario C over B).
On the other hand, option grant could create a substantial incentive for risk-tolerant executives to engage themselves in too much risk taking (Jolls, Ch., 1998). Investment Scenario Value of investment -100% -80% -60% -40% -20% 0% 20% 40% 60% 80% 100% A 100 0.5 0.25 0.25 B 1000 0.50 0.25 0.25 C 1000 0.50 0.25 0.25 D 1000 0.60 0.40 E 1000 0.60 0.40
Option-loaded CEOs have significant incentive to shift the primary focus on big gains and have almost no reason to be concerned with big loses. This shift is a consequence of stock-option asymmetric payoff; unlimited upside potential and no downside risk. (Sanders, G. and Hambrick C., D., 2007).Therefore, these CEOs prioritize different scenarios according to the expected value of the upside outcomes rather than from the expected value of full range of possibilities (upside and downside). If we accept the proposition that big losses are usually accompanied by substantial gains and consider that option-loaded CEOs have a tendency to ignore the likelihood and magnitude of these loses, then it is apparent why CEOs tend to choose projects characterized by a substantial chance of major loss (Milgrom & Roberts, 1992 in: Sanders, G. and Hambrick C., D., 2007). Looking at the Figure 1, which laid out different investment scenarios, we can expect that an option-loaded CEO will undertake Scenario E, because of the greatest expected value for gains, regardless of a negative overall expected value (negative NPV) and 60 percent probability of generating an extreme loss. In a certain
point, CEOs lose sight of the scenario’s downside effect and become "risk lovers" (Gomez-Mejia, L., R. And Wiseman, R. M., 1998). Several empirical studies confirmed the relationship between the excessive risk-taking and option grants. Hemmer et al. (2000) pointed out the convexity of risk-taking as a function of stock-option compensation. Furthermore, it appeared to be the case that the convexity (of the risk-taking effect) is decreasing when executives become more risk-averse. The other alternative of increasing firm’s volatility is to adjust capital structure by increasing leverage, hence increasing risk on the right hand side of the firm’s balance sheet (Gerhart B. and Milkovich, G., T., 1990). In this case however, the effect of stock-option compensation entirely depends on whether the company already is at the optimal capital structure. If that is indeed the case, increasing leverage any further is value-destroying rather than value-creating. In other words, depending on managers’ utility function (risk-averse or risk-tolerant), executive stock-option
compensation might cause value
destruction by motivating CEO to over-lever the company (Berger, P., Ofek, E., Yermack, D., 1997).
9| P a g e
3 Research method
Data
As suggested by Shavell, S., (1979) it appears to be the case that the effect of stock-option compensation on a CEO’s decision is closely associated with the increasing size of the company. Furthermore, it is common that larger companies tend to prefer stock-option compensation and are often subject to public information disclosure as well as increased public awareness. Therefore, the analysis is based on the Russell Top 200 index, which incorporates approximately 200 of the largest US securities with respect to their market capitalization. Not all 200 firms were used in the final analysis, because of mergers and missing compensation data or any other relevant financial data. Eventually, 126 companies out of 196 were used.
In total, three distinct databases were used in order to gather relevant data. The Execucomp database was used to obtain executive compensations (total, options and stocks) for the whole set of companies. Options are valued based on the Black-Scholes formula for valuing European call options, as modified by Merton (1973). The time span of data ranges from 1993 through 2000, since the comprehensive compensation data from Execucomp data base are not available before 1992 and
some other prior variables were used. Furthermore the time span of data was carefully chosen in order to avoid period of financial crisis and its influence on the data (the combination of variables does not exceed year of 2006).
Compustat database is used to obtain companies’ financial data including market value of equity, book value of equity, firm size, total assets, capital expenditure, R&D expenditure, acquisition, advertising, long term debt and others.
Finally company stock prices, total stock outstanding and returns are retrieved from CRSP database.
3.1 Empirical methodology
The idea to test the effect of executive stock-option compensation (ESO) on risk-taking and consequently risk-risk-taking effect of ESO on companies’ performance, introduces an essential problem of endogeneity in all three dependent variables of which two are subsequently used as independent variables. Moreover, in addition to ESO’s influencing executives’ risk-taking behavior in equation 2, the causality is highly likely to follow the other direction likewise (double causality). Therefore, to counteract these issues the 3-stage least squares technique was employed (reg3 command in Stata). In this methodology, three equations of option compensation, risk-taking and
company’s performance are studied. Thus the three dependent variables are explicitly treated as endogenous variables to the
system and are considered to be correlated with the disturbances in the equations.
