Pay Duration: Conflicting Interests of Institutional Investors
and Powerful CEOs
Michiel Engberts Student number: S2186667
Master thesis MSc Accountancy & Controlling (EBM869B20) MSc Finance (EBM866B20)
Faculty of Economics and Business University of Groningen
June 2016
Words: 14,503
Supervisors: dr. R.B.H. Hooghiemstra and prof. dr. C.L.M. Hermes
Abstract:
Some severe financial crises have highlighted the importance of properly structuring compensation packages. This study empirically investigates whether institutional investors are associated with more long-term horizon incentives in a compensation package of a Chief Executive Officer (CEO). In order to measure the horizon incentives in CEO’s compensation contracts, this research uses pay duration. Based on the managerial power theory, it also addresses the impact of a powerful CEO on her or his pay duration. Although institutional investors possess strong monitoring capabilities, it is further expected that the positive association of institutional investors is constrained by the presence of a powerful CEO. After examining 617 pay contracts of CEOs, the results do not support any of the predictions. The study results provide new practical and theoretical insights about the role of institutional investors, powerful CEOs and pay duration in the largest listed United States firms.
JEL classification: J33, G32, G35
1. I
NTRODUCTIONSome severe financial crises have recently hit economies worldwide. An important reason for these financial crises are ascribed to structures of compensation packages of executives (Cheng, Hong and Scheinkman, 2015). Many components in compensation packages are linked to short-term oriented firm performance, encouraging executives to primarily focus on such firm performances (Bebchuk and Fried, 2010). By including these short-term oriented firm performance components in compensation packages, executives obtain higher amounts of remuneration if these short-term performances are delivered. Consequently, this personal benefit incentivizes top management to take excessive short-term firm risks at the expense of long-run value with possibly disastrous consequences (see: e.g., bailouts of AIG and Bear Stearns). To mitigate these potential negative effects of compensation packages, there exists an ongoing debate among scholars and politicians about how to structure executives’ compensation packages. Proper structuring of such packages are key in corporate governance debates, since Jensen and Murphy (1990) document evidence that aligning interests of Chief Executive Officers (CEOs) and shareholders are about how you pay rather than how much you pay.
The aim of this study is to investigate compensation packages of CEOs. More precisely, the influence of institutional investor shareholding and CEO power on these packages are examined in detail. Although non-executive directors determine compensation packages (e.g., Fama and Jensen, 1983), Hartzell and Starks (2003) document evidence that institutional investors’ monitoring capabilities influence this pay-setting process significantly. Since institutional investors hold their shares for long periods of time, they are characterized as more long-term oriented. Therefore, institutional investors may provide longer horizon incentives to a CEO’s compensation contract. Horizon incentives in compensation packages are key to investigate, since Cadman and Sunder (2014) and Gopalan, Milbourn, Song, and Thakor (2014) claim that adapting horizon incentives in pay contracts is an effective instrument to influence the decision making process of CEOs. These horizon incentives derived from compensation packages are also termed as pay duration. In order to clarify why pay duration is powerful, consider a large option grant compared to an immediate bonus. The former is more likely to improve a long-term orientation than the latter. If, for example, the current market price of a share is below the exercise price of the granted option, then the large option grant is worth zero. Before the option becomes valuable in the future (i.e., due to a higher stock price), the executive should perform actions to improve firm performances in the future. In contrast, the immediately obtained bonus does not require any future action. Therefore an option grant is marked as more long-term oriented than a bonus payment. Moreover, if option grants are only exercisable in the distant future instead of any point in time, long-term incentives are improved even further, when compared to the immediate bonus and the regular option grant. Although potential effects of pay duration on executive decision making are beyond the scope of this study, these effects emphasize the overall relevance to examine whether institutional investors are able to reflect explicitly longer horizon incentives in a CEO’s pay contract.
However, Bebchuk and Fried (2004) predict circumstances that a powerful executive is able to influence her own pay-setting process. As a result, incentive contracting may be less feasible in solving the misalignment of interests between shareholders and executives. For several reasons (e.g., forfeiting of unexercised grants), Morse, Nanda and Seru (2010) claim that powerful CEOs prefer more short-term horizon incentives and, in fact, shortened pay duration. Therefore, this study also investigates whether more powerful CEOs are rewarded with more short-term incentives than less powerful CEOs. In addition, these arranged compensation packages may result in misalignment of interests, for example regarding investment decisions of particular shareholders and executives (Dikolli, Kulp and Sadatole, 2009). More specifically, this shortening of compensation contract duration predicts suboptimal compensation arrangements for executives compared to the more long-term oriented interests of institutional investors. Although institutional investors possess strong monitoring capabilities, and are thus able to influence the pay process, I expect that the influence of institutional investors is constrained by the presence of a powerful CEO.
