CEO RISK-TAKING INCENTIVES, CORPORATE RISK-TAKING AND
THE EFFECT OF THE DEGREE OF INTERNATIONALIZATION
Abstract: This thesis examines the relationship between CEO risk-taking incentives (vega and delta) and corporate risk-taking by analyzing a sample of 2,584 firms containing 20,996 firm-year observations over the period 1992 - 2013. In addition, I explore the moderating role of the degree of internationalization. In the jointly determined CEO stock and stock option compensation package, I find that a higher sensitivity of CEO wealth to stock price (delta) will decrease corporate risk-taking, whereas an increase in the sensitivity of CEO wealth to stock return volatility (vega) will increase corporate risk-taking. In contrast, I find that firms with higher degrees of internationalization will experience a positive effect of delta on risk-taking. Since risk-taking studies are prone to endogeneity, complementary robustness checks are included and find additional support for the results of the main relationship.
Keywords: CEO risk-taking incentives, corporate risk-taking, internationalization
2
TABLE OF CONTENTS
INTRODUCTION... 3
LITERATURE REVIEW ... 9
2.1 Prior studies ... 9
2.2 Corporate risk-taking, corporate risk and R&D intensity ... 12
2.3 A CEO’s risk appetite and the agency theory ... 14
2.4 The risk-taking incentives mechanism ... 15
2.5 Corporate risk-taking and international diversification ... 17
DATA AND METHODOLOGY ... 23
3.1 Variables ... 23
3.2 Data and sample ... 27
3.3 Methodology ... 29
3.4 Model estimation ... 29
REGRESSIONS AND RESULTS ... 31
4.1 Descriptive statistics ... 31
4.2 Correlation matrix and endogeneity concerns ... 33
4.3 Empirical results and discussion ... 36
4.4 Robustness checks ... 44
CONCLUSION AND LIMITATIONS ... 49
5.1 Conclusion ... 49
5.2 Limitations and future research ... 52
Acknowledgement ... 54
BIBLIOGRAPHY ... 55
APPENDIX A ... 60
3
INTRODUCTION
Corporate risk-taking is essential for firm growth, firm performance and hence its survival (Boubakri et al., 2013). At large, corporate investment policies reflect a company’s degree of risk-taking. A rise in corporate risk-taking could take the form of increasing engagements in risky projects that increase a firm’s future default risk, such as investment in R&D projects. Investment in research & development (R&D), i.e. R&D intensity, is perceived as more risky when compared to e.g. capital expenditures because it is capital intensive, time-consuming and the outcomes of R&D projects have a high chance of failure (Holmstrom, 1989; Manso, 2011). Hence, R&D intensity is a strong determinant of corporate risk-taking (Sanders et al., 2001; Coles et al., 2006; Francis et al., 2017).
4 value-increasing, but risky R&D projects (Honoré et al., 2015), which is suboptimal for the shareholder. Hence, it is essential to motivate the CEO to commit to risk-increasing R&D projects, aligning the interest of the risk-averse executive with those of the shareholders (Jensen and Meckling, 1976).
Corporate governance mechanisms, such as incentivization through the composition of the CEO’s compensation scheme, allow shareholders to alter the risk appetite of the CEO (Smith and Stulz, 1985; Guay, 1999). In this compensation scheme, there are two dominant types of compensation that determine the risk appetite of a CEO: the amount of equity-based (henceforth, stock) and equity-based option (henceforth, stock option) compensation. By means of tying the personal wealth of a CEO to stock or stock options, shareholders are able to exert influence on this aforementioned risk appetite.
The concept of risk-taking incentives, henceforth indicated as vega1 and delta2 is based on adding the mechanism of convexity in the CEO’s payoff structure (Hayes et al., 2012; Gormley et al. 2013). Guay (1999) explains convexity as “the sensitivity of managers’ wealth to the volatility of equity value” (p.44).3 Stock option compensation makes the CEO wealth sensitive to the
volatility of the stock returns and is measured in vega. Alternatively, stock compensation makes
1 Coles et al. (2006) define vega as: “the change in the dollar value of the executive’s wealth for a 0.01 change in the annualized standard deviation of stock returns.” (p.439). Coles et al. and many other scholars (e.g. Hayes et al., 2012; Francis et al., 2017) derived this definition from the seminal work of Guay (1999).
2 Coles et al. (2006) define delta as: “the change in the dollar value of the executive’s wealth for a one percentage point change in stock price”. (p. 439). Coles et al. and many other scholars (e.g. Hayes et al., 2012; Francis et al., 2017) derived this definition from the seminal work of Guay (1999).
5 the CEO wealth sensitive to changes in stock price, as measured by delta. Delta is expected to make the CEO more risk-averse because in case of a decrease (increase) in stock price, delta will directly decrease (increase) the CEO’s wealth and hence make the CEO very sensitive to the stock price. On the contrary, vega is expected to increase a CEO’s risk appetite because an increase in the stock return volatility will increase the value of the options held by the CEO and thus increase his or her current wealth (Coles et al., 2006; Hayes et al., 2012; Francis et al. 2017). Since stock options bear different features than stock, the CEO will only face the upside potential of the stock option. CEOs will be rewarded for creating equity-value. Consequently, when a CEO enjoys a higher vega, a CEO is expected to be more driven to select corporate investment policies that will increase the volatility of the firm’s stock returns, which drives up the risk premium of the firm, in favor of the shareholder. Hence, this thesis addresses the importance of CEO risk-taking incentives in corporate risk-taking.
6 relationship between risk-taking incentives, corporate risk-taking and internationalization exist. Studies by Krapl et al. (2014) and Gao and Chou (2015) argue that lower cost of capital will lead to increasingly inefficient investments, creating higher corporate risk for which the shareholders wills seek compensation. Hence, they potentially decrease vega or increase delta.
Based on the theory building above, a range of questions arise. How do CEO risk-taking incentives, as captured by vega and delta, affect the risk appetite of a CEO? How does this translate into the levels of R&D intensity within firms? Is this relationship different for more internationally diversified firms? Consequently, the research question of this thesis is as follows:
‘Do CEO risk-taking incentives exert influence on corporate risk-taking, and how would the degree of internationalization play a role in this relationship?’
Hence, this thesis examines whether and how risk-taking incentives have an effect on corporate risk-taking and tests for the role of the influence of the degree of internationalization in this relationship.
This thesis will add to existing literature for several reasons. First of all, the effect of risk-taking incentives on corporate risk-risk-taking4 is an extensively discussed topic in finance literature (Guay et al., 1999; Coles et al., 2006; Low. 2009; Hagendorff and Vallascas, 2011; Hayes et al., 2012; Gormley et al., 2013; Armstrong et al., 2013; Francis et al., 2017). However, recent studies
7 question this relationship or even find contradictory evidence (e.g. Hayes et al., 2012; O’Connor et al, 2013). Hence, this research will provide new evidence on the, in literature defined as ambiguous, relationship between risk-taking incentives and risk-taking. Thereby, I add to the literature on corporate decision policies, as well as corporate governance controls. Secondly, a limited amount of studies have explored the relationship between taking incentives, risk-taking and the effect of the degree of internationalization of a firm, creating a gap in literature. Therefore, this study contributes to the literature about the role of degree of internationalization. Third, prior publications point out that compensation studies are plagued with endogeneity issues and find proof of causal effects which are difficult to overcome (Coles et al., 2006; Low, 2009; Gormley et al., 2013). Hence, this thesis controls for endogeneity by incorporating the residual estimates of vega and delta into the model. These excess vega and excess delta will crosscheck the validity of the estimated relationships of vega and delta. Concluding, this thesis will provide unique evidence of the relationship between CEO risk-taking incentives and corporate risk-taking in a multinational perspective.