1. ESO T1 = f (Prior Stock Risk t-1, Dividend t-1, Managerial Ownership t-1, Leverage t-1, Capital Investment t-1, Size t-1, Unexercised Options t-1)
2. Risk T2 = f (ESO t, (ESO2)t, Managerial Ownership t, Unexercised Options t, Cash flow t, Size t, Market-to-Book ratio t, Tech Dummy)
3. Performance T3= f (Risk T2, Former Performance T2, Tech Dummy)
3.2 Regression coefficient expectations
All 3 equations contain several explanatory variables which are specified in the research done by Chen, Y-R. and Ma, Y., (2010). Even though we are mainly interested in coefficient of ESO in second equation and Risk T2 in third equation, which provide information about the effect of ESO on executive risk-taking and performance respectively, the theory
suggests the following;
Prior stock risk is a measure representing the volatility of the company. As suggested by Baker, Jensen and Murphy (1988), firms with higher volatility are believed to compensate executive to a greater extent than companies with lower volatility. Thus the theory suggests positive regression coefficient here. As to dividends variable, cash dividends are expected to reduce the agency problem (Jolls, Ch., 1998). Therefore increasing dividends imply less need to use option compensation as a tool
for ameliorating agency problem, thus the regression coefficient is expected to be negative. Similarly, if an executive owns a substantial portion of companies stocks, there is less incentive for shareholders to use options to mitigate the agency problem (Chen, Y-R. and Ma, Y., 2010). Thus, regression coefficient for Managerial ownership is expected to be negative. As to firm’s leverage, the regression coefficient is anticipated to be negative, since increasing level of financial leverage mitigates agency problem by for example avoiding the overinvestment issue, where managers cause their firm to grow beyond an optimal size. Issuing new debt increases interest payments which reduce free cash flow available to managers; hence managers are forced to choose their investment more cautiously (Berger, P., Ofek, E., Yermack, D., 1997). Companies with a substantial capital investment yield
11| P a g e
higher, profitable growth opportunities, which increase the need to reduce agency problems through executive ESO (Chen, Y-R. and Ma, Y., 2010). Therefore the regression coefficient for Capital Investment is likely to be positive. As indirectly mentioned in the section 2, large companies are expected to award substantially higher number of options in order to compensate CEOs for relatively higher responsibility. Thus the coefficient of size is expected to be positive (Shavell, S., 1979) Lastly, Unexercised Options variable is a reasonable proxy of future firm’s compensation policy. In other words, a high number of unexercised options suggests that the company has experience with awarding executives with stock-option compensation and is more likely to stick to the compensation from the current period onwards (Armstrong Ch., S. and Vashishtha, R., 2012). Therefore the variable is expected to have positive
regression coefficient.
The explanatory variable ESO (variable of interest) in equation 2 is predicted variable from equation 1. In accordance with literature, the regression coefficient of ESO should be positive, since option compensation should stimulate CEO to increased risk-taking due to several above mentioned reasons (Gomez-Mejia, L., R. &
Wiseman, R. M., 1998). This logically leads to the following hypothesis:
Hypothesis 1: Stock-option compensation is believed to increase CEO’s risk-taking, thus a positive relation between dependent variable Risk, and explanatory variable ESO (LIF) is expected.