My research will make several contributions to the literature. First, recent financial crises have shown the importance of proper structuring of compensation arrangements to overcome misalignment and potential distress (Cheng et al., 2015). Especially institutional investors have incentives to prevent such negative consequences of short-termism, because they hold shares for longer periods of time. It is suggested that institutional investors can provide more long-horizon incentives in compensation contracts of executives to take advantage in the long-run (Bushee, 1998). This study empirically investigates whether institutional investors are associated with more long-term horizon incentives to a CEO’s compensation package. In particular, new insights are obtained by addressing this question in a setting where equity holdings are dominated by institutional investors1. Another key difference with previous research is that this study more explicitly addresses the length of the vesting schedules of every component in the compensation packages of CEOs. Second, to the best of my knowledge, this study is the first to address the role of a powerful CEO and his or her horizon incentives. Therefore, this research also provides new insights about a firm’s pay setting process and CEO power. Finally, from a more practical perspective, potential outcomes could spur debates about good corporate governance and potential new regulation about compensation practices.
The paper proceeds as follows. The next section describes the theoretical background briefly. In addition, current literature is reviewed about the influence of institutional investors and powerful CEOs in the pay-setting process of executives. Furthermore, the hypotheses are developed. The third section discusses variables and the corresponding data applied in the empirical investigation. Then, the
1 In contrast, Cadman and Sunder (2014) investigate the effects of institutional ownership on horizon incentives
results of the empirical analysis are summarized and discussed in the fourth section. Finally, in the fifth section the conclusion and some final remarks are presented.
2. L
ITERATURE REVIEW AND HYPOTHESIS DEVELOPMENTIn this section the literature about compensation packages in corporate governance practices is reviewed. In addition, evidence on the influence of institutional investors and powerful CEOs in designing these compensation packages is analyzed. Furthermore, hypotheses will be formulated based on the literature.
2.1. Compensation of executives
In past decades, scholars have been exploring the agency theory of Jensen and Meckling (1976). A key principle in the agency theory is that everyone maximizes their own utility. If everyone maximizes his or her own utility, interactions between principals (i.e., shareholders) and agents (i.e., executives) result in so-called agency costs. These agency costs occur due to the misalignment of interests. Agents may extract personal benefits to the detriment of the principals. Examples are empire building, excessive compensation, shirking and unnecessary perks like corporate jets. If these benefits are one-sided and, as a consequence, harm the principal, then this is classified as agency costs (Fama, 1980). To prevent this sub-optimal behavior of executives, a board of directors could perform a monitoring role to align interests between shareholders and executives (e.g., Eisenhardt, 1989). These non-executives gather information about the performance of top executives. These ex post performance evaluations by a board of directors is part of the input used to determine the rewards for executives (e.g., Fama and Jensen, 1983; Chhaochharia and Grinstein, 2009). Furthermore, rewarding of executives may also provide ex ante directions to particular future firm behavior (Eaton and Rosen, 1983). A well-known complementary tool to incentivize agents and harmonize their interests with those of the principals is to design compensation contracts in specific ways (Jensen and Murphy, 1990). Hence, compensation contracts of executives reflect, for example, a certain mix of salary, bonuses, options and restricted stock grants.
2.2. Pay duration
Prior literature investigates the effects of several components in compensation contracts (e.g., Jensen and Murphy, 1990). On the one hand executives are provided with cash and bonus components, which intuitively are more likely to reward executives for current firm performance. On the other hand, executives are rewarded in form of stock-based compensation (i.e., restricted stock grants or option grants). Stock-based compensation is commonly referred as a solution to provide more long-term investment orientation of executives (Eaton and Rosen, 1983), whilst others also emphasize potential negative effects of these types of rewards in firm decisions2.