In order to test for the theorized relationship, this thesis considers a sample consisting out of 2,584 U.S. listed firms (20,996 firm-year observations) in the period of 1992 – 2013. The relationship is assessed empirically and estimated by an OLS approach correcting for clustering at the industry- and firm-level.
8 on risk-taking. I do not find any support for a significant role of the degree of internationalization in the relationship between vega and risk-taking.
9
LITERATURE REVIEW
This chapter provides a comprehensive discussion of the relevant literature and explains the development of the hypotheses. First, relevant literature is reviewed. Hereafter, I develop my central hypotheses H1a and H1b by discussing the relation between the agency theory, corporate risk-taking and risk-taking incentives. Lastly, I develop hypothesis 2a, 2b, 2c and 2d by devoting attention to the ambiguous role of the degree of internationalization.
2.1 Prior studies
In the past decades, prominent literature has shed light on the relationship between risk-taking incentives and various proxies of risk-risk-taking. Early studies exploring the relationship are conducted by Jensen and Meckling (1976), Smith and Stulz, (1985) Lambert et al. (1991) and Sanders et al. (2001). Coles (2006) argues that the early measures of risk-taking incentives were rather noisy proxies. Guay (1999) was one of the first scholars to adopt the ground-breaking vega and delta measure. In literature, vega and delta are examined as an independent, as well as a dependent variable, indicating a causal relationship which is addressed later in this thesis.
10 The seminal work of Coles (2006) explores the relationship between risk-taking incentives and managerial risk-taking in investment policy choices, and find a positive relationship between vega and riskier policy choice. Hence, they find that firms with a higher CEO vega increase investments in R&D, have less capital expenditures and experience higher levels of leverage. In addition, they stress that vega and delta are jointly determined and argue that in determining the effects of vega or delta, it is important to control for the other incentive as well. In a similar fashion, Armstrong and Vashishtha (2012) find a positive association between vega and risk-taking and a negative association between delta and risk-taking. Remarkably, many dominant studies focus on vega as the main explanatory variable, whereas delta is incorporated to control for pay-performance sensitivity. Hence, Chang et al. (2015) explored the relationship between non-executive employee’s vega and corporate innovation and discover a positive relationship. Francis et al. (2017) explored the relation between executive stock option remuneration and policy choices on exchange rate exposure and find a positive relationship. This indicates that the risk-taking incentives have effects on different types of risk-taking policies. The same applies to the work of Hagendorff and Vallascas (2011). They study a sample of U.S. acquiring banks and find that CEOs incentivized by vega increasingly engage in risk-induced mergers. In addition, Boulton et al. (2014) and Croci & Petmezas (2015) investigated the relation between the executive vega and investments in the acquisitions context and both find positive relations. Hence, all the previously discussed studies find evidence of a positive relationship between vega and different types of corporate risk-taking.
12 Based on my review of prior studies, I conclude that strong empirical foundations and theories for this thesis are set in existing literature. Ambiguous results about the risk-taking incentive mechanism as well as the direction of the relationship between risk-taking incentives and corporate risk-taking exists.
2.2 Corporate risk-taking, corporate risk and R&D intensity
It is essential for firms to take adequate levels of risk in order to develop economic benefits for the risk-neutral shareholder. Hence, corporate risk-taking is vital for firm growth, firm performance and hence the survival of the firm (Boubakri et al., 2013) In literature, corporate risk-taking and corporate risk are inherently related. Studies frequently related to corporate risk as the systematic risk and idiosyncratic risk of the firm (Armstrong and Vashishtha, 2012), whereas risk-taking refers to changes in policy choices that affect the level of corporate risk. 5 Economically, corporate risk and risk-taking enjoy feedback effects (Coles, 2006), meaning that they influence each other both directions. Increases (decreases) in corporate risk-taking generally lead to increases (decreases) in corporate risk. However, corporate risk exerts influence on the level of risk-taking as well. The shareholder wants to receive the appropriate returns for the amount of risk premium he or she faces from holding an interest in the firm. Therefore, changes in corporate risk will affect the level of risk-taking within the firm. For example, when a firm experiences more risk, investors want to be compensated for this increased risk premium in terms of a higher rate of return. Hence, to realize the expected returns on the investment, the risk-neutral shareholders are
13 in favor of higher corporate risk-taking, to which they adjust the corporate governance mechanisms within the firm.
Within a firm, the CEO has the responsibility to set the appropriate risk-taking policy. The risk-neutral shareholders expect the risk-averse CEO to increasingly engage into relatively more risky, but value-increasing activities such as mergers and acquisitions (Boulton et al., 2014; Croci et al., 2015), lower cash holdings (Gormley et al. 2013), higher leverage (Gormley et al., 2013; Coles et al., 2006) and less capital expenditures (Coles et al. 2006) and increasing R&D expenditures (Coles et al. 2006; Gormley et al., 2013; Francis et al., 2017)
14
2.3 A CEO’s risk appetite and the agency theory
15
2.4 The risk-taking incentives mechanism
16 and thus be demotivated to increase company risk-taking. As a result of this increased risk adversity, CEOs potentially reject projects with a positive Net Present Value (NPV) which will eventually hurt shareholder value (Croci & Petmezas, 2015).
Concluding, this section theorized that higher amounts of stock compensation increases the sensitivity of CEO wealth to stock price (delta) which is consequently associated with lower levels of corporate risk-taking, hence lower levels of R&D intensity. Therefore, I hypothesize:
H1a: The CEO delta is negatively related to R&D intensity.
On the contrary, equity-based options (stock options) bear a different influence on the CEO wealth. Whereas delta ties CEO wealth to the stock price, vega will tie CEO wealth to the stock return volatility. Hence, vega represents the CEOs sensitivity to stock return volatility. In case of a rise in stock price (when the pay-off of the option is > 0), the wealth of the CEO will rise jointly with the shareholders’ wealth. However, in the case of a drop in the firm’s stock price (when the pay-off of the option is <0), the CEO will experience no decrease in its wealth since the option will only be exercised in the event when the stock price is above the exercise price. The convex pay-off nature of these options create an asymmetric risk effect and hence a motivation for a CEO to increase its risk appetite, since the CEO will share the gains but not all of the loss (Jensen and Meckling, 1976; Guay, 1999; Coles et al., 2006; Gormley et al., 2013; Croci and Petmezas, 2015). In other words, by tying the sensitivity of CEO wealth to stock return, the CEO will experience an increase in systematic firm risk but his or her personal idiosyncratic risk will remain the same.
17 increasingly granting stock options and thereby increasing vega functions as an incentive to increase risk-taking behavior of the CEO (Sanders & Hambrick, 2007). This ‘solution’ to the agency problem theorizes that in response to offset the concavity of a risk-averse CEO wealth utility function, corporate governance structures should make the CEO’s compensation scheme follow a convex payoff structure. By that means, CEO wealth is more sensitive to firm risk. Thus, in order to offset both delta-induced risk-aversion and a CEO’s natural risk aversion, vega is used as a risk-taking incentive (Guay, 1999).
Accordingly, one can state that equity-based option compensation is crucial in incentivizing CEOs to engage more into corporate risk-taking, such as R&D projects. Hence, I hypothesize that an increase in stock options compensation and therefore an increase in the sensitivity of CEO wealth to stock return volatility (vega) has a positive relation to increases in corporate risk-taking. This leads to the following hypothesis;
H1b: The CEO vega is positively related to R&D intensity.
2.5 Corporate risk-taking and international diversification
This section first discusses the effect of international diversification on corporate risk and corporate risk-taking. Hereafter, I zoom in on the effects of international diversification on corporate risk-taking from the financial flexibility perspective. Lastly, I address the relation between vega, delta, international diversification and corporate risk-taking.
risk-18 taking incentives, corporate risk-taking and the moderating effect of the degree of internationalization6. Aabo et al. (2015) define multinationalism as “corporate expansion across different geographic areas and encounters with a multitude of economic, political and cultural environments.” (p. 66). Multinationalism, henceforth referred to as international diversification, is a type if corporate risk-taking7 on itself since it could either increase the total risk experienced by
a firm, but also has the potential to decrease corporate risk. Hence, over the past decades, scholars present ambiguous opinions and arguments on the influence of the degree of internationalization8 on firm risk and corporate risk-taking.
