Firm performance, CEO compensation, and managerial
risk-taking: empirical evidence
Master’s Thesis
International Financial Management
January 2020
Author: Thom Boelens
1Supervisor: Dr. N. Selmane
Co-assessor: Dr. A. de Ridder
This study examines the effect of CEO compensation and risk-taking incentives on corporate investment and debt policy using a sample of 1,255 U.S. firms in the 2007-2018 period. It provides evidence that higher sensitivity between CEO wealth and stock price volatility (vega) is associated with higher R&D expenditures and lower capital expenditures. This finding provides empirical support to the notion that CEOs implement risk-associated corporate policies based on their personal stock and option holdings in the firm. Compared to findings from before the 2008 financial crisis, the influence of vega on corporate policy decisions has weakened. Finally, this paper finds that CEO compensation is mostly determined based on stock returns and less on ROA, and that the degree of a firm’s internationalization is positively associated with the level of CEO compensation.
Field Key Words: Executive compensation; Managerial incentives; Risk taking;
Risk-taking incentives; Investment opportunities; Firm performance
1Author: Thom Willem Johannes Boelens (s2755173) Master Student International Financial Management.
2
1
Introduction
There is a substantial amount of literature on Chief Executive Officer (CEO hereafter)
compensation. This thesis combines two strands of CEO compensation literature in order to
analyze the topic in a comprehensive way. The first strand investigates the effects of firm
performance on the level and structure of CEO compensation. The second strand analyzes how
CEO compensation affects managerial risk-taking. Surprisingly, few studies investigate these
relationships after the 2007-2008 financial crisis. In particular, I contribute to the literature by
analyzing these two strands of literature in a contemporaneous timeframe (2007-2018).
CEO compensation has always been an important and controversial topic of debate. It
has always been important, because the day-to-day decisions of the CEO heavily influence the
firm’s financial performance and its business strategy. For publicly traded firms these decisions should be aimed at increasing the market value of the firm. However, CEOs can also make
decisions that do not increase market value and instead make decisions that benefit their own
wealth, at the expense of shareholders (Mehran, 1995). The Board of Directors, who work on
behalf of the shareholders, therefore go to great lengths to design CEO compensation in such
a way that the self-interests of the CEO align to that of shareholders and to mitigate the natural
risk-aversity of the CEO, in order to increase the market value of the firm (John, Litov, &
Yeung, 2008). CEO compensation has also been controversial. It is often seen by critics as
“excessive” because of the supposedly weak link between firm performance and compensation, and the fact that powerful CEOs can extract very high levels of pay (Hill, Lopez, and Reitenga,
2016). Although CEO compensation levels are high across the globe, they are especially high
in the U.S. In 2018 median U.S. CEO compensation was $5.3 million, while this was $4.5
million in Europe and $3.2 million in Canada (Ludwig, 2019). Between 1978 to 2014,
3
than double the stock market growth in that same period (Mishel & Davis, 2015). What’s more,
U.S. CEO compensation grew much more than the top 0.1 wage earners in the U.S., suggesting
that the increase in pay does not only reflect “market for talent” mechanisms at work, but rather that CEOs are powerful enough to extract “rents” and negotiate a substantially higher pay than they would’ve gotten in a free market (Mishel and Davis, 2015).
Critics also argue that the level and structure of CEO compensation encourages
excessive risk-taking. For example, one of the causes of the Enron accounting scandal and its
subsequent bankruptcy is attributed to the way executives and other employees at the company
were compensated. (Dharan and Bufkins, 2008). As a reaction to Enron’s and other accounting
scandals, the Sarbanes-Oxley act was put into U.S. federal law in 2002, which – among others,
made strict guidelines for reporting stocks and stock options as executive compensation
(Volcker and Levitt, 2004). Furthermore, FAS 123R was introduced in 2005 by the U.S.
Securities and Exchange Commission in order to increase the transparency of director and
executive compensation (Hayes, Lemmon, and Qiu, 2012).
Prior literature generally finds a positive relationship between firm performance,
measured as market- or accounting-based performance (i.e. Conyon, 2006; Conyon and
Murphy, 2000; Jensen and Murphy, 1990; Rosen, 1990). The second strand of literature, which
has received less attention, analyzes how the sensitivity between CEO wealth and stock price
volatility relates to executives’ risk choices (e.g. Armstrong, Larcker, Ormazabal, and Taylor, 2013; Chava & Purnanandam, 2010; Coles, Daniel, & Naveen, 2006). Prior literature has
shown that CEO compensation structure does affect managerial risk-taking. To analyze this
relation in a contemporaneous timeframe, I construct a very direct measure of CEOs’ incentives
to increase firm risk given their personal stocks and options in their firm, called vega. I also
construct delta, the sensitivity between CEO wealth and stock price, as a proxy for managerial
4
After covering the two strands of literature described above, I can provide empirical
evidence to answer the main research question of this paper, which is:
“How are U.S. CEOs rewarded for increasing firm performance, and to what extent are they incentivized to make risk-increasing corporate decisions?”
This paper makes several contributions to the academic field. This paper is one of the few
studies that measures risk incentives in a contemporaneous timeframe using vega and delta.
Almost all studies use relatively simple and rudimentary proxies of managerial incentives, such
as the value or number of options granted, stock held, stock granted, etc. Using such proxies
of managerial incentives provides only a limited estimation of incentives faced by executives
(Coles et al., 2006). Computing vega allows to accurately determine the effect of CEO
compensation and CEO option holdings on risk-taking, because vega accounts for the
multidimensional characteristics of option valuation, such as time to maturity, stock price
volatility, exercise price and dividend payouts.
In addition to the contributions in the academic field described above, this research has
several managerial implications. First, corporate boards, shareholders, and regulatory bodies
pay close attention to the relation between compensation and managerial behavior. However,
less attention seems to be paid to CEOs’ wealth, and how it affects their behavior and decision
making. This paper gives more insight into how executives’ personal wealth can affect the
risk-taking preferences of CEOs. Second, this thesis can help compensation committees to find a
better balance between firm performance, executive compensation, and risk preferences.
The findings indicate that CEOs receive their compensation predominantly based on
stock performance, and less based on accounting-based performance, i.e. ROA. Furthermore,
5
CEOs with higher vega implement more R&D expenditures and lower capital expenditures.
These results indicate that CEOs with higher vega move resources away from risk-averse assets
towards riskier assets, underscoring the long-held belief that CEOs act in their own
self-interest, based on their personal stocks and options in the firm and given how their corporate
policy decisions affect the value of their own wealth.
The remainder of this study is organized as follows. Section 2 outlines the literature and the
hypotheses. Section 3 describes the sample collection and descriptive statistics and section 4
explains the methodology. Section 5 analyzes the results on the CEO pay-performance analysis
and how it relates to internationalization. Section 6 provides the results for the CEO
compensation and risk-taking analysis. Concluding remarks are given in section 7 by
summarizing the main findings, contributions, limitations and directions for future research.
2
Literature and hypotheses
This section provides a synthesis of the literature surrounding the main areas of research to
which this thesis aims to contribute. I cover the relevant literature on executive compensation,
firm performance, internationalization, and managerial risk-taking.