The impact of Managerial Ownership variable on the risk taking might be either positive or negative depending on which type dominates; tolerant or risk-averse. The coefficient of ESO2 should be in principle negative, however it might be the case that the CEO is risk-tolerant to such a degree that the coefficient turn out to be positive. (for certain range of ESO values). The Unexercised Options variable has, in this equation, similar effect as ESO, since increasing the number of unexercised options should further stimulates CEO to increase risk-taking (Chen, Y-R. and Ma, Y., 2010). Thus the coefficient is expected to be positive indeed. Variable Cash flow in this case represents capital available for any sort of further investment. Therefore, the higher the cash flow, the lower the investment costs such as interest expense on raising capital through debt issuance. Consequently the lower the investment costs the lower the investment risk which is reflected in negative regression
coefficient (Hayes, M., R., Lemmon, M. and Qiu, M., 2012). The effect of variable Size is in this equation genuinely different from equation 1. According to the theory of firm, the bigger the firm, assuming the same industry, the less risky it is compared to its smaller alternative. It is commonly believed that smaller firms often struggle to keep up with regulatory requirements and cannot fully exploit economies of scale and scope (Schuler, R., S. and Rogovsky, N., 1998). As a result, the coefficient should be negative. The coefficient of market-to-book ratio is expected to be positive, since the variable itself technically represents the growth opportunities, which are proportionate with higher risk (Chen, Y-R. and Ma, Y., 2010). Lastly, different industries tend to differ in underlying risk. Therefore the dummy variable is used in order to reflect the higher risk for technology industry (Schuler, R., S. and Rogovsky, N., 1998). Therefore the coefficient is expected to be positive. Finally, the third equation measures the effect of Risk on corporate performance. The variable Risk is predicted variable from the equation 2. On one hand, since the research previously
conducted substantially differs in conclusions it is entirely reasonable to expect positive or negative regression coefficient depending on the assumption of CEO’s decisions with respect to negative NPV projects. On the other hand, the theory suggests that compensation committees are prone to provide ESOs to inspire assumption of risky but positive NPV projects which further improve company’s performance. Thus the second hypothesis is as follows:
Hypothesis 2: Assuming that the CEO does not undertake negative NPV projects, a positive relationship between company’s performance and CEO’s risk-taking is expected.
The effect of previous performance should be negative given that the assumption of mean reversion theory holds. (Chen, Y-R. and Ma, Y., 2010). Lastly, the dummy variable effect is similar to the equation number 2, therefore the regression coefficient is expected to be positive, since technology companies tend to perform better relatively to other industries (Schuler, R., S. and Rogovsky, N., 1998).
13| P a g e
3.3 Variable specification
3.3.1 Measuring Executive Option Compensation.
Even though a substantial proportion of the previous papers focused on the ESO compensation of top five executives within a company, this analysis is focused on CEO compensation only. According to Rajgopal S. and Shevlin, T., (2001) there are at least three rationales which justify in particular the focus of the analysis on CEO. Firstly, CEO assumes overall responsibility for the company including its operations. Furthermore the decisions regarding the operating risks are most probably done at this level not lower. Secondly, CEO stock options and other relevant financial variables are easily accessible from Execucomp, whereas for other top executives it might be quite problematic. Thirdly, since substantial number of papers is in particular dedicated to CEOs, it is of great advantage to be able to compare results to previous research results. Furthermore, researchers such as Aboody (1996 in: Rajgopal S. and Shevlin, T., 2001) argued that the differences between using top five executives’ measure vs CEO’s measure are in essence often negligible. Therefore, the incentive effect of the stock-option compensation is defined and measured as a Black-Scholes value of CEO’s stock-options awarded to the value of total
compensation obtained (referred as
Incentive fraction ( IF)). The IF is further
transformed using natural logarithm function. Furthermore, because some
companies award stock-options
periodically rather than on yearly basis and because stock-option grants have motivational effects over several upcoming years, LIF is measured over for year period and averaged out (LIF (t, t+3)) (Chen, Y-R. and Ma, Y., 2010).
3.3.2 Measuring firm’s Risk
Most previous studies measured risk as a volatility (standard deviation) of future predicted stock returns using either CAPM or other asset pricing model. This could be according to Armstrong et al ( 2012), highly problematic when examining the effect of CEO’s risk-taking incentives, since future volatility of expected returns mirrors not only the effect of CEO’s risk-taking decisions but also firm-specific environmental characteristics or information disclosure. Furthermore, the CEO’s compensation structure cannot be assumed to be stable in the future because of unexpected changes in company’s contracting environment or alternatively the company can change CEO who has significantly different risk-characteristics.
Therefore a measure of risk based on future realized volatility, would not necessarily represents current risk preferences of CEO and would thus introduce biased and fallacious correlation between CEO risk-taking incentive and firm’s risk. As a result, to construct an appropriate measure for risk while at the same time avoiding above mentioned biases, the standard deviation of average yearly stock returns composed from monthly stock returns (StSD (t, t+3)), measured over a 4 year period horizon is used. Choosing this specific measure for risk allows for more straightforward test of the desired relationship between Stock-option compensation effects on risk-taking behavior of CEO.