The classical corporate international diversification theory argues that international diversified firms enjoy risk-reduction when compared to purely domestic firms (Hughes et al., 1975; Fatemi, 1984). Recent studies by Krapl et al. (2014) and Aabo et al. (2015) argue that multinationalism is a form of diversification of risk due to limited dependence on the firms home country. This could partially be explained by the increased diversification of funding multinational firms enjoy, henceforth referred to as financial flexibility. Gamba and Triantis (2008) describe financial flexibility as “the firm’s ability to access and restructure its financing at low cost, by
6 A study by Sanders and Carpenter (1998). investigated the association between risk-taking incentives and
internationalization. In a corporate governance study, they examined the relationship between risk-taking incentives and internationalization from an agency theory perspective. Through the lens of information asymmetries, they find that due to the increased board self-monitoring, the degree of internationalization positively affects the proportion of total compensation paid in long-term compensation (such as equity options).
7 The majority of literature investigating international diversification is focused on its relation to corporate risk, not explicitly risk-taking. As mentioned in this thesis before, corporate risk and corporate risk-taking are inherently related and enjoy feedback effects from each other. Increased corporate risk-taking leads to increased corporate risk, whereas increased corporate risk could also lead to an altering in corporate risk-taking: in cases of e.g. higher corporate risk compared to the actual risk premium, shareholders want the firms to compensate for this by taking extra risks in order to drive up the volatility and hence returns.
19 means of which firms become able to avoid the costs of financial distress and to fund investment when profitable opportunities arise” (p. 2263).
Studies by Shapiro (2008), Gamba and Triantis (2008) and Jang (2017) find a positive association between international diversification and access to foreign sources of capital: the multinational is able to borrow at cheaper rates, is less sensitive to the domestic capital market and is also less dependent on the domestic capital market (Bruno and Shin, 2014; Jang, 2017). Jang (2017) argues that when a firm’s location is relevant for financing, internationally operating firms should have easier access to different sources of foreign capital in comparison to domestic firms. Consequently, multinational firms are able to access international capital markets at a lower cost. Bruno and Shin (2014) highlight that due to a higher degree of internationalization, firms are able to lend in more countries. Due to spillover effects, credit conditions are different all over the world. Therefore, the firm is able to choose the lowest rate, leading to lower risk-adjusted lending rates. Consequently, a multinational firm will apply lower discount rates, which will increase the Net Present Value (NPV) of the investment decision. Thus, internationally diversified firms enjoy lowers risks and therefore lower costs of the capital resources allocated to the investments. Accordingly, multinational firms are expected to engage in more investment projects for any given profile of expected cash flows. Bruno and Shin (2014) also argue that, since multinational firms are relatively more diversified as compared to domestic firms, the returns of the multinationals will be less correlated with the domestic market.
20 benefit from the lower costs and to make up for the risk premium they face. In doing so, the shareholders will motivate the CEO to increase the investments and in doing so, increase corporate risk-taking (Jang, 2017). Hence, it could be argued that the shareholders will incentivize the CEO by increasing the CEO vega but decreasing the CEO delta. Another important argument is that in times of e.g. negative shocks, a financial flexible firm is better able to avoid financial distress by exploiting alternative funding sources (Gamba and Triantis, 2008). Hence, multinational firms will, due to their higher financial flexibility, have more incentives to continue engaging in investment opportunities to which the CEO compensation scheme will be adjusted accordingly, and thereby increase their corporate level of risk-taking.
21 H2a: The relationship between vega and R&D intensity will be strengthened by the firm’s degree of
internationalization
H2b: The relationship between the delta and R&D intensity will be weakened by the firm’s degree of internationalization
On the contrary, it is argued that international diversification has the potential to weaken (strengthen) the effect of vega (and delta) on corporate risk-taking. Several studies find evidence that international diversification leads to increased firm risk. From an agency perspective strong evidence is found that multinationals face more information frictions and agency conflicts leading to increased corporate risk (Aabo et al., 2015; Doukas and Pantzalis, 2003). Other scholar find that international expansion leads to an increase in firms risk, since firms face e.g. a higher exchange rate exposure (Francis et al, 2017). Krapl et al. (2014) find that a higher degree of internationalization leads to the increase of the complexity of the firm’s operating environment, an increase in political exposure all leading to an increase in agency problems such as higher information asymmetries and monitoring issues.
22 but investments that increase corporate risk. Moreover, Krapl et al. (2014) and Aabo et al. (2014) find that corporate international diversification increases idiosyncratic equity risk. Hence, due to the increased corporate risk the multinational firm faces, I expect the shareholder to reduce the amount of risk taken by the company and a decrease in the CEO vega and an increase the CEO delta.
Accordingly, it can be concluded that international diversification, in contrast to aforementioned findings, has the potential to increase corporate risk by facilitating inefficient investments and increasing idiosyncratic equity risk. This lead to lower levels of risk-taking and hence weaken (strengthen) the effect of vega (delta) on corporate risk-taking.
Hence, the effect of the degree of internationalization on the relation between vega, delta and corporate risk-taking is ambiguous. Therefore, which effect trumps is yet to be researched.
Consequently, I also hypothesize:
H2c: The relationship between vega and R&D intensity will be weakened by the firm’s degree of internationalization
23
DATA AND METHODOLOGY
The data and methodology is organized as follows. The first part presents the variables used in the analysis and hence discuss their relevance and operationalization. Hereafter, the data and sample are described. Consequently, the methodology is explained and lastly the model estimations are presented.
3.1 Variables
3.1.1 Independent variables
Vega and delta are the explanatory variables in this study. Including vega as well as delta in my regressions enables me to isolate the effect of the taking incentives on corporate risk-taking. Since the data on vega and delta in this study is retrieved from a study by Croci and Petmezas (2015)9, I follow their definition which is based on the studies of Guay (1999) and Coles et al. (2006). Croci and Petmezas (2015) define vega as “the change in the dollar value of the CEO wealth for a 1% change in the annualized standard deviation of stock returns”, whereas delta is “the dollar change in CEO wealth for a 1% change in stock price.” (p.4). The computation of the variables follow the methodology of Core and Guay (2002), who use the Black-Scholes option valuation model. Coles et al. (2006) argue that in order to overcome noisy proxies for vega and delta, computations should be based on an estimation of the CEO’s complete portfolio (Guay, 1999) and not on the number or value of options or stock held or granted. However, the option vega is much larger than the stock vega. Therefore, the vega of the option portfolio is used to
24 measure total vega of the option and stock portfolio (Croci & Petmezas, 2015). To control for the skewness of the data, the outliers and the observations where vega and delta take a value of zero, I transform the data prior to the regression to the natural logarithm of one plus vega or delta (Low, 2009; Croci et al., 2017; Francis et al., 2017; Armstrong et al. 2013).
3.1.2 Dependent variable
As discussed earlier, scholars use several proxies to measure corporate risk-taking. Coles et al. (2006) and Gormley et al. (2013) measure corporate risky policy choices by assessing proxies such as R&D intensity, cash holdings, capital expenditures and leverage. In this study, corporate risk-taking will be captured by R&D intensity. R&D intensity is commonly used as a proxy due to the risky nature of the investment in R&D activities. Following studies by Mansfield et al. (1981), Kotabe et al. (1990, 2002), Coles et al. (2006), Low (2009) and Honoré et al. (2015), this variable will be operationalized as R&D expenditure over total assets. Scaling the R&D expenditure by assets will account for the variation between firms, e.g. firm size and profitability. Following Coles et al. (2006) and Low (2009), missing values on R&D expenditure will take a value of zero.