Furthermore, to strengthen the long-term incentives in compensation packages of executives, stock-based compensation rewards are more likely to become available over a (fixed) period of time. For instance, a fixed amount of options are rewarded in a particular year, whilst (parts of) these options become exercisable only after some period. The period after which these awarded options become exercisable is called the vesting period. Previous studies document that compensation packages consisting of grants with long vesting schemes are able to reduce myopic interests of executives in a firm’s investment decisions (e.g., Brisley, 2006; Laux, 2010; Laux, 2012; Edmans, Fang and Lewellen, 2015).
Despite a vast amount of literature about composition of compensation contracts, including vesting schemes, little is known about whether compensation packages of executives put more emphasis on short-term incentives or on long-term incentives. In order to be able to determine the horizon of incentives in pay contracts, Gopalan et al. (2014) have recently developed a new measure termed pay duration. Pay duration is claimed to be a more important method for influencing executive behavior than previous instruments discussed in the literature. A similar statement about pay duration is documented by Cadman and Sunder (2014). In particular, it is demonstrated that varying the underlying components and vesting schemes of pay duration, the decision horizon of managers can be strongly influenced. Furthermore, it is stated that horizon investors explicitly provided their executives with short-horizon incentives. Studies of Gopalan et al. (2014) and Cadman and Sunder (2014) stress the need to design a proper compensation package in order to align interests over longer periods of time. Thus, addressing the duration in compensation contracts is a new and important approach for boards to align the interests between shareholders and executives.
2.3. Institutional investors
Simultaneously with their increased ownership in United States’ firms, institutions or individuals which own a substantial equity stake in a firm, have evolved to a more active role in corporate governance practices in the past decades (Gillan and Starks, 2000)3. To demonstrate, in the past institutional investors sell their shares, if they did not agree with business policies. Recently, institutional investors have taken their responsibility and address corporate policies before proceeding to ‘exit’ (i.e., sell) their stakes. In particular, if a corporate policy is a sensitive issue in society, institutional investors are expected to express their ‘voice’ before ‘exit’ (Hirschman, 1970; 1980). More specifically, institutional investors are likely to exert pressure through direct negotiations with a firm’s management team, by submitting proxy proposals and/or using media influence (Gillan and Starks, 2000; Westphal and Bednar, 2008; Aggerwal, Saffi, Sturgess, 2015). Because an ‘exit’ drives the share
3 The Security and Exchange Commission (SEC) classifies any entity or natural person as institutional investor if
he or she invests, or obtains discretion, over more than $100 million of securities. Some examples of institutional investors are investment advisors, banks, insurance companies, pension funds and mutual funds. See: https://www.sec.gov/about/forms/form13f.pdf (22-02-2016; 15:35).
price down (Holthausen, Leftwich and Mayers, 1990; Brown and Brooke, 1993; Chan and Lakonishok, 1993; Sias, Starks and Titman, 2002; Dey and Radhakrishna, 2015), firms executives are likely to meet the demands of institutional investors4. Hence, institutional investors have become an important group to monitor corporate policies.
There are three reasons why institutional investors have developed incentives to monitor firms. First, institutional investors may want to bolster their reputation (Wang and Mao, 2015). Second, institutional investors are motivated to safeguard the value of their equity stakes (David, Kochbar and Levitas, 1998). Because simply selling shares results in price pressures, institutional investors also suffer financial losses from an ‘exit’. Third, Jensen (1993) notes that alternative monitor mechanisms as capital markets have become less effective in addressing the misalignment between shareholders and management, and such misalignment comes with costs5. However, there are also drawbacks for institutional investors in monitoring corporate actions. For instance, institutional investors incur large costs (Grossman and Hart, 1980). In addition, monitoring is prone to free-rider problems by other shareholders (Shleifer and Vishny, 1986). Nevertheless, Gillan and Starks (2000) document that the trade-off between active (‘voice’) and passive (‘exit’) monitoring is profitable for (large) institutional investors (i.e., due to more efficient distribution of time and effort)6.
Given that institutional investors are key to influence corporate policies, numerous studies have attempted to specify the interests of institutional investors. Particularly, Bushee (1998) claims that institutional investors prefer more long-term oriented behavior in corporate policies of executives, since they holds shares for longer periods of time7. Furthermore, Wahal and McConell (2000) document that firms with high institutional ownership shift their investments towards more long-term profitable projects. In other words, an important interest of institutional investors is to focus on maximizing long-term firm value. As highlighted before, proper structuring of compensation contracts is key to mitigate short-term behavior of executives (Edmans et al., 2015). Thus, among a variety of corporate policies to influence, the impact of institutional investors is expected to be observable in a firm’s pay-setting process. This view is supported by findings of Hartzell and Starks (2003), who document that
4 Please note that this suggests that institutional investors are also able to indirectly influence the pay-setting
process by their trading behavior. The indirect effect of institutional investors is however beyond the scope of this study. For a clear description of this indirect effect, see Hartzel and Starks (2003, p. 2370).