3.3.3 Measuring corporate performance
The corporate performance in equation 2 is measured on the basis of stock performance rather than accounting performance for two main reasons. Firstly, accounting measures are more often subject to manipulation practices such as income smoothing, than stock performance measures. Secondly, ESO is defined as forward looking measure, whereas accounting performance is a backward looking indicator of formerly realized company’s policies (Sanders, G. and Hambrick C., D., 2007). Despite these disadvantages, Chen et al (2010)
conducted an empirical analysis employing both types of measures separately. They reported substantially similar results using both methods.
3.3.4 Control variables
The model uses several control variables, which are further described below. Furthermore the descriptive statistics of both dependent and control variables is showed in Table 1.
Equation 1
Prior Stock Risk: Standard deviation of
average yearly stock returns, measured over a five-year period, composed on monthly basis (t-5, t-1). The raw data is
downloaded from CRSP database.
Dividend: Dividend per share over stock
price. Dividend per share is downloaded from Compustat database and further adjusted by dividing by appropriate fiscal
share price.
Managerial Ownership: Number of total
shares owned by CEO (excluding options) to number of total shares outstanding. Both sub variables are downloaded from Compustat database.
Leverage: Total long-term debt to the book
value of Equity, both obtained from Compustat database.
Capital Investment: The ratio of R&D
expenditures, acquisitions, advertising and capital expenditure to total assets. Each
15| P a g e
one of them obtained from Copustat.
Size: Natural logarithm of the market value
of Equity. Sub variable obtained from Compustat.
Unexercised Options: Number of unexercised options owned by CEO to total shares outstanding. The first sub variable is obtained from Compustat, the latter from Execucomp database.
Equation 2
Cash Flow: The ratio of cash flow from
operation to total value of assets. Both
obtained from Compustat database.
Market-to-Book ratio: Ratio of market
value of equity to book value of equity. Both retrieved from Compustat database.
Technology Industry Dummy: Dummy
variable equals to 1 if firm operates in Industrial, IT or Communication sector, 0 otherwise.
Equation 3
Former Performance: Average yearly
stock returns, composed from monthly returns (
AvRet
(t, t+3)). Stock returns are retrieved from CRSP database.Paramaters Number of
Observation
Mean Standard Deviation Median
LIF(t,t+3) 484 -1.02971 0.59740 -0.9612
LIF2(t,t+3) 484 1.416458 1.743848 0.923914
Stock Volatility StSD (t,t+3)
484 0.08299 0.032913 0.075973
Previos Stock Volatility
(t-5,t-1) 484 0.076069 0.027632 0.068825 Managerial Ownership (t-1) 484 2.07374 8.252147 0 Unexercised option (t-1) 484 0.01228 0.00246 0.000561 Capital Investment (t) 484 0.100329 0.078688 0.084848 Leverage (t-1) 484 0.605616 1.398409 0.45607 Did. Yield (t-1) 484 0.021753 0.029167 0.01945 Size (t-1) 484 9.085955 1.116384 9.010799 Size (t) 484 9.28198 1.119762 9.180417 Managerial Ownership (t) 484 1.842665 7.238568 0 Unexercised Options (t) 484 0.001473 0.002822 0.000636 Cash Flow (t) 484 0.120448 0.077551 0.117471
Table 1 Descriptive statistics of dependent and control variables.
3.3.5 Model specification
With the specification of both dependent and control variables above, the empirical model looks as follows;
1. LIF (t, t+3) = f (Prior Stock Risk (t-5,t-1), Dividend t-1, Managerial Ownership t-1, Leverage t-1,
Capital Investment t-1, Size t-1, Unexercised Options t-1)
2. StSD (t, t+3) = f (LIF (t, t+3), (LIF 2(t, t+3)), Managerial Ownership t, Unexercised Options t,
Cash flow t, Size t, Market-to-Book ratio t, Tech Dummy)
3. AvRet (t+4, t+8) = f (StSD (t, t+3), AvRet (t, t+3), Tech Dummy)
Notation t stands for base year (t = 1993-1997). In equation 1 and 2 respectively, the dependent variables, LIF and StSD are measured over 4-year time period. In equation 3, the dependent variable AvRet is measured over 5-year time period following the 4-5-year of Executive stock-option award (LIF).