𝑅𝐼𝑆𝐾𝑇𝐾 = (𝑅&𝐷 𝑒𝑥𝑝𝑒𝑛𝑑𝑖𝑡𝑢𝑟𝑒 𝑇𝑜𝑡𝑎𝑙 𝑎𝑠𝑠𝑒𝑡𝑠 )
3.1.3 Firm-level moderator
25 DOI of exporting firms with no or little foreign assets. Furthermore, counting the number of countries a firm operates in or the share of foreign employees does rather capture the number of markets a firm accessed than the actual size of the market, where the latter one is more important in order to determine the scale of the operations (Kafouris et al., 2008). Therefore, I argue that measuring the DOI by foreign assets scaled by local assets would be a stronger proxy. However, due to limited available data, I focus on the income from foreign operations scaled by total income. Jang et al. (2017) use a similar proxy; pretax foreign income divided by pretax total income.
Consequently, in this study the DOI is operationalized by pretax income from foreign operations over net income. However, it is a ratio where the numerator is an integral part of the denominator, potentially outweighing domestic (foreign) losses (gains) with foreign (domestic) gains (losses). Hence, all observations taking value >1 or <-1 will be corrected to 1 or -1 in order to deal with outliers without compromising the measurement of the DOI. Moreover, the variable only measures in absolute values, since a negative DOI does not exist. My this means, foreign losses are also taken into account.
𝐷𝑂𝐼 = ( 𝑃𝑟𝑒𝑡𝑎𝑥 𝑖𝑛𝑐𝑜𝑚𝑒 (𝑙𝑜𝑠𝑠) 𝑓𝑟𝑜𝑚 𝑓𝑜𝑟𝑒𝑖𝑔𝑛𝑜 𝑝𝑒𝑟𝑎𝑡𝑖𝑜𝑛𝑠
𝑁𝑒𝑡 𝑖𝑛𝑐𝑜𝑚𝑒 (𝑙𝑜𝑠𝑠) )
3.1.4 Control variables Firm control variables
26 with a firm’s corporate risk-taking policy (Coles et al. 2006). Firm control variables as well as a CEO specific control will be employed to isolate the effects of my independent variables on the dependent variable. First of all, firm size is included. Studies by Coles et al. (2006) and Francis et al. (2017) suggest that firm size negatively relates to risk-taking, since risk management increases within larger firms. Hence, larger firms are motivated to invest in risk-taking activities (Fernandes et al., 2012). The control variable is operationalized by the natural logarithm of sales (Hall and Murphy, 2002; Coles, 2006; Shen and Zhang, 2013; Croci and Petmezas, 2015), which is a common proxy for firm size. Secondly, I control for growth opportunities. A higher degree of a firm’s growth and investment opportunities is expected to increase investments and thereby increase the R&D intensity. I incorporate a proxy to measure growth and investment opportunities of a firm in a similar fashion as Coles et al. (2006) and Francis et al. (2017) by including the market-to-book ratio.10 Hence, I expect a positive effect of the market-to-book ratio. Thirdly, according to Gao and Chou (2015) the capital structure of a firm influences corporate risk-taking due to the trade-off between tax benefits and costs of times of financial distress. Since the risk-taking proxy is measured by investment in R&D, high levels of leverage can negatively influence the main relationship and should be accounted for (Coles, 2006; Hayes, 2012; Croci and Petmezas, 2015). The leverage ratio is operationalized by total debt divided by the book value of total assets at the end of the fiscal year. The following formulas are created to reflect the definitions of the firm control variables of this study:
𝑆𝐼𝑍𝐸 = ln (𝑡𝑜𝑡𝑎𝑙 𝑠𝑎𝑙𝑒𝑠)
27 𝑀𝑇𝐵 =𝑀𝑎𝑟𝑘𝑒𝑡 𝑣𝑎𝑙𝑢𝑒 𝑜𝑓𝑎𝑠𝑠𝑒𝑡𝑠
𝐵𝑜𝑜𝑘 𝑣𝑎𝑙𝑢𝑒 𝑜𝑓 𝑎𝑠𝑠𝑒𝑡𝑠 𝐿𝐸𝑉 =𝑆ℎ𝑜𝑟𝑡𝑡𝑒𝑟𝑚 + 𝑙𝑜𝑛𝑔𝑡𝑒𝑟𝑚 𝑑𝑒𝑏𝑡
𝐵𝑜𝑜𝑘 𝑣𝑎𝑙𝑢𝑒 𝑜𝑓 𝑎𝑠𝑠𝑒𝑡𝑠
CEO control variable
Extant literature find that a CEO’s personal characteristics impacts the CEO’s view on risk, hence his or her risk appetite. The effect of CEO tenure is a two-edged sword. Longer-tenured CEOs often have more power than newly appointed CEOs (Croci and Petmezas, 2015). Hence, a longer tenure within the position forces the risk appetite of the CEO into the corporate investment policies. Accordingly, a study by Shen & Zheng (2014) finds that tenure has a positive effect on risk-taking behavior. In contrast, a CEO’s tenure is also used as a proxy for risk-adversity. Berger et al. (1997) and Coles et al. (2006) find that CEOs with longer tenures are more likely to avoid risk because the CEO is found to be more entrenched in the firm when he or she has a longer tenure. Following Guay (1999) and Hayes et al. (2012), tenure is defined as the numbers of years the CEO has been in the position.
𝑇𝐸𝑁 = (𝑛𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑦𝑒𝑎𝑟𝑠 𝑡ℎ𝑒 𝐶𝐸𝑂 ℎ𝑎𝑠 𝑏𝑒𝑒𝑛 𝑖𝑛 𝑝𝑜𝑠𝑖𝑡𝑖𝑜𝑛)
3.2 Data and sample
28 companies are trendsetters in executive remuneration. In addition, financial and executive specific data on U.S. firms is retrieved from one of the most tracked stock indexes and will therefore provide more complete and accurate availability of data as compared to global data. Moreover, the time-frame in question is a relevant period due to the experienced growth in stock-option incentives over the past 30 years (Bergstresser and Philippon, 2006; Frydman & Jenter, 2010; Gormley et al., 2013). The data on the variables described above is retrieved from two sources. First, data on delta, vega and tenure is retrieved from Croci and Petmezas (2015)11. Secondly, financial data is retrieved from WRDS Compustat Capital IQ. The datasets are matched by the firm unique GVKEY indicator. Firm-year observations not matching a CEO identification number (co_per_rol) are excluded from the sample. Consistent with prior literature, financial and utility firms are excluded because they are often subject to heavy federal regulations (Low, 2009; O’Connor et al., 2013). Following Coles et al. (2006), Hayes et al. (2012), Gormley et al. (2013) and Boulton et al. (2014), two-digit SIC codes dummies are created to classify industries in order to account for the possibilities of omitted variables. The dataset is based on both cross-sectional and time series, enabling the results to generalize. Finally, an unbalanced panel dataset containing 20,996 firm year observations is constructed. To further deal with outliers, all independent variables are winsorized at the 1st and 99th percentile respectively (Coles et al, 2006; Low, 2009; Croci & Petmezas, 2015).
29
3.3 Methodology
In order perform an empirical analysis on the six hypotheses specified in the previous chapter, I estimate ordinary least square (OLS) regressions based on the unbalanced panel dataset. To mitigate the issues of endogeneity in this analysis, I take the lagged version of all independent continuous variables (Croci & Petmezas, 2015). In order to account for the possibility of omitted variables and to control for industry effects, year and industry fixed effects are included. All variables employed in this thesis are reported in section 3.1.