2.1 The agency problem and executive compensation
The agency problem between managers and shareholders is extensively studied in public firms
where ownership and control are separated (Bebchuk and Weisbach, 2010). In the context of
my thesis, the agency problem results from conflicting interests and incentives between the
CEO (the agent) and shareholders (the principal), who are both value-maximizing parties. The
6
wealth, instead of maximizing firm and shareholder value. Because managers are
self-interested, they could make strategic choices that would benefit them personally at the expense
of shareholder value. In other words, managers of public firms with external shareholders make
strategic choices that result in lower firm value than it would be if managers were the only
owners of the firm (i.e. Haugen & Senbet, 1981; Jensen & Meckling, 1976).
The agency theory has important implications on executive compensation and there are
two main views on this relation. First, supporters of the ‘optimal contracting view’ suggests
that executive compensation can be a (partial) solution to the agency problem. This view
emphasizes that managers are subject to agency problems, and that managers do not
automatically want to maximize shareholder value (Bebchuk and Weisbach, 2010). Boards
therefore attempt to give executives a cost-effective incentive through arm’s length contracting,
in order to maximize shareholder wealth. Second, the ‘managerial power view’ argues that
executive compensation is not a potential tool to solve the agency problem, but rather that it is
part of the problem itself. Supporters emphasize that various pay arrangements enable
managerial rent-seeking rather than providing efficient incentives. For example, Fama and
Jensen (1983) argue that the board often does not do a good job at monitoring top management.
Executives therefore could negotiate compensation packages substantially larger, and less
proportionate to firm performance, than what they would have gotten if they were received via
arm’s-length contracting (Bebchuck and Weisbach, 2010). Research in fact shows that characteristics of the board affect the level and structure of executive compensation. For
example, Core and Guay (1999) and Conyon and Murphy (2000) find that a significant positive
relationship between board size and CEO compensation. The composition of the board also
impacts CEO compensation, as a higher degree of outside directors is related to higher CEO
7
Executive compensation can be used to solve agency problems, by aligning executives’
interests to that of shareholders. These interests are aligned by rewarding desirable managerial
behavior that increases shareholder value and aligns the risk orientation of both parties. The
Board of Directors is ultimately responsible for determining the level and structure of executive
compensation on behalf of the shareholders. The compensation of a typical executive often
consists of several components. First, executives receive a base salary that is not tied to
personal or firm performance, and an annual bonus plan that is paid based on single-year
personal and firm performance. Second, they receive equity-based compensation in the form
of stocks and options, and finally, other forms of compensation such as restricted stocks,
long-term incentive plans and retirement plans (Murphy, 1999).
In short, executives do not have the same interests as shareholders by nature. Executives
don’t necessarily want to maximize shareholder value, but they are more likely to do so when it benefits them personally. Moreover, executives are naturally risk-averse which can result in
missed opportunities to increase firm value. Executive compensation can therefore be used to
incentivize desirable behavior. Some argue, however, that powerful executives can negotiate
higher compensation that is not necessarily proportionate to the performance they realize.
2.2 Executive compensation and firm performance
Increasing the sensitivity of executive compensation to firm performance and
shareholder wealth can create incentives for managers to increase shareholder value and wealth
(Coles et al., 2006). An effective way to accomplish this is with equity-based compensation.
By compensating executives in equity, using stocks and options, executives become part-owner
of the firms they manage. The rationale is that executives who receive a significant portion of
their compensation in equity work harder or more effectively to increase firm- and shareholder
8
Jensen and Murphy (1990) find a statistically significant relationship between CEO
compensation and firm performance. Using a sample of 1,049 firms from 1973-1986, they find
that for every $1,000 change in shareholder wealth, CEO pay- and stock related wealth changes
with $3.25 in the same direction. They explain this relatively small sensitivity by hypothesizing
that public and private forces impose constraints that reduce the sensitivity. Considering that
the sensitivity in their research is $1.85 for large firms, and $8.05 for small firms, this could be
true because large firms are often more exposed to public and private forces. Rosen (1990)
finds for large U.S. firms an accounting-based pay-for-performance elasticity of 1.0, meaning
a 10% increase in ROA leads to a 10% increase in executive pay. When replacing ROA with
stock market returns this sensitivity is much smaller, namely 0.1, meaning a 10% increase in
stock return translates to a 1% increase in total executive compensation. Ozkan (2011) also
finds a positive association between firm performance and CEO compensation in a sample of
390 UK nonfinancial firms in a sample from 1999-2005. Pay- performance elasticity for CEOs
is 0.095 for total direct compensation, indicating that a 10% increase in shareholder return
corresponds to an increase of 0.095% in CEO total direct compensation.
In addition to ROA and market-based performance measures as determinants of
executive compensation, firm size is also a determining factor of executive compensation. CEO
compensation is determined in a competitive talent market, and the size of that market is
determined by the size of the firms to which these CEOs are related (Gabaix and Landier,
2008). For example, between 1980 and 2003 CEO compensation increased by a six-fold, which
can be fully attributed to the six-fold increase in market capitalization of large U.S. firms in
that period according to Gabaix and Landier (2008). Using the 2007-2009 financial crisis as an
exogenous shock in a later research, these findings still stand (Gabaix, Landier, and Sauvagnat,
9
There are some papers that explore the effect of CEO compensation on firm
performance. For example, using a sample of 153 randomly selected manufacturing firms in
the 1979-1980 period, Mehran (1995) finds statistically significant and positive coefficients
between CEOs equity based compensation as an independent variable, and both Tobin’s Q and ROA as dependent variables. These results can indicate that firms where CEOs’ compensation
is sensitive to firm performance, those CEOs produce better firm outcomes and higher returns
for shareholders than firms where the sensitivity between CEO compensation and firm
performance is weak. Mehran (1995) argues that this supports Jensen and Meckling (1976),
who argue that CEOs incentives to work harder increases when their stake in the firm rises.
However, the positive relationship between compensation and firm performance found by
Mehran (1995) is not necessarily causal in nature, because the relation has a dual causality.
Firms that have higher performance give their CEO higher compensation, and firms that give
out higher compensation have higher performance on average (Jensen and Meckling, 1976).
Because previous research generally finds a positive relationship between firm
performance and executive compensation, I hypothesize the following:
Hypothesis 1: Higher firm performance is associated with higher CEO compensation
2.2.1 Executive compensation and firm internationalization
This paper refers to internationalization as the degree of firms’ foreign involvement, its
dependence on international customers, markets, production factors, the ability to create value
abroad, and to the geographical dispersion of such dependencies (Sanders and Carpenter,
1998). I measure internationalization as foreign sales intensity, calculated as the ratio of foreign
10
There are numerous reasons why firms with more international activities compensate
their CEOs differently than firms that do not operate internationally. Increasing the degree of
foreign operations increases managerial complexity of running the firm. As a firm moves
beyond the borders of its domestic country complexity increases, because it encounters new
regulations, customers, competitors (i.e. Brahm, 1994), exchange rate exposures, and a greater
diversity of cultures (i.e. Hofstede, 1980). The ability of a CEO to process such complexities
is a scare and valuable resource, and firms therefore need to compete for CEOs that possess
such abilities. Because internationalized firms are more complex, they need to find CEOs that
can process such complexity, which they can attract by giving higher compensation to their
CEOs. Furthermore, CEOs of internationalized firms are more exposed to exogeneous risks,
which can lead them to demand a risk premium that can translates to a higher level of
compensation (Oxelheim and Randøy, 2005). Therefore, I expect the following:
Hypothesis 2: Higher firm internationalization is related with higher CEO compensation.