4.1 Results from empirical analysis
Results from equation 1
Using 3-stage least squares methodology, the following results, shown in Table 2, were obtained. Most of the regression coefficients have in general predicted sign (the 2 regression coefficients of the utmost interest are LIF (ESO) in second equation and StSD in third equation) and the overall results are consistent with the results from majority of previous papers.
Specifically in equation 1, as previously expected the Prior stock risk regression coefficient is positively related to stock-option award (ESO) with a statistical significance of 1 %. Capital investment regression coefficient is positively correlated to LIF, which indicates that
CEO’s magnitude of
stock-Market-to-Book (t) 484 4.227924 5.314448 3.104079
Average return (t,t+3) 484 0.009029 0.010576 0.00834
Tech. industry dummy 484 0.019797 0.013863 0.018497 Average Return (t+4.t+8) 484 0.338843 0.473806 0
17 | P a g e Parameters Equation 1 (ESO equation) Equation 2 (Risk equation) Equation3 (Performance equation) Intercept -2.500646 (0.2755621) 0.300275 (0 .0387751) -0.0058674 (0.0020048) LIF(t,t+3) 0.2001549 *** (0.0445706) LIF2(t,t+3) 0.0495549 *** (0.0139216) Stock Volatility StSD (t,t+3) 0.2807624 *** (0.0300404)
Previos Stock Volatility
(t-5,t-1) 6.602475 *** (1.010943) Managerial Ownership (t-1) -0.0044888 (0.0030418) Unexercised options (t-1) 28.37972 *** (10.33038) Capital Investment (t) 0.6050031 ** (0.3092181) Leverage (t-1) -0.0118958 (0.01732) Did. Yield (t-1) -2.460641 *** (0.8625869) Size (t-1) 0.1036195 *** (0.0250569) Size (t) -0.0069462 *** (0.0021503) Managerial Ownership (t) 0.0006853 ** (0.0003162) Unexercised Options (t) -0.8780202 (0.8485498) Cash Flow (t) -0.1001159 ** (0.0410821) Market-to-Book (t) -0.0010938 ** (0.0005484) Average return (t,t+3) -0.4145488 *** (0.0457888)
Tech. industry dummy -0.000552
(0.0054073)
-0.0005811 (0.0010937 )
Table 2.
The table presents results from 3-stage least squares estimation of the 3 equations. Coefficients include time specification t. Robust standard errors are reported in parenthesis below the regression coefficients. The following notation is used for different level of significance; at 10% , 5% and 1% as *, **, *** respectively.
option compensation necessary to mitigate agency costs increases as firm’s future prospects (represented by Capital investment) improve. Furthermore, the coefficient is statistically significant at 5% with a p-value of 0.05. The coefficient of Dividend variable is significant at 1% level with a p-value of 0.004 and is negatively related to ESO exactly as the theory suggests. As to Managerial ownership variable, even though the predicted effect of Managerial ownership variable on ESO corresponds to the obtained negative sign, it appears to be the case that the regression coefficient is statistically insignificant even at 10% level with a p-value of 0.128. Similarly, the coefficient of leverage is highly statistically insignificant with a p-value of 0.421 with a negative sign. Given the insignificance of these two control variables, the different 3-stage least squares regression was conducted excluding these two variables. Although the results are not substantially different than before (no change in significance or magnitude of coefficient), it is not reasonable to exclude the variables out of the model since both play a substantial role as control variable.
Results from equation 2
In the risk equation, equation 2, the regression coefficient of stock-option compensation (LIF) is positively related to
stock volatility and statistically significant at 1% level. Consequently it appears to be the case that there exists positive relationship between ESO and CEO’s risk –taking. This finding is highly consistent with both theoretical and empirical literature and supports the aforementioned hypothesis 1. It is necessary to note, that there is statistically significant and positive
relationship between Managerial
ownership variable and risk, which suggests (as mentioned before in section 3.2) that in this case the risk-taking incentive dominates risk-aversion effect. The coefficient of stock-option compensation squared (LIF2) is positively related to risk, which implies that the incentive effect of ESOs on CEO’s risk-behavior is exponentially increasing in stock-option compensation. This situation is the exact opposite of what would have happened under the risk-aversion
dominance (negative Managerial
Ownership coefficient), when the incentive of ESOs are increasing for relatively small value of ESO. In other words, given the risk-aversion dominance, it appeared to be the case that overused ESOs have an adverse effect on managerial risk-taking behavior. The coefficient of Cash flow is negatively correlated with risk with a statistical significance of 5% and p-value of 0.015. Similarly as cash flow, the
19| P a g e
coefficient of Size exhibits a statistically significant (1%) and negative relationship with dependent risk variable. Similarly as in equation 1, there appear to be some insignificant control variables such as Unexercised Option holdings and Market-to-Book ratio. Both are according to theory determinants of Risk and thus are important to the model itself. Therefore cannot be excluded due to misspecification issue. The coefficient of dummy variable, representing Technology sector, is highly insignificant with a p-value of 0.919. Thus using the specific data set and model, it appears to be the case that there is no appreciable risk differences between Technology industry and the other industries when controlling for other variables.