3.4 Model estimation
In this thesis, I formed several hypotheses. The next step is to test these propositions empirically, where the ordinary least square approach will be used. The following models are used to test for the proposed hypotheses:
Model (1) will assess hypothesis H1a and H1b. In order to test for H1a and H1b, I analyze coefficient 𝛽1 (𝛽2) of model (1) which measures the sensitivity of R&D intensity to changes in vega (delta) respectively.
(𝟏)𝑹&𝑫𝑰𝑵𝑻𝒊𝒕 = 𝜷𝟎 + 𝜷𝟏𝑳𝑵(𝟏 + 𝑽𝑬𝑮𝑨)𝒊𝒕−𝟏+ 𝜷𝟐𝑳𝑵(𝟏 + 𝑫𝑬𝑳𝑻𝑨)𝒊𝒕−𝟏+ 𝜷𝟑𝑫𝑶𝑰𝒊𝒕−𝟏 + 𝜷𝟒𝑺𝑰𝒁𝑬𝒊𝒕−𝟏+ 𝜷𝟓𝑴𝑻𝑩𝒊𝒕−𝟏+ 𝜷𝟔𝑳𝑬𝑽𝒊𝒕−𝟏+ 𝜷𝟕𝑻𝑬𝑵𝑼𝑹𝑬𝒊𝒕−𝟏
30 Consequently, model (2) takes into account interaction effects between the degree of internationalization and vega or delta and R&D intensity and hence test hypothesis H2a, H2b, H2c and H2d. The coefficients 𝛽4 (𝛽5) represent the sensitivity of R&D intensity to vega (delta) for the degree of internationalization a firm experiences.
(𝟐)𝑹&𝑫𝑰𝑵𝑻𝒊𝒕 = 𝜷𝟎 + 𝜷𝟏𝑳𝑵(𝟏 + 𝑽𝑬𝑮𝑨)𝒊𝒕−𝟏+ 𝜷𝟐𝑳𝑵(𝟏 + 𝑫𝑬𝑳𝑻𝑨)𝒊𝒕−𝟏+ 𝜷𝟑𝑫𝑶𝑰𝒊𝒕−𝟏
+ 𝜷𝟒(𝑳𝑵(𝟏 + 𝑽𝑬𝑮𝑨)𝒊𝒕−𝟏∗ 𝑫𝑶𝑰) + 𝜷𝟓(𝑳𝑵(𝟏 + 𝑫𝑬𝑳𝑻𝑨)𝒊𝒕−𝟏∗ 𝑫𝑶𝑰) + 𝜷𝟔𝑺𝑰𝒁𝑬𝒊𝒕−𝟏+ 𝜷𝟕𝑴𝑻𝑩𝒊𝒕−𝟏+ 𝜷𝟖𝑳𝑬𝑽𝒊𝒕−𝟏+ 𝜷𝟗𝑻𝑬𝑵𝑼𝑹𝑬𝒊𝒕−𝟏 + ∑ 𝒚𝒆𝒂𝒓 𝒅𝒖𝒎𝒎𝒊𝒆𝒔 + ∑ 𝒊𝒏𝒅𝒖𝒔𝒕𝒓𝒚 𝒅𝒖𝒎𝒎𝒊𝒆𝒔 + 𝜺𝒊𝒕
In addition, I control for the endogeneity bias12. I focus on 𝛽1 and 𝛽3 (𝛽2 and 𝛽4) of model
(3), which measures either the sensitivity of R&D intensity to changes in the predicted or excess vega (delta) in order to control for endogeneity.
(𝟑)𝑹&𝑫𝑰𝑵𝑻𝒊𝒕= 𝜷𝟎 + 𝜷𝟏𝑬𝑿𝑪𝑬𝑺𝑺 𝑳𝑵(𝟏 + 𝑽𝑬𝑮𝑨)𝒊𝒕−𝟏 + 𝜷𝟐𝑬𝑿𝑪𝑬𝑺𝑺 𝑳𝑵(𝟏 + 𝑫𝑬𝑳𝑻𝑨)𝒊𝒕−𝟏+ 𝜷𝟑𝑷𝑹𝑬𝑫𝑰𝑪𝑻𝑬𝑫 𝑽𝑬𝑮𝑨𝒊𝒕−𝟏 + 𝜷𝟒𝑷𝑹𝑬𝑫𝑰𝑪𝑻𝑬𝑫 𝑫𝑬𝑳𝑻𝑨𝒊𝒕−𝟏+ 𝜷𝟓𝑫𝑶𝑰𝒊𝒕−𝟏+ 𝜷𝟔𝑺𝑰𝒁𝑬𝒊𝒕−𝟏+ 𝜷𝟕𝑴𝑻𝑩𝒊𝒕−𝟏 + 𝜷𝟖𝑳𝑬𝑽𝒊𝒕−𝟏+ 𝜷𝟗𝑻𝑬𝑵𝑼𝑹𝑬𝒊𝒕−𝟏+ ∑ 𝒚𝒆𝒂𝒓 𝒅𝒖𝒎𝒎𝒊𝒆𝒔 + ∑ 𝒊𝒏𝒅𝒖𝒔𝒕𝒓𝒚 𝒅𝒖𝒎𝒎𝒊𝒆𝒔 + 𝜺𝒊𝒕
31
REGRESSIONS AND RESULTS
This section presents the results from the empirical analysis. First, the descriptive statistics are presented. Hereafter, the correlation and endogeneity issues are discussed. Lastly, the results testing the hypotheses are presented and discussed.
4.1 Descriptive statistics
Table 1 Descriptive statistics
This table represents the descriptive statistics of all variables used in the analyses of this study, including the dependent variable, independent variables, the moderator and control variables. Except tenure, all variables are winsorized at the 1st and 99th percentile. This table includes the number of observations, the mean of the sample, the standard deviation of the sample and the values for the variables for the minimum, 25th percentile,
median, 75th percentile and maximum respectively.
Variable Obs Mean SD Min P25 P50 P75 Max
R&D intensity 20,996 0.0352 0.070 0.000 0.000 0.003 0.0442 2.091 Vega ($000s) t-1 20,996 114.980 193.946 0.000 13.074 43.094 123.041 1152.602 Delta ($000s) t-1 20,996 648.188 1526.999 1.181 78.615 199.624 533.666 15562.630 DOI t-1 20,996 0.278 0.364 0.000 0.000 0.055 0.516 1.000 Firm size t-1 20,996 1.932 0.240 1.170 1.791 1.949 2.098 2.405 Market-to-book t-1 20,996 2.076 1.370 0.764 1.242 1.634 2.360 8.566 Leverage ratio t-1 20,996 0.214 0.180 0.000 0.050 0.198 0.326 0.838 CEO tenure t-1 20,996 8.091 7.321 0.000 3.000 6.000 11.000 54.000
32 114.980 (648,188) and the vega (delta) median is 43.094 (199.624). This indicates that the CEOs in this sample enjoy on average approximately an increase of $114.980,- in his/her wealth for a 1% increase in stock volatility, whereas his or her wealth will enjoy an increase of $648.188,- for a 1% increase in stock price. The means of vega and delta are consistent with the range described in the studies by Coles et al. (2006), Croci and Petmezas (2015) and Francis et al. (2017) where vega (delta) take a mean value between 80 - 149 (600-824). Moreover, the standard deviation is about 1.7(2.3)x the mean of vega (delta), consistent with the previous mentioned studies. All in all, it can be concluded that the descriptive statistics of vega and delta are consistent with extant literature.
As for the firm control variables (market-to-book ratio, leverage ratio and size) values are comparable to previous mentioned studies, without any outliers . The average tenure of the CEOs in this sample is 8 years with a median of 6 years. From this, one can conclude that the CEOs are relatively long tenured within their position, meaning that the vega and delta are computed over wealth accumulated over a relatively long period (Croci and Petmezas, 2015).