5 For example, due to downward trends in takeovers, capital markets are ineffective in replacing bad performing
managers.
6 Dasgupta and Placentino (2015) argue that this trade-off also depends on the amount of individual’s own
investments of fund managers in a particular firm. These money fund managers are likely to be classified as institutional investors in this analysis. A discrimination is however beyond the scope of this study.
7 Bushee (1998) distinguishes two different types of institutional investors due to different trading behavior. In
brief, on the one hand there exist transient institutional investors who are more short-term oriented, whereas on the other hand dedicated institutional investors are classified as more long-term oriented.
institutional investors possess capabilities to review compensation contracts. More precisely, institutional investors influence compensation contracts in several ways and can be described as follows.
Hartzell and Starks (2003) document that due to institutional monitoring, a decrease in the total wealth is observable. Furthermore, the composition of compensation components shifts from cash components to more long-term oriented equity-based compensation. In addition, according to a study of Cadman, Rusticus, and Sunder (2013), institutional investors also provide longer vesting schemes to their executives of the firms in which they invest. Taken together, as theoretically predicted by Bushee (1998), institutional investors can indeed implicitly provide longer horizon incentives in compensation packages. Consistent with this expectation, Cadman and Sunder (2014) find empirical evidence that institutional investors restrict a large, dominant, short-term oriented shareholder in its ability to reduce the horizon incentives in compensation contracts of CEOs. Therefore, in a setting where equity holdings are dominated by (large) institutional investors, it is hypothesized that due to these institutional investors, compensation contracts of CEOs consist of more long-term incentives. Or, alternatively formulated, pay duration is lengthened by higher institutional ownership:
Hypothesis 1: Institutional ownership is positively associated with a longer CEOs pay duration. 2.4. Managerial power theory
Bebchuk and Fried (2004) cast doubt on the optimal contracting assumption of classical agency theory, that is, that boards provide executives with incentives in compensation contracts to align interests between shareholders and executives effectively. Although boards should structure compensation packages in the best interests of shareholders, a number of studies document that powerful CEOs may be able to obtain significant control over proposed firm policies of boards, including their compensation practices (e.g., Adams, Almeida, and Ferreira, 2005; Garvey and Milbourn, 2006; Morse et al., 2011). The managerial power theory (MPT) suggests that incentive contracting is a manifestation of the agency problem rather than a solution (Van Essen, Otten and Carberry, 2015). So, if an executive is able to influence her own pay-setting process, incentive contracting is less appropriate in solving the misalignment of interests between shareholders and executives.
According to Bebchuk and Fried (2004), there are three reasons why directors act in the interests of executives rather than the interests of shareholders. First, CEOs possess power to influence director selections. Since fulfilling a director position is considered a desirable function, directors have incentives to retain it and ensure reelection by adhering to the CEOs desires. Second, due to social and psychological interactions (e.g., friendships and reciprocity effects) between CEOs and directors, refuting arguments of CEOs may be more difficult (O’Reilly and Main, 2010). Finally, CEOs are able to adapt remuneration of directors. Given these three reasons, powerful CEOs who control board members will be able to influence compensation structures.
It is further important how CEOs would change their compensation packages if these CEOs gain control over their board members and are thus likely to influence their compensation packages in their own interests. According to the entrenchment skimming theory, higher compensation of powerful CEOs is attributed to the fact that these powerful CEOs exert their power over boards in order to capture rents from the companies (Morse et al., 2011). In fact, prior literature documents two ways how this rent skimming may be expressed8.
First, Baker, Jensen, and Murphy (1988) claim that executives in general prefer compensation arrangements consisting of higher cash payments, which are less dependent on performance. Executives have incentives to prefer this, since executives have liquidity needs or are risk-averse (Walker, 2010). Indeed, Morse et al. (2011) state that powerful executives obtain higher amounts of cash compensation. Thus, powerful CEOs are more likely to obtain higher payments, and more cash payments in comparison to CEOs with less power.