2.3 Executive compensation and risk-taking
Palmer and Wiseman (1999) argue that there are two important types of risk in organizations:
managerial risk and organizational risk. First, managerial risk is defined by the authors as
“management’s proactive strategic choices involving the allocation of resources”. Strategic choices made by CEOs involve a certain level of uncertainty, and thus riskiness, because it can
change how an organization is structured and how it operates. Some choices are relatively
risky, while others are relatively risk-averse. Because it is difficult to determine whether
11
The rationale is that risk-averse firms choose a mode of operation that reduces cash flow risk,
which results in lower volatility of corporate earnings (Acharya, Amihud and Litov, 2011).
Vice versa, riskier firms have more volatile returns to capital (John et al., 2008). Research on
this type of risk is important, because volatile corporate earnings can negatively affect firm
value (Rountree, Weston, & Allayannis, 2008).
Although executives might work harder because of equity-based compensation as
discussed in the previous section, it does have a negative side-effect. Specifically, tying
manager’s wealth to firm performance discourages risk-taking (Chava & Purnanandam, 2010;
Coles et al., 2006; Smith & Stulz, 1985). Wiseman and Gomez-Meija (1998) consider
principals (i.e. shareholders) as risk-neutral to firm actions, because they can diversify their
shareholdings across firms. Inversely, agents (i.e. executives) are risk-averse, because their
compensation is tied to stock performance, which they cannot diversify2 and is to some extent
beyond their control (Coles et al., 2006; Low, 2009; Mehran, 1995). For CEOs to reduce their
personal compensation risk, they can choose to make corporate decisions that reduce firm risk.
Furthermore, CEOs can choose to avoid risky projects if they believe these projects can hurt
their personal career, even if such projects have a positive net present value and can be firm
value-enhancing (Hirshleifer and Thakor, 1992; John et al., 2008).
Jensen and Meckling (1976) originally identified the role of stock options in solving
the agency problems by aligning executives’ and shareholders’ interests. Options provide
CEOs with convex payoffs which could stimulate their willingness to participate in risky
projects (Kini and Williams, 2011 ; Coles et al., 2006). In other words options, preserve
incentives while mitigating risk aversion (Panousi & Papanikolaou, 2009). Risk aversion is
2 Although I don’t have the necessary data to compare the investment portfolios of executives with that of
12
further mitigated by options because option value increases with stock price volatility. Because
firm risk is associated with stock return volatility, this incentivizes managers to engage in
risk-taking (Smith & Stulz, 1985). On the empirical side, Guay (1999) and Coles et al. (2006) make
important contributions. Guay (1999) finds a positive relation between the sensitivity of CEO
wealth to stock return volatility, vega, and investments in riskier assets and more aggressive
debt policies. This suggests that vega encourages executive risk-taking.
My main expectation is that CEOs who benefit from increasing firm risk given their
personal stocks and options holdings of the firm, will implement riskier corporate policies. I
expect the reverse to happen when managers have an incentive to decrease firm risk. I use two
proxies two test this hypothesis. The first, vega, is calculated as the sensitivity of executive
wealth to stock return volatility. Vega refers to the dollar gain of an executive’s personal
portfolio when the firm’s stock return volatility increases with 0.01. Hence, vega is a relatively
direct measure of a risk-increasing incentive, because it incentivizes a manager to make more
risk-increasing decisions that result in more volatile earnings and higher stock return volatility
(Guay, 1999). This leads me to formulate the following hypothesis:
Hypothesis 3a: CEOs with higher vega implement riskier corporate policies
The second proxy to test managerial risk-taking is delta. It is the sensitivity of executive wealth
to stock price and refers to the dollar gain of an CEOs personal portfolio when the firm’s stock
price increases with 1%. As noted, linking executive compensation to stock price can cause
executives to work harder to increase shareholder wealth, but it does discourage risk-taking.
This leads me to assume that managers with higher delta are likely to favor low risk firm
13
Although firm risk is directly observable by measuring for example stock return
volatility, executive risk-taking is not directly observable by measuring one variable. In this
thesis I therefore examine four important strategic decisions that influence firm risk, and I study
how risk-increasing incentives influence these decisions. One way to increase risk is to increase
the level of leverage (Coles et al., 2006). As higher vega implies higher risk, I expect the
following:
Hypothesis 3b: Higher vega is related to higher leverage.
According to Froot, Scharfstein and Stein (1993), higher cash balance allows the firm to
smooth its investment decision better by weakening its dependence on external funding. Firms
can hold more cash to better cope with adverse shocks when access to capital is restrained and
more costly (Bates, Kahle, & Stulz, 2009). Holding cash can therefore be a precautionary
motive for firms in order to create a financial buffer. Hence, I consider higher cash holdings to
imply a risk-averse strategic decision by the CEO, and vice versa. Finally, cash balance can be
seen as negative debt (Chava & Purnanandam, 2010). This means that all the arguments about
the effects of executive incentives on the firm’s leverage also directly apply to cash holdings. Because higher vega indicates a risk-increasing incentive, I expect that it will lead to lower
cash holdings.
Hypothesis 3c: Higher vega is related to lower cash holdings.
Investments into R&D are typically seen as high-risk, as opposed to capital expenditures into
14
much less certain than, for example, a tangible machine that has a production rate that is known
before investing in it. Hence, I expect the following:
Hypothesis 3d: Higher vega is related to higher R&D expenditures.
Contrary to investments in R&D, capital expenditures (CAPEX) are low-risk. Capital
expenditures imply investments in fixed assets such as machines, in order to maintain and
increase production and productivity. They are low risk because the outcome is more
predictable than for example R&D, they are directly visible and easily measurable (Coles et
al., 2006). The amount of capital expenditures can also be relatively easily adjusted by
management without generating significant losses, which reduces the risk profile of capital
expenditures.
Hypothesis 3e: Higher vega is related to lower CAPEX
3
Sample collection and summary statistics
I begin this section by describing the sample collection process in section 3.1. Thereafter the
summary statistics are described in section 3.2
3.1
Sample collection
To investigate the pay-for-performance relationship and the influence of executive
compensation on risk-taking, I compile a sample of U.S. firms in a sample period of 12 years,
15
the Compustat, Execucomp, CRSP, and Datastream databases. I use the S&P 1500 in order to
start with a sufficiently large sample size, as I need to merge various databases and expect that
observations will be lost in this process.