33
4.2 Correlation matrix and endogeneity concerns
4.2.1 Correlations
Table 2 Correlation matrix
This table represents the information about the correlation of all variables used in the analyses of this study, including the dependent variable, independent variables, the moderator and control variables. Except tenure,
all variables are winsorized at the 1st and 99th percentile. * denotes statistical significance at the 1% level RDINT VEGA t-1 DELTA t-1 DOI t-1 SIZE t-1 MTB t-1 LEV t-1 TEN t-1 RDINT 1.000 VEGA t-1 -0.005 1.000 DELTA t-1 -0.024* 0.368* 1.000 DOI t-1 0.064* 0.170* 0.023* 1.000 SIZE t-1 -0.361* 0.433* 0.217* 0.223* 1.000 MTB t-1 -0.276* 0.087* 0.0247* -0.071 -0.228* 1.000 LEV t-1 -0.214* 0.042* -0.040* -0.014 -0.250* -0.249* 1.000 TEN t-1 -0.012 0.011 0.257* -0.065* -0.069* 0.028* -0.054* 1.000
Table 2 represents the Pearson correlation matrix describing relationships between the dependent, independent, moderating and control variables used in the regression models. This matrix is included in order to identify potential multicollinearity, using the threshold of a correlation of 0.5 (-0.05). In case of multicollinearity, two or more explanatory variables have a high correlation which could lead to biases in the regression results, since the impact of incremental changes could be magnified in the model. Since the coefficients take a value smaller than 0.5, one can assume that multicollinearity is not an issue in this thesis.
4.2.2 Endogeneity concerns
34 difficult to overcome (Low, 2009; Gormley et al., 2013). Gormley et al. (2013) indicate that “.. identifying the causal effect of option-based incentives on corporate risk-taking is difficult given the obvious endogeneity of the relationship between options and risk. Because managers' compensation is arguably designed in anticipation of a particular risk environment, the possibility of reverse causality is hard to exclude.” (p. 80). So, one can assume that risk-taking incentives and R&D intensity are endogenously related and hence could give biased estimates.
In addition, vega and delta are expected to be related as well. The CEO compensation package consists out of a variety of observable, as well as unobservable facets of the firm (Shen and Zhang, 2013). As mentioned in the previous chapter, vega is partially used as a remedy to the delta-induced risk aversion and hence depend on each other: they are both elements of the same compensation package and therefore jointly determined. O’Connor (2013) finds that, even when the risk-taking incentives variables are predetermined and thus unrelated to the past, models are likely to suffer from endogeneity issues.
35 I follow Coles et al. (2006) and Shen and Zhang (2013) in selecting the (due limited data availability) best available determinants13 of vega and delta: the market-to-book ratio, firm size, leverage and tenure. Theory predicts that bigger firms tend to compensate their CEOs with more equity-based (option) compensation, since larger firms require more talented managers and therefore will attract them with more equity-based compensation (Smith and Watts, 1992). Moreover, within larger firms, CEOs have to deal with a greater number of responsibilities and should be compensated for this. A higher market-to-book ratio indicates a undervalued firm and thus more investment opportunities, hence firms should grant their CEOs more equity-based compensation in order to encourage the risk-taking behavior.14 Firms with lower levels of leverage are expected to have more freedom to invest and hence grant their CEOs more equity-based compensation. Hence, I used an OLS approach15 to regress vega and delta separately on their determinants and created the predicted vega (delta) and the excess vega (delta), of which the latter refers to the residuals from the regression model16. I find that the relationships between the coefficients of the determinants of vega and delta and the predicted versions of vega and delta are consistent with my expectations.
13 Hence, these determinants are the second-best selection of determinants due to data limitations
14 Firms with rich growth opportunities will be hurt the most if managers are overly risk averse and underinvest in risky, but value-adding projects. Hence, firms with more growth opportunities often grant more equity incentives, otherwise CEOs would be inclined to take on negative NPV projects (Armstrong and Vishishtha, 2012).
36
4.3 Empirical results and discussion
This table presents the OLS regression result for model 1. The sample period is from 1992 - 2013. Definitions of all variables are provided in the methodology. All models use year and industry dummies.
Except tenure, all variables are winsorized at the 1st and 99th percentile. Standard errors are denoted in bracket. ***, **, * denotes statistical significance at the 1%, 5% and 10% levels respectively.
Dependent variable: R&DINT
Model 1a Model 1b Model 1c ln (1+VEGA) t-1 0.005*** [0.000] 0.006*** [0.000] ln(1+DELTA) t-1 0.001*** [0.000] -0.002*** [0.000] DOI t-1 0.004*** [0.001] 0.006*** [0.001] 0.004*** [0.001] SIZEt-1 -0.097*** [0.002] -0.083*** [0.002] -0.094*** [0.002] MTB t-1 0.007*** [0.000] 0.007*** [0.000] 0.007*** [0.000] LEV t-1 -0.016*** [0.003] -0.016*** [0.003] -0.017*** [0.003] TEN t-1 -0.000*** [0.000] -0.000*** [0.000] -0.000 [0.000] Constant 0.173*** [0.006] 0.0153*** [0.006] 0.169*** [0.006]
Year FE Yes Yes Yes
Industry FE Yes Yes Yes
Adjusted R2 0.329 0.318 0.329
Observations 20,996 20,996 20,996
38 presents a positive relationship coefficient of 0.001 between delta and R&D intensity at a 1% significance level. This is not in line with the expected negative relationship. However, the positive coefficient could potentially be explained by the agency theory. In the classical agency theory, equity-based compensation is proposed to align the interest of the CEO with the interests of its shareholder. Therefore, a stand-alone delta (not controlled for vega) should align the risk appetites of the principle with the agent. This means that, it could be argued that higher levels of sensitivity to stock price, thus a higher amount of stock in the portfolio of the CEO, increasingly aligned the principle with the agent and hence led to more risk-taking and thereby increased R&D intensity.
However, the results of model 1a and 1b have relatively low economic relevance since vega and delta are jointly determined in the compensation package (Coles et al., 2006; Armstrong and Vishishtha, 2012; O’Connor et al., 2013). Therefore, only model 1c will test for hypothesis H1a and H1c, since this model provides results including both vega and delta and hence has stronger economic meaning. First of all, compared to model 1a, it is observed that the effect of vega on R&D intensity becomes stronger with an increase in the value of the coefficient of 0.00117 and can be concluded that vega enjoys a larger effect on R&D intensity when controlled for delta. On the contrary, when estimating the coefficient of delta in model 1b (0.001) and model 1c (-0.002) and controlling for vega, the signs of the coefficients change, indicating the important relationship vega and delta have to each other. In the jointly determined model, the delta coefficient indicates a negative relationship with R&D intensity, corresponding to my expectations.
Based on the results in model 1c, I can state that a higher vega leads to higher values of R&D intensity with a relation coefficient of 0.006 meaning that an increase in of 1% in the natural
39 of 0.6%18 at a significance level of 1%. In addition, delta takes a value of -0.002 at a significance
level of 1%, meaning that an increase of 0.01 in the natural logarithm of a CEO’s stock price sensitivity leads to a decrease in R&D intensity of 0.2%.19
In model 1c, the degree of internationalization holds a positive relationship with R&D intensity (0.004) at a 1% significance level, indicating that firms with a higher degree of internationalization enjoy higher levels of R&D intensity. Theoretically, this results corresponds to a range of theories indicating a positive relationship between the DOI and R&D intensity.
This model includes several control variables in order to control for the forces that drive both vega and delta together with corporate risk-taking policy (Coles et al. 2006). Considering model 1c, firm size assumes a negative relation to R&D intensity (-0.094) corresponding to my expectations. The market-to-book ratio takes on a positive relation with R&D intensity (0.007) at a 1% significance level. This corresponds with the theoretical predictions: a higher market-to-book ratio indicates a undervalued firm and thus more investment opportunities. The leverage ratio presents a negative relationship with R&D intensity (-0.017), as expected. Higher degrees of leverage will indicate less freedom to invest and hence decreases risk-taking. As for tenure, I did not find any relationship, meaning that in this model, tenure does not affect the level of risk-taking. The adjusted r-squared in model 1c is 0.337, which means that 33,7% of the variance is explained by the model. Studies by Coles et al. (2006) and Hayes et al. (2012) show a similar levels of adjusted r-squared and hence this value is accepted as an adequate level.