Second, for several reasons, executives may prefer short-term over long-term vesting schemes. For example, due to the threat of potential CEO turnover, grants of CEOs forfeiture that have not vested yet result in lost income (Laux, 2012). In addition, due to shorter vesting terms, risk-averse executives are able to trade their shares or options more rapidly. Furthermore, if an executive possesses more private information about the company, then shortened vesting schedules are beneficial because the executive is able to trade grants opportunistically (Cadman et al., 2013). As a consequence, powerful CEOs negotiate for shortening or even eliminating their entire vesting periods. Cadman et al. (2013) indeed provide evidence that powerful CEOs shorten vesting schemes in their compensation contracts. At the same time, however, powerful CEOs may fear the threat of public scrutiny when all long-term incentives in their compensation contracts are removed. Nevertheless, powerful CEOs are more likely to obtain shortened vesting schedules in contrast to CEOs with less power.
To conclude, it seems reasonable to expect that powerful CEOs prefer to emphasize higher riskless payments over risky compensation payments. However, if risky long-term components are included, then it is likely that these forms of compensation have shorter vesting periods or that the corresponding dollar value of these types of incentives are relatively low in comparison to their immediate payments9. These predictions support the entrenchment skimming theory. Summarizing, all
8 Recent studies also suggests a third reason. To illustrate, Morse et al. (2011) state that powerful CEOs dictate
firm’s to switch from difficult long-term targets to more easier long-term targets by arranging compensation packages. Furthermore, Abernethy, Kuang, and Qin (2015) find that powerful CEOs can mitigate incentives in contracts which in potential may result in lower benefits. Thus, these disguised cash payments are ‘camouflaged’ to indicate long-term incentives for outsiders. A review of particular details in contracts are beyond the scope of this study, since it does not directly affect pay duration.
9 Intuitively, this may also be an expression of symbolic compensation (Westphal and Zajac, 1994). For example,
to mitigate the likelihood of public scrutiny, powerful CEOs may arrange high dollar values to grants with shortened vesting scheme (perhaps with easy targets). By doing this, CEOs mask disguised cash payments. Alternatively, a powerful CEO may deliberately negotiate a long vesting scheme to hide short-term incentives, whilst the corresponding value regarding this lengthened vesting scheme is negligible. Both symbolic effects are
these possibilities of components and terms are captured in pay duration, which as a result is likely to be decreased. Thus, it is reasonable to expect that long-term horizon incentives in compensation packages of powerful CEOs are reduced. More specifically, the following is hypothesized:
Hypothesis 2: More powerful CEOs are negatively associated with pay duration. 2.5. Institutional investors and powerful CEOs
Studies of Hartzell and Starks (2003) and Cadman et al. (2013) document that both institutional investors and powerful CEOs are able to influence the pay-setting process. To illustrate, institutional investors want to lengthen horizon incentives in compensation contracts, whereas a powerful CEO may constrain the ability of institutional investors suppressing their interests. Thus, institutional investors and powerful CEOs have conflicting interests regarding pay arrangements.
According to Westphal and Bednar (2008) powerful CEOs possess tactics, which are useful in other circumstances such as negotiations with institutional investors10. More specifically, Westphal and Bednar (2008, p.1) claim the following: “CEOs’ ingratiation and persuasion tactics towards institutional fund managers reduce the effect of institutional ownership on specific changes in … CEO compensation”. Consistent with this view, Janakiraman, Radhakrishnan and Tsang (2010) find that institutional investors are more influential in the pay process when a CEO is characterized as less powerful11. The effectiveness of institutional investors’ pay interventions are thus reduced as CEO power increases. Therefore, it is reasonable to expect that institutional investors provide more long-term horizon incentives in a CEO’s compensation packages, whilst this relation is mitigated by the presence of a powerful CEO. More specifically, the following is hypothesized:
Hypothesis 3: The positive association with institutional ownership on pay duration is weakened by the presence of a more powerful CEO.