I obtain CEO compensation and personal information from the Excecucomp database
using the annual CEO identifier variable. CEOs are not properly identified by Excecucomp; in
those cases, I determine the CEO by looking at the dates when a CEO assumed and left office
and classify CEOs accordingly for each firm-year combination. I collect data on the board size
from the Datastream database. I collect company level data from Compustat and stock price
information from CRSP.
Consistent with the literature, I exclude financial firms (SIC codes 6000-6999) from
the sample. I exclude these firms to avoid bias in the model, since firms in the financial sector
generally have different capital structures, corporate goals, and growth rates (Coles et al., 2006;
Guay, 1999).
I choose the 2007-2018 timeframe because large changed in accounting and report
standards occurred in the U.S. in 2006, which affects the output of compensation variables in
Execucomp. The new reporting format changes the valuation of options and shares, which are
fundamental concepts in equity-based compensation research. In 2006 16% of the firms still
report using the old standard, and in 2007 all firms report in Execucomp using the new format
(Coles, Daniel, & Naveen, 2013). Hence, starting the timeframe in 2007 means that changes in
the reporting formats do not adversely affect my model. Moreover, by choosing a sample
period that effectively starts in the 2007-2008 financial crisis, I can analyze how the
pay-for-performance relationship, the level and composition of CEO compensation, and CEO
16
3.2 Summary statistics
The final sample consists of 1,255 unique U.S. firms in the 2007-2018 period after all data is
merged and I applied all the above-mentioned criteria. Table 1 provides summary statistics for
the key variables in my thesis. The summary statistics show that the average vega in my sample
is $130,029 and the average delta is $477,100. This means that when the stock price volatility
changes with 0.01, the value of the CEOs option portfolio changes on average with $130,0293.
Moreover, when the stock price changes with 1% the value of the CEOs stock and stock options
portfolio changes on average with $477,100. Compared to Coles et al. (2006), whose timeframe
is 1992-2002, vega has increased substantially, from $80,000 in their sample to The descriptive
statistics furthermore show that the average CEO in my sample earns $1,011,000 in cash
compensation, $3,848,295 in equity-based compensation, and $5,121,534 in total
compensation. The average CEO is almost exactly 56 years old and has an average tenure of
about 7 years and 7 months.
The leverage of 23% is similar and just above other studies such as Chava and
Purnanandam (2010). Likewise, the average cash holdings of 16% are similar and just above
the 14% reported by Chava and Purnanandam (2010). Average R&D expenditures are 3% ,
which is slightly lower than the 4% reported by for example Coles et al. (2006). CAPEX is 3%,
which more than twice as low as Coles et al. (2006).
Regarding the other firm-specific variables, total average sales are $7.24 billion and
total average assets are $8.96 billion. I am using only S&P 1500 firms in the sample, the largest
1,500 firms in the U.S., which explains such high numbers. ROA is 4% and the
3 I do not calculate the vega value for stocks, as the value change of stocks resulting from stock price volatility is
17
Book ratio is 1.93, representative of other studies using the S&P1500 as the sample (i.e. Chava
and Purnanandam, 2010), and the average board size is 9.74. Finally, the correlation
coefficients, which are provided in the appendix, show the absence of multicollinearity.
Table 1
Summary statistics
Vega refers to the dollar change in CEO wealth for every 0.01 change in stock price volatility. Delta refers to the dollar change in CEO wealth for every 1% change in stock price. Cash compensation is the sum of salary and bonus. Equity-based compensation is the sum of rewards in stocks and stock options. Total compensation is the sum of cash, equity-based and other compensation. Data on policy measures is from Compustat. R&D expenditures is the ratio of R&D expenditures to the book value of total assets. CAPEX is the ratio of net funds used for additions to property, plant and equipment to the book value of total assets. Leverage is the ratio of long- plus short-term debt to the book value of total assets. Cash holdings is the ratio of cash and short-term investments to the book value of total assets. All variables are winsorized at the 1st and 99th levels, firm risk is
winsorized at 5th and 95th levels.
(1) (2) (3) (4) (5) (6)
VARIABLES N mean SD median min max
CEO characteristics Vega ($000s) 9,862 130.03 245.9 53.95 0 1,455 Delta ($000s) 9,862 477.10 479.9 102.28 0 2,928 Total compensation ($000s) 9,862 4,974 4,651 3,550 330.4 26,601 Cash compensation ($000s) 9,862 977 632.2 857 110 4,500 Equity-based compensation ($000s) 9,862 3,745 4,112 2,451 0 23,444 Stock-based compensation ($000s) 9,862 2,413 3,000 1,383 0 16,063 Option-based compensation ($000s) 9,862 1,254 1,897 603 0 11,327 Age 9,862 56.09 6.95 56 28 88
CEO corporate policy decisions
18
3.3 Evolution of compensation data over time
Figure 1 shows the evolution of CEO compensation over time of firms in the sample in the
2007-2018 time period. Table 2 displays the same information, plus information about the
evolution of the CEO vega and delta, firm stock returns, size in total assets, and return of assets
(ROA).
By glancing at Figure 1, we immediately see that the average total CEO compensation has
increased significantly since the start of the timeframe in 2007. Note the temporary decrease
in compensation in 2009, right after the 2007-2008 financial crisis.
Figure 1 also shows that compensation in the form of restricted stocks increases in
relative and in absolute popularity. In 2007 only 36% of the total compensation value consisted
of stocks, whereas in 2018 it increases to 56%. Table 2 reports that the average yearly growth
of stock options value is more than 9%, whereas cash compensation (salary plus bonus) and
other compensation only change on average in value per year by 0.8% and -2.9%, respectively.
We can see that stock options remain a popular form of compensation with an average annual
33% 34% 32% 31% 29% 29% 27% 28% 25% 24% 22% 23% 35% 36% 39% 40% 43% 45% 48% 48% 53% 54% 56% 59% 23% 23% 24% 24% 23% 21% 21% 19% 18% 17% 17% 15% 9% 7% 4% 5% 6% 5% 5% 5% 4% 4% 5% 4% $1.000 $2.000 $3.000 $4.000 $5.000 $6.000 $7.000 $8.000 $9.000 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 CEO co m p en sa ti o n in th o u sa n d s
Figure 1: Evolution of the absolute levels and relative structure of average U.S. CEO compensation between 2007-2018
19
growth of 4.2% in value, but we see in Figure 1 that it concedes some of its relative popularity
to stocks. One of the explanations could be the introduction of FAS 123R in 2005, which
increased the accounting costs of awarding stock options to executives (Hayes et al., 2012). By
comparing the 2007 with the 2018 bar in Figure 1, we clearly see that options, stocks and cash
compensation are somewhat equally distributed in 2007 (33%, 35%, 23%, respectively), but
that stocks and options together form 82% of the average annual CEO compensation in 2018.
4
Methodology
In this section I explain the research methodology of the analyses in this study.
4.1
Specification of the empirical model and measurement of variables
This study performs two main regression analyses: the pay-performance analysis and the CEO
incentive-managerial risk-taking analysis. Both models are estimated using Ordinary Least
Squares (OLS) regressions robust to heteroskedasticity.