41 Table 4 presents the regression of model 2, which incorporates the moderating effect of degree of internationalization (DOI) into the analysis. In order to test for the influence of the DOI in the main relationship between vega or delta and R&D intensity, I test for interaction effects. The coefficients of the interaction effects (henceforth referred to as VEGA*DOI or DELTA*DOI) covers the simultaneous effect of vega (or delta) and the DOI on R&D intensity.
This table presents the OLS regression result including the moderating effect of the degree of internationalization, model (2). The sample period is from 1992 - 2013. Definitions of all variables are
provided in the methodology. All models use year and industry dummies. Except tenure, all variables are winsorized at the 1st and 99th percentile. Standard errors are denoted in bracket. ***, **, * denotes
statistical significance at the 1%, 5% and 10% levels respectively. Dependent variable: R&DINT
Model 2a Model 2b Model 2c ln(1+VEGA) t-1 0.005*** [0.000] 0.006*** [0.000] ln(1+DELTA) t-1 -0.003*** [0.000] -0.003*** [0.000] DOIt-1 -0.000 [0.003] -0.019*** [0.000] -0.019*** [0.000] ln(1+VEGA)*DOI t-1 0.001* [0.001] -0.001 [0.001] ln(1+DELTA)*DOI t-1 0.006*** [0.001] 0.005*** [0.001] SIZEt-1 -0.094*** [0.002] -0.091*** [0.002] -0.091*** [0.002] MTB t-1 0.007*** [0.000] 0.007*** [0.000] 0.007*** [0.000] LEV t-1 -0.018*** [0.003] -0.019*** [0.003] -0.019*** [0.003] TEN t-1 -0.000* [0.000] -0.000* [0.000] -0.000* [0.000] Constant 0.168*** [0.006] 0.157*** [0.006] 0.170*** [0.006]
Year FE Yes Yes Yes
Industry FE Yes Yes Yes
Adjusted R2 0.341 0.342 0.342
42 interaction term is due to the interaction and not due to the DOI itself. In model 2c the DOI coefficient (-0.019) projects a negative and very significant direct relationship to R&D intensity. The directional sign is different from model 1, indicating that when the interaction effects VEGA*DOI and DELTA*DOI are incorporated into the model, the DOI exerts a negative influence on R&D intensity. Furthermore, the coefficients of the effect of the main independent variables in relation to R&D intensity are consistent with model 1, where vega enjoys a positive relationship and delta is negatively related. The control variables firm size, market-to-book, leverage and CEO tenure remain consistent with model 1 and are therefore consistent with the extant literature.
Model 2a (2b) show the results of the regression that includes the moderating effect of VEGA*DOI (DELTA*DOI) in the relationship between vega (delta) and R&D intensity respectively. In model 2a, the coefficient of the interaction effect VEGA*DOI is positive (0.001) and weakly significant at the 10% level. This indicates that model 2a finds weak support for the observation that the impact of vega (when not controlled for delta) on R&D intensity is greater for firms with higher degrees of internationalization. Model 2b finds a strongly significant relationship between the coefficient of the interaction effect DELTA*DOI and R&D intensity (0.006), indicating that the impact of delta (when not controlled for vega) on R&D intensity is greater for firms with higher degrees of internationalization.
43 H2a and H2c respectively.
As for DELTA*DOI, the coefficient of the interaction effect in model 2c yields a highly significant result at the 1% significance level, indicating a marginal effect of 0.006. This implies that effect of delta is higher for firms with a higher degree of internationalization. Economically this means, that when firms enjoy higher degrees of internationalization, delta will have a stronger relationship with R&D intensity. However, in model 2c the coefficient of delta on R&D intensity is -0.003 meaning that the actual effect of delta on R&D intensity, when moderated by the degree of internationalization, takes a coefficient of 0.00320. Hence, this would mean that for firms with a higher degree of internationalization, an increase in delta will actually project a positive effect on R&D intensity. Therefore, I can reject the null-hypothesis of H2d and state that the degree of internationalization weakens the hypothesized relationship between delta and R&D intensity. Subsequently, H2c is rejected. Concluding, the isolated DOI has a negative and significant effect on R&D but when firms enjoy higher degrees of internationalization, higher level of delta will increase R&D intensity significantly.
45 Table 5
Model 3: Excess vega and excess delta
This table presents the OLS regression result for model 3. The sample period is from 1992 - 2013. Definitions of all variables are provided in the methodology section. All models use year and industry dummies. Except tenure, all
variables are winsorized at the 1st and 99th percentile. Standard errors are denoted in bracket. ***, **, * denotes statistical significance at the 1%, 5% and 10% levels respectively.
Dependent variable: R&DINT
Model 3a Model 3b Model 3c Model 3d Model 3e Model 2f EXCESS ln(1+VEGA)t-1 0.006*** [0.000] 0.006*** [0.000] 0.006*** [0.000] 0.006*** [0.000] EXCESS ln(1+DELTA)t-1 0.001** [0.000] -0.002** [0.000] 0.001** [0.000] -0.001 [0.000] PREDICTED ln(1+VEGA) t-1 0.006*** [0.001] 0.052*** [0.002] PREDICTED ln(1+DELTA) t-1 -0.020*** [0.002] -0.100*** [0.004] DOI t-1 0.004*** [0.001] 0.006*** [0.001] 0.004*** [0.001] 0.004*** [0.001] 0.006*** [0.001] 0.004*** [0.001] SIZEt-1 -0.076*** [0.002] -0.076*** [0.002] -0.075*** [0.002] -0.006*** [0.009] -0.093*** [0.005] -0.105*** [0.010] MTB t-1 0.007*** [0.000] 0.007*** [0.000] 0.007*** [0.000] 0.009*** [0.000] 0.006*** [0.001] 0.003*** [0.001] LEV t-1 -0.018*** [0.003] -0.018*** [0.003] -0.018*** [0.003] -0.018*** [0.003] -0.016*** [0.003] -0.009*** [0.003] TEN t-1 -0.000*** [0.000] -0.000*** [0.000] -0.000*** [0.000] -0.000*** [0.000] -0.001*** [0.000] -0.005*** [0.000] Constant 0.144*** [0.006] 0.144*** [0.006] 0.143*** [0.006] 0.057*** [0.012] 0.159*** [0.007] 0.140*** [0.015]
Year FE Yes Yes Yes Yes Yes Yes
Industry FE Yes Yes Yes Yes Yes Yes
Adjusted R2 0.342 0.328 0.343 0.344 0.329 0. 358 Observations 20,996 20,996 20,996 20,996 20,996 20,996
46 bias (Shen and Zhang, 2013). The excess vega and excess delta in this model represent the component of a CEO’s vega or delta orthogonal to the determinants of a CEO’s compensation and is therefore isolated from its determinants. Model 3a, 3b and 3c represent the model when only incorporating the excess components in the analysis. In model 3a, 3b and 3c, the excess vega and excess delta project significant results with signs in similar directions as found in model 1a, 1b and 1c, consistent with my previous findings.
To interpreted the results from model 3 and hence draw conclusions on the endogeneity in this thesis, I compare the directional signs of the coefficient of the predicted vega (delta) with the coefficient of the excess vega (delta). In model 3d and 3f, both the predicted as well as the excess vega enjoy a positive relation to R&D intensity, meaning that that the components of vega that are orthogonal to its determinants do have explanatory power In model 3f, both predicted as well as excess delta carry a positive sign, inferring that the components of delta that are uncorrelated to the explanatory variables, have explanatory power. However, the excess delta does not find statistical support and therefore no conclusions about delta could be drawn. On the contrary, in model 3e where delta is isolated from vega, the excess and predicted delta project dissimilar directional signs. In this case, it would mean that significant support is found that the components of delta that are orthogonal to the other independent variables do not have explanatory power.