3. M
ETHODOLOGY AND DATA3.1. Sample selection
The sample of firms for this research consists of the Standard & Poor’s 500 (S&P 500) firms. The S&P 500 comprises the 500 largest listed firms on the New York Stock Exchange (NYSE), American Stock Exchange (AMEX) and NASDAQ in the United States (U.S.). Because of revised regulation in accounting treatment and compensation disclosures after 2005, the empirical analysis is
captured in pay duration. However, to the best of my knowledge, there is no literature available who address or validate these symbolic effects simultaneously. Although scant literature about pay duration in this context, it is quite logical, because pay duration has recently been introduced to the literature.
10 See Westphal and Bednar (2008) for a precise description of tactics.
11 Janakiraman et al. (2010) investigate CEO power by using managerial ownership as proxy, whilst this study
conducted over a 2002-2005 period12,13. Data are collected from several sources. First, consistent with prior research on CEO compensation (e.g., Gopalan et al., 2014; Van Essen et al., 2015), Compustat ExecuComp (hereafter: ExecuComp) provides details of remuneration components in CEO compensation contracts. Second, data about institutional investor shareholdings are obtained from the Thomson Reuters Institutional (13F) Holdings database. Third, Investors Responsibility Research Center (IRRC) database provides details of individual board directors. Fourth, from Compustat the annual firm financials are collected. Both the Thomson Reuters Institutional 13F Holdings, IRRC and Compustat databases are commonly used in prior research (e.g., Cheng, 2008; Cheng et al., 2015). Finally, in accordance with prior equity-based compensation literature (e.g., Cadman et al., 2013) vesting details of the compensation contracts of CEOs are hand-collected from proxy statements in the EDGAR SEC database.
The sample starts with a full set of CEO observations for four years. Then, all firm-years with missing compensation observations (e.g., missing Black-Scholes option values) are removed. Furthermore, firms with missing vesting terms and different compensation policies are deleted from the sample14. Finally, 617 fully available CEO-firm-year observations are included in the final sample. In addition, industries are defined using the Fama and French 12 industry classification15. The sample characteristics are depicted in table 1 in three different panels. First, the selection process is summarized in Panel A. Second, an overview of CEO firm-year observations between industries are described in Panel B. Third, the number of firms (in)completely covered in the sample are presented in Panel C.
3.2. Dependent variable: Pay duration
The dependent variable used in the regression model is pay duration. Whereas prior literature (e.g., Cadman et al., 2013) frequently analyzes particular items such as cash-based components or vesting details of compensation packages, in this study the total compensation package is considered. Thus, both compensation components and vesting details of equity-based compensation are taken into account.
12 For details about overhauled disclosure regulation and accounting treatment, see respectively the following:
http://www.sec.gov/rules/final/2006/33-8765.pdf and http://www.fasb.org/cs/BlobServer?blobkey=id&blobnoca che=true&blobwhere=1175823287357&blobheader=application%2Fpdf&blobcol=urldata&blobtable=MungoBl obs (both available at 11 January 2016, 15:15).
13 One may cast doubt on the relevance of this sample period. However, recent studies as Cheng et al. (2015) show
the importance by investigating similar time periods. In addition, due to revised regulation by SEC and Financial Accounting Standard Board (FASB) about compensation disclosure and accounting treatment of options (i.e., SFAS 123(R)), respectively, in U.S. firm starting in 2006, many variables became uncollectible in ExecuComp for more recent years (e.g., restricted stock values and Black-Scholes option values).
14 For example, some firms report different incentive plans with additional restrictions dependent on firm or stock
performances.
15 Although the Fama and French 12 industry classification is less common in the literature than the Fama and
French 48 industry classification, it is an appropriate method to allocate sufficient firms to particular industries given the small research sample (Eckbo and Kisser, 2015). See: http://mba.tuck.dartmouth.edu/pages/faculty/k en.french/Data_Library/det_12_ind_port.html (consulted, 5 April 2016, 09:35)
16 For example, data of General Electric are fully available in all sample years 2002-2005. In contrast, data of
NVIDIA Corporation are only available during 2002-2004. Potential explanations for differences are a changing composition of the S&P500 and previous restrictions in data presented in Panel A of table 1.