The dataset that I use in my thesis consists of multiple years and firms, which makes it
panel data. More specifically, it is an unbalanced panel because for some firms I do not have
data on all the years in the timeframe (2007-2018). Panel data is usually analyzed using a
random effects model or a fixed effects model. Fixed effects control for omitted variables by
using group dummies. Although most of the literature uses fixed effects models (e.g. Chava &
Purnanandam, 2010; Coles et al., 2006; Jensen & Murphy, 1990; Ozkan, 2011), I decide which
model is more appropriate in the next paragraph.
One of the things to consider when deciding on a fixed or random effects model, is if
the model suffers from omitted variable bias. In the case of this thesis, managerial incentives,
20
wide range of variables, which I try to capture using control variables. Still, there might be
omitted variables not present in my models that result in biased results. Specifically, I expect
the industry in which firms operate to influence the dependent variables. Furthermore, a fixed
effects model is more suitable when the entities within a model change over time. I suspect that
this will be true in my timeframe, because my timeframe starts in 2007 during the financial
crisis. This suspicion is enforced by figure 1, which shows that the level and structure of CEO
compensation in my sample has changed a lot in the timeframe. Collectively, there is a real
change my models suffer from omitted variable bias and that the entities in my model change
over time. A fixed effects model seems therefore more appropriate than a random effects
model.
In addition to the question of a fixed effects vs a random effects model, two other
problems need to be addressed with the pay-performance and the incentives-strategic choices
analyses. First, there is a real concern these main relationships in this study suffer from dual
causality and endogeneity issues. Consistent with the literature (i.e. Chava and Purnanandam,
2010; Hartzell and Starks, 2003) I use the lagged (t-1) values of independent and control
variables to alleviate some of the dual causality concerns. For example, if I want to say that
firm characteristics lead to certain levels of executive compensation, but I only use data from
the same time period t, then it is impossible to differentiate between the executive
compensation influencing firm characteristics or firm characteristics influencing executive
compensation. Furthermore, it is logical to assume the effects of managerial incentives, i.e.
vega and delta, do not immediately influence firm outcomes. I cluster the regression models at
the panel level to adjust the standard deviations, because the Wooldridge test shows
autocorrelation (Drukker, 2003). Second, the distribution is skewed for all compensation
variables (e.g. total compensation, cash compensation, etc.), also after winsorizing them at the
21
variables show the variables are positively skewed. To address this violation of the normality
assumption in the fixed effects regression model, I compute the logarithm of the compensation
variables, consistent with the literature (i.e. Conyon, 2006; Ozkan, 2011).
The regression models are given below. In line with previous literature, this study
maintains a linear relationship between the dependent and independent variables. The Controls
are in line with previous studies and are explained later in this section. The year and industry
fixed effects are 𝜆 and 𝜇, respectively. The constant is 𝛼, and the standard error term is represented by 𝜀.
The first regression for the estimation of the level and structure CEO compensation is based
on the following model:
𝐶𝑜𝑚𝑝𝑒𝑛𝑠𝑎𝑡𝑖𝑜𝑛𝑖,𝑡 = 𝛼𝑖+ 𝛽1∗ 𝑆𝑅𝑖,𝑡−1+ 𝛽2∗ 𝑅𝑂𝐴𝑖,𝑡−1+ 𝛽3∗ 𝐶𝑜𝑛𝑡𝑟𝑜𝑙𝑠𝑖,𝑡−1+ 𝜆𝑖,𝑡+
𝜇𝑖,𝑡+ 𝜀𝑖,𝑡 (1)
Where SR refers to the annual stock returns, i.e. the market-based corporate performance, and
ROA refers to the return on assets, i.e. the accounting-based corporate performance.
The second regression for investigating the influence of internationalization on the level of
CEO compensation is based on the following model:
𝐶𝑜𝑚𝑝𝑒𝑛𝑠𝑎𝑡𝑖𝑜𝑛𝑖,𝑡 = 𝛼 + 𝛽1∗ 𝑀𝑢𝑙𝑡𝑖𝑛𝑎𝑡𝑖𝑜𝑛𝑎𝑙𝑖𝑡𝑦𝑖,𝑡−1+ 𝛽2∗ 𝑆𝑅𝑖,𝑡−1+ 𝛽3∗
𝑅𝑂𝐴𝑖,𝑡−1+ 𝛽4∗ 𝐶𝑜𝑛𝑡𝑟𝑜𝑙𝑠𝑖,𝑡−1+ 𝜆𝑖,𝑡+ 𝜇𝑖,𝑡+ 𝜀𝑖,𝑡 (2)
The third regression for the estimation of the effect of CEO incentives on managerial
risk-taking is based on the following model:
𝐶𝑜𝑟𝑝𝑜𝑟𝑎𝑡𝑒 𝑝𝑜𝑙𝑖𝑐𝑖𝑒𝑠𝑖,𝑡 = 𝛼 + 𝛽1𝑣𝑒𝑔𝑎𝑖,𝑡−1+ 𝛽2𝑑𝑒𝑙𝑡𝑎𝑖,𝑡−1+ 𝛽3𝐶𝑜𝑛𝑡𝑟𝑜𝑙𝑠𝑖,𝑡−1+ 𝜆𝑖,𝑡+
22 4.1.1 Dependent variables
Compensation level. Total compensation is calculated as the sum of cash compensation, equity-based compensation (which is the sum of stock- and options compensation), and other
compensation (not analysed in this study). Cash compensation refers to the sum of annual base
salary and bonus.
Compensation structure. The compensation structure variables are measured as the compensation variable divided by level of total compensation.
Corporate policies. The primary measure of managerial risk-taking in this study is based on four corporate policy decisions that are related to risk, namely leverage, cash holdings, R&D
expenditures, and CAPEX. First, leverage is defined as the value of book leverage, measured
as the ratio of total debt to the total assets. Using market leverage would not be appropriate in
this study, because its value can change due to stock price variations, rather than the CEOs’
decisions. Welch (2004) shows in fact that stock returns are a first order determinant of
market-based debt ratios. Second, cash holdings are measured as the ratio of cash and short-term
investments to the book value of total assets following Opler, Pinkowitz, Stulz, and Williamson
(1999) and Bates, Kahle, and Stulz (2009). Third, R&D expenditures are R&D expenditures
scaled by total assets (Coles et al., 2006), and fourth, CAPEX is defined as the annual funds
used for additions to property, plant and equipment minus the cash inflows from the sale of
property, divided by total assets. (Coles et al., 2006).
4.1.2 Independent variables
Stock returns & ROA. For regression one and two, the independent variables are market-based performance, i.e. stock returns, and accounting-based performance, i.e. ROA. Because
23
stock returns as the primary measure of firm performance, consistent with Murphy (1999). I
measure the annual stock returns by calculating the annual change in the company
end-of-fiscal-year stock prices. I use the logarithmic value of this ratio because of two main reasons.