47 vega, delta and R&D intensity.
4.4.2 Alternative proxy for risk-taking
Table 6 Model B CAPEX
This table presents the OLS regression result for model (B). The sample period is from 1992 - 2013. Definitions of all variables, except for CAPEX, are provided in the methodology section. The definition of CAPEX is provided in Appendix B. All models use year and industry dummies. Except
tenure, all variables are winsorized at the 1st and 99th percentile. Standard errors are denoted in bracket. ***, **, * denotes statistical significance at the 1%, 5% and 10% levels respectively.
Dependent variable: CAPEX
Model (B)1 Model (B)2 Model B(3) ln(1+VEGA) t-1 -0.001*** [0.000] -0.002*** [0.000] ln(1+DELTA) t-1 0.002*** [0.000] 0.004*** [0.000] DOI t-1 -0.008*** [0.001] -0.008*** [0.001] -0.008*** [0.001] ln(1+VEGA)*DOI t-1 ln(1+DELTA)*DOI t-1 SIZEt-1 0.005** [0.002] -0.007*** [0.002] -0.003 [0.002] MTB t-1 0.005*** [0.000] 0.004*** [0.000] 0.004*** [0.000] LEV t-1 -0.023*** [0.002] -0.021*** [0.002] -0.021*** [0.002] TEN t-1 0.000* [0.000] -0.000 [0.000] -0.000*** [0.000] Constant 0.085*** [0.006] 0.098*** [0.006] 0.092*** [0.006]
Year FE Yes Yes Yes
Industry FE Yes Yes Yes
Adjusted R2 0.334 0.346 0.348
48 case, R&D intensity is replaced by capital expenditures. Capital expenditures, hence referred to as CAPEX, are tangible and hence perceived as less risky (Hölstrom, 1989; Kothari et al., 2002; Manso, 2011). Therefore, Coles et al. (2006) proposes that increasing degrees of CAPEX decrease risk-taking. Hence, regressing vega and delta on CAPEX should give coefficients with relational signs in the opposite directions as compared to R&D intensity. The output of the robustness check is presented in table 622. In model B1 and model B3, I finds that vega carries a negative coefficient
in its relation to CAPEX, meaning that higher degrees of vega will decrease the level of CAPEX, consistent with my predictions. Moreover, delta presents a positive relationship to CAPEX in model B2 and model B3, indicating that increasing levels of delta will make the CEO more risk-averse and hence will increase less risky investments such as CAPEX. The control variables project consistent predictions in its relation to CAPEX: the larger the growth opportunities (as presented by the MTB), the more investments, also increasing the level of CAPEX. Moreover, higher levels of leverage are negatively related to CAPEX, making economically sense since there is limited freedom to invest in general, with no exception to capital expenditures. This robustness check adds additional support to my previous findings regarding H1a and H1b.
From the robustness checks it can be concluded that the residual values of vega but the results on delta remain unsupported. Therefore, I can conclude that endogeneity remains an issue in this thesis. Moreover, robustness check considering CAPEX as the dependent variable finds significant support for the main hypothesis.
21 Please find the equation for model (B) and the corresponding descriptive statistics in Appendix B
49
5.1 Conclusion
The purpose of this thesis is to provide empirical evidence on the relation between two risk-taking incentives, vega and delta, and corporate risk-taking. In addition, it examines the unexplored moderating role of the degree of internationalization.
Corporate risk-taking is critical for a firm’s survival. Hence, shareholders demand adequate levels of corporate risk-taking. However, whereas the risk-neutral shareholder is able to diversify away any firm-specific risk, the CEO can only put his or her effort into one job leaving him or her relatively undiversified and more dependent on the firm. Hence, an agency conflict between the risk-averse CEO and the risk-neutral shareholder arises. Consequently, it is important that corporate governance mechanisms motivate the risk-averse CEO to set the optimal risk-taking policies and to incentivize him or her to make the correct value-critical investment choices. Risk-taking policies influence investments in R&D, since R&D expenditures are perceived as investments bearing relatively high risk.
50 including both stock and stock options, I find statistical evidence that increased levels of delta will refrain the CEO from committing to higher levels of risk-taking. I theorize that this is due to the increase in idiosyncratic risk he or she experiences. Consequently, the risk appetite of the shareholder and delta-induced CEO can be aligned by adding convexity into the pay-off structure of the CEO. Hence, by means of increasing vega, the CEO wealth is only tied to the upside of the stock options and therefore the CEO experiences a decrease in idiosyncratic risk. Hence, in this thesis I find empirical evidence that when a CEO experiences a higher vega, he or she will be motivated to increase corporate risk-taking which is in favor of the shareholder.
When firms are internationally diversified, they are expected to be more financially flexibly. Hence, the international firm is able to borrow at cheaper rates, is less sensitive to the domestic capital market and is also less dependent on the domestic capital market. Consequently, firms are able to obtain capital resources at lower costs, hence increasing the Net Present Value (NPV) of projects and decreasing corporate risk. Due to the risk premium the shareholders face, they are motivated to increase corporate risk. Hence, I expect the shareholders could alter the level of the risk-taking incentives to align the CEO’s risk appetite to theirs. I examined whether higher degrees of internationalization will exert influence the effect of vega and delta on R&D intensity. In this thesis, I find support for the claim that firms with a higher degree of internationalization will experience a higher delta, positively influencing the level of risk-taking. On the contrary, I do not find support for the claim that the degree of internationalization will influence the relation between vega and corporate risk-taking.
51 First of all, managers should a take into account that CEO risk-taking incentives are effective corporate governance tools within a firm which can be altered by the board of directors. Hence, by means of these tools, a firm can reduce the agency conflict between the risk-averse CEO and the shareholder. This is important for the firm’s long-term growth. By means of creating an optimal balance between the risk-taking incentives, a firm is better able at controlling the level of corporate risk-taking. Moreover, the relation between risk-taking incentives and risk-taking is very relevant since equity-based option within the CEO compensation package has seen a large growth in the past decades. Consequently, CEOs are nowadays increasingly sensitive to stock return volatility. Managers should take into account that this increase could potentially lead CEOs to take inadequate levels of risk-taking, such as excessive risk-taking. This can have major consequences for the survival of firm, but also for (financial) markets on a global magnitude (e.g. the global financial crisis). Lastly, managers can derive from this thesis that firms with a higher degree of internationalization experience different effects from the risk-taking incentives mechanism as compared to firms only operating in the U.S. and keep in mind that these results can differ for non-U.S. firms.
52 options are an effective tool of motivating managers to alter their risk incentive behavior.
5.2 Limitations and future research
Despite the contributions of this thesis, it faces some limitations too. First of all, this study encountered, like many other studies, difficulties in data availability. As a result, limitations in calculating vega and delta aroused. Hence I used a pre-calculated vega and delta with observations until the fiscal year 2013. The relevance of this thesis could be improved by the use the most recent data available. Moreover, due to this limited data availability and the with coming limitations in calculating vega and delta, this study is based on U.S. firms only. Therefore, one has to be careful with generalizing results since U.S. firm. When compared to firms from other countries, U.S. firms might possess unique country-specific features influencing the main relationship of this study. Another limitation to this study is my proxy for risk-taking. R&D intensity is a solid proxy, but does not encompass all the risk-taking of a company. Lastly, but perhaps most important limitation to my thesis is the level of endogeneity the models face. Although similar studies encounter the same issues and I attempted to correct for these concerns, endogeneity issues will remain and need to be kept in mind.