Results from equation 3
Finally and most importantly in the performance equation, equation 3, the regression coefficient of Risk (StSD) is positively related to corporate performance with a p-value close 0. This indeed suggests that Risk-taking behavior, induced by executive stock-option compensation ESO, improves company’s future performance. This finding is consistent with hypothesis number 2. The regression coefficient of previous performance demonstrates negative
relation to future performance. Both coefficients are statistically significant at 1 % level. Lastly, the dummy variable appears to be repeatedly statistically insignificant with a p-value of 0.595, which demonstrates that the sample exhibits no performance differences related to the industry selection when controlling for other variables.
4.2 Omitted variable bias in risk-equation (OVB)
According to Yermack (1995), other cash compensation (salary or bonus) could also create an incentive for CEO to behave differently and thus should be included in the model determining the risk-incentive of executives. The issue of OVB could arise specifically in this model, which ignores the other compensation (excluding stocks) to be a fully-fledged determinant of risk-taking incentive (StSD). Therefore, a regression including the variable Other Compensation was conducted in order to control for a potential effect on StSD. The data was obtained from Execucomp and appropriately adjusted in order to fit the specified model. Despite the argument, no significant results were found; the coefficient of Other Compensation was statistically insignificant with a p-value of 0.191. Furthermore, no substantial changes in other explanatory variables were made.
Table 3 The table contains regression results including additional control variable Other
Compensation in equation 2. The coefficient of regression for the additional appeared to be insignificant. Despite the minor changes in regression coefficient of control variables in equation one and equation two there were no substantial changes either in significance or values of (LIF), (StSD)(coefficients of interest)
Paramaters Eqaution 1 (ESO equation) Equation 2 (Risk equation) Equation3 (Performance equation) Intercept -2.501472 (0.2757267) 0.2962923 (0 .0375521) -0.0057249 (0.0019963) LIF(t,t+3) 0.1978168 *** (0.0436913) LIF2(t,t+3) 0.0488761*** (0.0136179) Stock Volatility StSD (t,t+3) 0.2778845 *** (0.0299028)
Previos Stock Volatility
(t-5,t-1) 6.620602 *** (1.011751) Managerial Ownership (t-1) -0.0044669 (0.0030433) Unexercised options (t-1) 28.61177 *** (10.33493) Capital Investment (t) 0.5875571 * (0.3091526) Leverage (t-1) -0.0103857 (0.0174072) Did. Yield (t-1) -2.502747 *** (0.8615016) Size (t-1) 0.1037136 *** (0.0250647) Size (t) -0.006971 *** (0.0021232) Managerial Ownership (t) 0.0006983 ** (0.0003126) Unexercised Options (t) -0.8604558 (0.8368811) Cash Flow (t) -0.0973782 ** (0.0402937) Market-to-Book (t) -0.0010229 * (0.0005333) Average return (t,t+3) -0.4101114 *** (0.0455723)
Tech. industry dummy -0.0003784
(0.0053319) -0.0005562 (0.0010901) Other Compensation(t) 0.0337515 (0.0258311)
21| P a g e
5 Conclusion
The main purpose of the paper was to analyze the risk-taking effects of ESOs (executive stock-option compensation) on company’s performance. For this purpose a sample of Russell top 200 index companies was used. The empirical analysis was conducted by applying a 3-stage least squares methodology in Stata (reg3). The model used, composed by Yenn-Ru Chen and Yulong Ma, consists of 3 different equations; Stock-option equation, (ESO or LIF) Risk-taking equation measured as volatility of stock StSD and finally performance equation measured as subsequent average stock return AvRet. The results found are to a great extent consistent with the exiting literature and theory. It appeared to be the case that for the specified sample of 200 biggest companies represented by their CEOs, listed at New York Stock Exchange (NYSE), that a positive non-linear relationship exists between stock-option compensation and risk-taking of CEO (consistent with hypothesis 1). In other words, the increasing level of ESOs has a positive and increasing effect on risk-taking of CEOs. This finding is highly justifiable since the sample exhibits dominance of the risk-taking effect over risk-aversion (Managerial Ownership coefficient in equation 2 is positive).