First, logarithmic returns can be interpreted as continuously compounded returns (Hudson,
2010). This is advantageous, because I consider multi-period returns and the multi-period
continuously compounded return is simply the sum of all the single-period compounded
returns. Second, the logarithmic function of stock returns is very similar to the size of the
simple formula of stock returns (Rozeff and Kinney, 1976). I use return on assets as an
accounting-based firm performance measure. ROA is calculated as net income divided by the
total assets.
Internationalization. I calculate internationalization as the ratio of foreign sales to total sales. I do not divide the sample into two groups (i.e. as in Aabo et al., 2015), because using a ratio
instead of a binary dummy variable gives a more refined variable for analysis.
Vega & delta. For regression three, vega is the main independent variable. Vega refers to the dollar change of a CEOs personal stock and options portfolio when stock return volatility
changes with 0.01. Delta refers to the dollar change of CEOs personal stock and options
portfolio when the stock price changes with 1%. Vega is a proxy for risk-increasing incentives
whereas delta is a proxy for risk-decreasing incentive.
To construct vega and delta, I follow the methods of Core and Guay (2002), which are
also described in detail by Coles et al. (2013). The values of vega and delta are based on the
CEO’s stocks and options holdings. Particularly for vega, the process of calculating the variable involves an extensive process that requires complex calculations and merging five
different datasets (Compustat, Execucomp, CRSP, and the U.S. Federal Reserve to obtain
risk-free rates). The steps I follow to calculate vega and delta are described in Appendix B in order
24
I use vega and delta as proxies for risk-increasing and –decreasing incentives, because
alternative measures do not sufficiently capture the multidimensional characteristics of CEO
compensation and risk-taking incentives. For example, using the total number of options as an
alternative measure for incentives ignores cross-sectional variation in the characteristics of
options such as time-to-maturity, volatility, or exercise price. If I use only vested options to
measure incentives then my model ignores the incentives provided by the unvested option
(Coles et al., 2006). A number of authors have recognized vega to be the most direct measure
of risk-increasing incentives, which enforces the choice of vega as the main independent
variable in this study (i.e. Chava & Purnanandam, 2010; Coles et al., 2006; Guay, 1999).
Moreover, by using vega and delta simultaneously, I can isolate the effects of each one of these
two variables on the dependent variable, CEO corporate policy decisions, and achieve a cleaner
measurement of CEO risk incentives. Specifically, I use vega as an independent variable and
delta as a control variable.
4.1.3 Control variables
In the first and the second models regressing firm performance and internationalization on CEO
compensation, I control for various firm- and CEO-specific characteristics that have been
shown in prior research to be determinants of CEO pay (Conyon, 2006; Mehran, 1995; Ozkan,
2011; Rosen, 1990). These include firm size (natural logarithm of total sales), growth
opportunities (market-to-book ratio), board size, and CEO age. Firm size influences CEO
compensation, for example because CEO pay is determined in a competitive talent market, and
the size of that market is determined by the size of the firms to which these CEOs are related
(Gabaix & Landier, 2008; Murphy, 1999). Research also shows growth opportunities relate to
CEO compensation (e.g. Conyon & Murphy, 2000). Furthermore, larger boards can be more
25
become more vulnerable to CEO power in setting their own compensation (Ozkan, 2011). CEO
age can influence their compensation because older CEOs can become more powerful in setting
their own compensation level and structure (e.g. Hill & Phan, 1991; Ryan & Wiggins, 2001).
In the third model regressing CEO risk-taking incentives on the corporate policy
decisions, I control for the level of CEO risk aversion (delta), firm size (natural logarithm of
total sales), growth opportunities (market-to-book ratio), leverage, profitability (ROA),
because previous research shows these variables influence the corporate policies (e.g. Chava
& Purnanandam, 2010; Coles et al., 2006).
Chava and Purnanandam (2010) argue that cash holdings are de facto negative debt.
Hence, the assumptions about the effects of managerial incentives on leverage are also true for
cash holdings, but with an inverted coefficient sign. Therefore, I use the same control variables
for the two models. I control for firm size because larger firms have better access to external
capital, and larger firms benefit more from economies of scale resulting from holding a larger
cash balance (Opler et al., 1999). I control for profitability because more profitable firms have
better access to external capital, and profitability also influences the precautionary motives for
holding cash (Boutchkova, Doshi, Durnev, and Molchanov, 2012). In the R&D model, I control
for leverage because it is suboptimal for firms engaging in R&D activities to also have high
levels of leverage. R&D activity is often an intangible and firm-specific asset that cannot easily
and quickly be converted into cash or tangible assets to satisfy debtor claims in the event of
financial distress (Bhagat & Welch, 1995). Furthermore, I control for the market-to-book ratio
because growth opportunities have been shown to be associated with R&D (Bhagat & Welch,
1995; Coles et al., 2006). In the CAPEX model, I control for firm size because research shows
that firm size is related to capital expenditures (e.g. Servaes, 1994). I control for firm
profitability and growth opportunities because firms with more investment opportunities tend
26
be used to acquire property, plant and equipment, and because these are tangible assets that can
be used to satisfy debtor claims (Bhagat & Welch, 1995).
5
Firm performance and CEO compensation
5.1 CEO compensation and firm performance analysis
The regression results in table 2 show that firms with higher stock returns provide higher levels
of CEO compensation. The main independent variable, stock returns, is positive for all
compensation variables and highly significant at the 1% level in columns for total,
equity-based, and options compensation. Stock returns has a highly significant coefficient of 0.065 on
log (Total compensation), indicating that a 10% increase in stock returns translates to a 0.67%
(e0.065 – 1) increase in total CEO compensation. Stock returns are also positive for cash and
stock compensation, but at the 5% level. Collectively, the evidence in table 2 supports the claim
that CEO compensation is positively related to market-based performance, i.e. stock returns,
consistent with the literature (i.e. Conyon, 2006). These results also support the claim that firm
performance mostly drives equity-based compensation, compared to cash compensation
(Frydman and Jenter, 2010), as indicated by the higher coefficients in columns 3-5 compared
to column 2 in table 2. Furthermore, the results in columns 3 and 4 in table 3 show that firms
with higher stock returns pay less stocks and more options as a percentage of total
compensation.
The results in table 2 only partly support the claim that CEO compensation is related to
ROA, i.e. accounting-based firm performance. The results show that higher ROA is related to
lower equity-based and lower option compensation to the CEO, at the 10% and 5% significance
27
and cash compensation to the CEO. This is inconsistent with Murphy (1999), who posits that
compensation is determined using both market- and accounting-based performance metrics.
Table 2 also provides evidence of the claim that larger firms pay more CEO
compensation (i.e. Murphy, 1999). This is shown by the positive coefficients at all
compensation variables, significant at the 1% level. The results also support the claim that
firms with higher growth opportunities, i.e. a higher market-to-book ratio, pay higher CEO
compensation. This claim is not supported for cash compensation, as only that coefficient is
not significant. This could be explained by the fact that the market-to-book ratio is a proxy for
future growth, and that it is therefore rewarded with equity-based compensation which is seen
as a long-term incentive (Mehran, 1995).
Table 2 also provides empirical evidence that firms with larger boards pay their CEO
more equity-based, cash, and total compensation, but not that larger firms pay more option
compensation. An explanation for the positive relationships could be that larger boards can be
more complex, and thus less effective in their monitoring function. Consequently, the board
can become more vulnerable to CEO power (Ozkan, 2011).
Table 2 shows that older CEOs receive more cash compensation, but less compensation
in the form of equity and stocks. This can be inconsistent with the entrenchment argument that
older CEOs become more entrenched and gain more power in determining their own
compensation (Hill and Phan, 1991). One explanation however, could be that older CEOs have
accumulated enough stock and therefore seek to receive less equity-based compensation than
younger CEOs. The evidence in table 3 supports this explanation, showing that older CEOs
receive a higher percentage of their compensation in the form of cash, and less in the form of
equity and stocks. This argument is also supported by the findings of Ryan and Wiggins (2001),
who also report significantly negative coefficients between equity-based compensation and
28
performance is positively related to CEO compensation. Although Stock performance is in line
with the expectations, I find no conclusive evidence on the effect of ROA on CEO
compensation. The results only indicate that ROA is negatively related to equity-based
compensation and stock compensation, which supports the notion that equity-based
compensation is predominantly determined using market-based performance measures.
Table 2
Panel regressions of CEO compensation level on stock returns
The dependent variable is the natural logarithm of given compensation variables. Stock returns is the logarithm of yearly stock returns, and ROA is net income scaled by total assets. Firm size is the logarithm of total sales. Market-to-book is total market value scaled by book value per share. Board size and age are absolute values. t-statistics based on robust standard errors are within parentheses. ***, **, and * indicate statistical significance at the 1%, 5%, and 10%, respectively.
(1) (2) (3) (4) (5)
Independent variables Log (Total compensation) Log (Cash compensation) Log (Equity-based compensation) Log (Stock compensation) Log (Option compensation) Stock returnst-1 0.065*** 0.032* 0.085*** 0.066*** 0.106*** (0.023) (0.016) (0.028) (0.024) (0.033) ROAt-1 -0.371 -0.063 -0.516* -0.352 -0.497** (0.236) (0.109) (0.297) (0.245) (0.242) Firm sizet-1 0.385*** 0.176*** 0.463*** 0.449*** 0.401*** (0.019) (0.013) (0.020) (0.016) (0.033) Market-to-bookt-1 0.088*** -0.008 0.162*** 0.128*** 0.227*** (0.024) (0.014) (0.028) (0.020) (0.035) Board size 0.027*** 0.013* 0.023** 0.012 0.008 (0.010) (0.007) (0.010) (0.013) (0.010) Age -0.001 0.008*** -0.008** -0.006 -0.001 (0.002) (0.001) (0.003) (0.004) (0.003) Constant 4.870*** 4.890*** 4.214*** 4.042*** 3.531*** (0.157) (0.118) (0.226) (0.283) (0.224) Observations 8,574 8,574 8,574 8,574 8,574 R2 0.518 0.428 0.522 0.477 0.413 Adjusted R2 0.513 0.422 0.518 0.471 0.405
Industry FE YES YES YES YES YES
29
Table 3
Panel regressions of CEO compensation structure on stock returns
The dependent variable is the given compensation variable divided by total compensation. Stock returns is the logarithm of yearly stock returns, and ROA is net income scaled by total assets. Firm size is the logarithm of total sales, and MTB is total market value scaled by book value per share. t-statistics based on robust standard errors are within parentheses. ***, **, and * indicate statistical significance at the 1%, 5%, and 10%, respectively.
(1) (2) (3) (4)
Independent variables Cash compensation ratio Equity-based compensation ratio Stock compensation ratio Options compensation ratio Stock returnst-1 -0.004 0.013 -0.012* 0.024*** (0.008) (0.010) (0.006) (0.008) ROAt-1 0.082 -0.106 -0.011 -0.094 (0.086) (0.098) (0.053) (0.076) Firm sizet-1 -0.055*** 0.056*** 0.037*** 0.019*** (0.005) (0.006) (0.006) (0.005) Market-to-bookt-1 -0.016*** 0.023*** -0.020*** 0.042*** (0.006) (0.007) (0.006) (0.006) Board size -0.007** 0.008** 0.009** -0.000 (0.003) (0.004) (0.004) (0.002) Age 0.003*** -0.004*** -0.004*** -0.000 (0.001) (0.001) (0.001) (0.001) Constant 0.654*** 0.336*** 0.302*** 0.037 (0.052) (0.053) (0.079) (0.062) Observations 8,574 8,574 8,574 8,574 R2 0.232 0.228 0.231 0.113 Adjusted R2 0.225 0.221 0.225 0.106
Industry FE YES YES YES YES
Year FE YES YES YES YES
5.1.1 CEO compensation and firm internationalization
Because the control variables in this section are the same as the pay-performance models in
section 5.1, I will not extensively interpret the control variables in this regression model. It
should be noted though, that the controls in this section are similar to previous studies analyzing
the relationship between internationalization and executive compensation (e.g. Conyon, Core,
and Guay, 2010; Conyon and Murphy, 2000). The results in table show support for the claim
that firms with more foreign sales pay higher CEO compensation. The coefficient on the
internationalization variable of 0.891 in column 2 of table 4 implies that the predicted total
CEO compensation increases with 14.4% (e0.891 – 1) when the ratio of foreign sales to total
sales increases by 0.1. The coefficient of 0.570 in column implies a 4.8% increase in CEO cash
30
Table 4
Regressions of firm internationalization on U.S. CEO compensation level
The dependent variable in all models is the logarithm of the given compensation levels +1, to avoid unlimited negative values resulting from taking the logarithm of zero. Internationalization is total foreign sales scaled by total sales. ROA is net income scaled by total assets, stock returns is the logarithm of yearly stock returns, firm size is the logarithm of total sales. MTB is total market value scaled by book value per share. Board size and CEO age are constant values. t-statistics based on robust standard errors are within parentheses. ***, **, and * indicate statistical significance at the 1%, 5%, and 10%, respectively.