Furthermore, a significantly positive effect was found between risk-taking effect of ESOs and a company’s future performance (consistent with hypothesis 2). This indeed suggests that given the specific sample, the executive stock-options compensation provides incentives to managers which are value-increasing if given a certain time to observe the effect (4 years in this case). Furthermore, no significant changes were found by testing the theory of Yermack 1995, who suggested that Other Compensation also play a substantial role in determining the risk-incentive. In this case the results were found insignificant. An important caveat of the research is a relatively small sample of 200 companies, which are used for the core analysis and thus the external validity of this research is questionable. (Different results might apply for smaller companies). On the other hand, the analysis should be a good estimation of the effect of ESO on risk-taking which in turn influences company’s performance for larger corporations absent any significant disturbances such as economic recession. As result, future work could assess the effect for larger and smaller companies separately or even include cross-industry differences. Alternatively, trying to assess the real reason why stockholders award stock-option compensation (mitigating agency problem or perhaps for accounting
23| P a g e
References
Armstrong Ch., S. and Vashishtha, R., (2012), Executive stock options, differential risk taking incentives and firm value, The Journal of Financial Economics, pp. 70- 88.
Baker G., P., Jensen C., M. And Murphy J., K., (1988), Compensation and Incentives: Practice vs Theory, The Journal of Finance, Vol. 43, pp. 593-616.
Berger, P., Ofek, E., Yermack, D., (1997). Managerial entrenchment and capital structure decisions. Journal of Finance 52, pp. 1411–1438.
Brian, H., J. and Hall, J., K., 2002, Stock Options for Undiversified Executives, The journal
of Accounting and Economics, Vol. 33, pp. 1-42.
Chen, Y-R. and Ma, Y., (2010), Revisiting the risk-taking effect of executive stock options on firm performance, The Journal of Business Research, pp. 640-648.
Efendi, J., Files, R., Ouyang, B. and Swanson E., P., (2013), Executive Turnover Following Option Backdating Allegations, The Accounting Review, pp. 75-105.
Gerhart B. and Milkovich, G., T., (1990), Organizational Differences in Managerial
Compensation and Financial Performance, The Academy of Management Journal, Vol. 33, pp. 663-691.
Gomez-Mejia, L., R. & Wiseman, R. M., 1998, A behavioral agency model of managerial risk taking. Academy of Management Review, 23: pp. 133-153.
Hayes, M., R., Lemmon, M. and Qiu, M., (2012), Stock options and managerial incentives for risk taking: Evidence from FAS 123R, Journal of Financial Economics, pp. 174-190.
Jensen C., M. and Murphy J., K., (1990), Performance pay and top-management incentives,
Journal of Political Economy, pp. 225–64
Jolls, Ch., 1998, “Stock Repurchases and Incentive Compensation,” NBER Working Paper, pp. 1-19.
Low, A., (2009), Managerial risk taking behaviour and equity-based compensation, Journal
of Financial Economics, pp. 470–490.
Rajgopal., S. and Shevlin, T., 2001,Empirical Evidence on the Relation between Stock Option Compensation and Risk Taking, Journal of Accounting and Economics, Vol. 33, pp. 145-171
Sanders, G. and Hambrick C., D., (2007), Swinging for the Fences: The Effects of CEO Stock Options on Company Risk Taking and Performance, The Academy of Management
Schuler, R., S. and Rogovsky, N., (1998), Understanding Compensation Practice Variations across Firms: The Impact of National Culture, Journal of International Business
Studies, Vol. 29, pp. 159-177
Shavell, S., (1979), Risk sharing and incentives in the principal and agent relationship. Bell
Journal of Economics, pp. 55–73.
Yermack, D., (1995), “Do Corporations Award CEO Stock Options Effectively?, ” Journal of