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) Independent variables Log (Total comp) Log (Total comp) Log (Cash comp) Log (Cash comp) Log (Equity-based comp) Log (Equity-based comp) Log (Stock comp) Log (Stock comp) Log (Option comp) Log (Option comp) Internationalizationt-1 2.949*** 0.891*** 1.162*** 0.389** 3.772*** 1.018*** 3.904*** 1.407*** 3.597*** 1.052** (0.251) (0.310) (0.149) (0.165) (0.379) (0.342) (0.347) (0.347) (0.576) (0.517) Stock returns t-1 0.067** 0.009 0.087** 0.062* 0.081** (0.031) (0.020) (0.034) (0.031) (0.037) ROA t-1 -0.715*** -0.167 -0.797** -0.690** -0.893** (0.223) (0.136) (0.299) (0.271) (0.350) Firm size t-1 0.364*** 0.173*** 0.451*** 0.440*** 0.407*** (0.025) (0.018) (0.019) (0.020) (0.025) MTBt-1 0.070*** -0.029* 0.147*** 0.122*** 0.215*** (0.021) (0.015) (0.023) (0.028) (0.030) Board size t-1 0.031** 0.015* 0.012 0.007 0.002 (0.012) (0.009) (0.012) (0.016) (0.012) Age t-1 -0.001 0.006*** -0.006 -0.009 0.000 (0.003) (0.002) (0.004) (0.006) (0.004) Constant 8.112*** 5.077*** 6.724*** 5.016*** 7.772*** 4.361*** 7.409*** 4.294*** 6.998*** 3.516*** (0.013) (0.183) (0.008) (0.188) (0.019) (0.296) (0.018) (0.361) (0.029) (0.222) Observations 9,862 5,306 9,862 5,306 9,862 5,306 9,862 5,306 9,862 5,306 R2 0.173 0.513 0.211 0.466 0.168 0.540 0.189 0.478 0.136 0.460 Adjusted R2 0.164 0.507 0.202 0.459 0.159 0.533 0.179 0.472 0.122 0.450
Industry E YES YES YES YES YES YES YES YES YES YES
31
6
CEO incentives and risk-taking
6.1
CEO incentives and corporate policy decisions
Higher vega is a risk-increasing incentive for managers, therefore I expect that higher vega is
associated with riskier corporate policy decisions. Delta is a proxy for risk-decreasing
incentives, which I expect to be associated with risk-averse corporate policy decisions. Hence,
I expect a CEO with a higher vega to adopt a more aggressive capital structure with higher
leverage, to implement higher R&D, lower cash holdings, and lower CAPEX. Vega is the main
independent variable, and I use delta and other control variables to isolate the effect of vega on
the dependent variable.
Table 5 reports the regression results. Column 1 shows that CEO vega has a coefficient
of 0.064 on leverage and is significant at the 5% level. This provides support for the claim that
CEOs who have higher vega implement a more aggressive, i.e. riskier, debt policy. This result
is consistent with hypothesis 3b. However, column 2 shows that after controlling for firm size,
MTB, and ROA, the coefficient of vega on leverage is no longer significant. Hence, it would
be inappropriate to make any inferences on this coefficient. Delta is significant at the 1% level
and has a coefficient of 0.025 on leverage in column 2, meaning that CEOs who have a higher
delta have a more conservative, i.e. risk-averse, debt policy. As expected, larger firms have
higher leverage, and more profitable firms have lower leverage, which could show that more
profitable firms have less need to attract external capital. Collectively, the evidence in column
1 and 2 partly supports the claim higher vega relates to lower leverage, higher managerial
risk-taking, and the evidence fully supports the claim that delta is related to lower leverage, i.e.
lower managerial risk-taking (consistent with e.g. Chava & Purnanandam, 2010; Coles et al.,
32
Columns 3 and 4 show that CEO vega has a significantly negative coefficient of -0.045
on corporate cash holdings, as expected, but the coefficient of vega on corporate cash holdings
is no longer significant after including the control variables. These results only partially support
hypothesis 3c, and only partially support the notion that higher vega relates to less
risk-decreasing managerial decisions.
Column 5 shows that the coefficient of CEO vega is not significant on R&D
expenditures, so it would be inappropriate to make any inferences on the coefficients. Column
6, which includes delta and the other controls, shows that CEO vega has a coefficient of 0.032
on R&D and is significant at the 1% level. This provides evidence of the claim that CEOs with
risk-increasing incentives have higher R&D expenditures, which is a riskier policy decision.
This result is consistent with hypothesis 3d.
Column 7 reports that vega has a coefficient of -0.019 on CAPEX and is significant at
the 1% level, lending support to the claim that higher CEO risk incentives lead to fewer
investments in risk-averse corporate policies. These results are consistent with hypothesis 3e.
Column 8 reports the results for the regression on CAPEX including control variables. Vega
has a coefficient of -0.017 on CAPEX and is significant at the 1% level. Furthermore, delta has
a coefficient of 0.003 on CAPEX and is significant at the 10% level. These results are also
fully consistent with hypothesis 3e and with the prior literature.
The possibility should be considered that CEOs with higher delta that implements
risk-averse corporate policy decisions are not motivated by their risk aversion per se. For example,
a CEO with high delta can also implement lower leverage in order to achieve shareholder value
maximization (Chava and Purnanandam, 2010). After all, delta represents the CEOs’
sensitivity between the stock price and his personal wealth. To address this concern, I need to
33
efforts by CEOs or on their risk aversion. The fact that there is a consistent pattern in the
significant coefficients of delta on leverage and CAPEX, which meet the expectations, gives
reasonable supports to the risk-aversion theory.
In order to check for robustness, I run the above analysis using simultaneous equation
models (3SLS) including the above dependent variables, vega, and delta. The robustness check
further alleviates some of the dual causality concerns regarding firm outcomes and CEO
compensation. Taken collectively, the findings in this section show that CEOs can implement
corporate policies based on the effect of those policies on their personal wealth. It should be
noted that the findings are not necessarily undesirable, because these incentives can be the
result of goal alignment between the CEO and shareholders, in order to induce corporate
policies that maximize shareholder value. It is interesting that for two of the four analyzed
corporate policies in the model, vega is such a significant explanatory variable after the
financial crisis.
6.1.1 Comparison of the results with prior literature
To determine how the influence of CEO incentives on managerial risk-taking has changed, I
will now compare the results in table 5 with the results from Coles et al., (2006), whose
timeframe is 1992-2002 with a comparable sample. It should be noted that Coles et al. include
more control variables in their regression models, something that was not desirable in this
thesis. This makes the results not directly comparable, because the additional controls influence
the magnitude of their coefficients, but it is still worthwhile to compare the results so see if
there is a pattern. In addition to the controls in this study, Coles et al. (2006) control for tenure,
cash compensation, surplus cash, sales growth and stock returns. All regression analyses of
34
variables can be compared to their results. First, in my model (table 5) the coefficient of CEO
vega on leverage with control variables in not significant, so I cannot make any inferences
based on that comparison. Second, in my model the coefficient of CEO vega on R&D is 0.034,
and in Coles et al. this coefficient is 0.080. The coefficient of CEO vega on CAPEX in my
model is -0.017, and in Coles et al. this coefficient is -0.038.
Hence, the magnitude of the coefficients of vega on the corporate policies has decreased
since 2002. Furthermore, I only show partly significant coefficient results whereas the previous
studies find significant results across all four corporate policy variables. This could show that
since the 2002, the influence of CEO incentives, i.e. vega, on managerial risk-taking has
decreased. After all, the corporate policy decisions represent managerial risk choices. The
absolute level of vega has increased since the study of Coles et al. (2006) and Chava and
Purnanandam (2010)4. While higher vega means that CEOs get paid more for increasing firm
risk, i.e. stock return volatility, this doesn’t seem to have translated to higher managerial risk-taking. Since the study of Coles et al. FAS 123R has been introduced in 2005, which imposed
new rules on compensation in the form of options, but this has not affected the riskiness of
corporate policy choices (Hayes et al., 2012). Another explanation could be that compared to
the timeframe of the previous studies, the power of CEOs has decreased to implement corporate
policies based on their personal stock and options holdings in the firm.