Table 1: Sample characteristics
Panel A: Sample selection
Sample size Firm-years available in ExecuComp for S&P 500 firms in 2002-2005
period
1,642
Less restrictions:
Firm-years with missing Black Scholes ($) values 3
Firm-years with missing Compustat Fundamental data 803
Firm-years with uncollectible vesting terms 140
Firm-years with missing institutional investor shareholding 32
Firm-years with missing board sizes 47
Final sample 617
Panel B: Distribution of firm-years between industries
2002 2003 2004 2005 Total
Business equipment 41 46 44 39 170
Chemicals and allied products 10 9 8 9 36
Consumer durables 9 10 9 10 38 Consumer non-durables 11 12 13 13 49 Energy 6 6 6 5 23 Finance 2 4 5 7 18 Healthcare 20 19 21 21 81 Manufacturing 22 23 23 22 90 Other 3 3 3 3 12
Telephone and television transmission 0 0 1 2 3
Utilities 0 0 0 0 0
Wholesale, retail and some services 25 27 31 14 97
Total 149 159 164 145 617
Panel C: Covered sample firms16 firms (#) fully available in
sample period
firms (#) incompletely available in sample period
Business equipment 32 17
Chemicals and allied products 7 3
Consumer durables 6 2 Consumer non-durables 11 3 Energy 4 3 Finance 2 6 Healthcare 16 8 Manufacturing 17 9 Other 2 2
Telephone and television transmission 0 2
Utilities 0 0
Wholesale, retail and some services 11 21
Pay duration is a method to explicitly summarize the short and long-term horizon incentives in compensation packages (Cadman and Sunder, 2014; Gopalan et al., 2014). In fact, it is an instrument to express whether a compensation package is more or less long-term focused. By calculating pay duration, the length of the vesting schedules of every component in the compensation packages is explicitly considered. According to Gopalan et al. (2014, p.10) pay duration can be estimated as follows17:
𝑃𝑎𝑦 𝑑𝑢𝑟𝑎𝑡𝑖𝑜𝑛𝑖= (𝑆𝑎𝑙𝑎𝑟𝑦+𝐵𝑜𝑛𝑢𝑠)×0+∑𝑁𝑠𝑖=1(𝑅𝑒𝑠𝑡𝑟𝑖𝑐𝑡𝑒𝑑 𝑠𝑡𝑜𝑐𝑘𝑖×𝑡𝑖)+ ∑ (𝑂𝑝𝑡𝑖𝑜𝑛𝑗×𝑡𝑗) 𝑁𝑜 𝑗=1 𝑆𝑎𝑙𝑎𝑟𝑦+𝐵𝑜𝑛𝑢𝑠+ ∑𝑁𝑠𝑖=1(𝑅𝑒𝑠𝑡𝑟𝑖𝑐𝑡𝑒𝑑 𝑠𝑡𝑜𝑐𝑘𝑖)+ ∑ (𝑂𝑝𝑡𝑖𝑜𝑛𝑗) 𝑁𝑜 𝑗=1 (1)
where subscript i signifies a restricted stock grant and subscript j signifies an option plan. Restricted
stocki is expressed in dollar value of restricted stock grants with a vesting period ti. Optionj is option grant j expressed in the Black-Scholes dollar value with associated vesting period tj. Since firms may have issued other stock grants or options with different vesting periods (ti) during the year, Ns or No sum up the total number of stock grants or options, respectively. Consistent with Cadman and Sunder (2014) and Gopalan et al. (2014), Salary and Bonus are expressed in dollars and determined annually, implying that vesting periods of zero are assumed18.
In order to clarify the duration calculations with respect to restricted stock grants and option grants, the following formula is stated by Cadman and Sunder (2014, p. 1308):
𝑡𝑖𝑡 =
1
𝑅𝑒𝑠𝑡𝑟𝑖𝑐𝑡𝑒𝑑 𝑠𝑡𝑜𝑐𝑘 𝑖𝑡 (#)∑ 𝑡 × 𝑉𝑒𝑠𝑡𝑖𝑡
𝐾
𝑡=0 (2)
where Restricted Stock (#) (Options (#)) denotes the number of shares of restricted stock (number of options) granted during the fiscal year for executive i. K denotes the vesting period over which the entire grant of restricted stocks (options) is vested for executive i. Vestit denotes the number of restricted stocks
(options) that is vested in year t for executive i19. Regarding the collection of vesting data of equity-based compensation components, a precise description including assumptions and data corrections is included in Appendix B.
17 Please note that pay duration excludes other forms of compensation, such as life insurance premiums and
post-retirement benefits (401(k)). In particular, post-retirement grants may be characterized by additional restricted stocks or option grants (Gopalan et al., 2014). However, corresponding vesting data are unavailable in proxy statements. Thus, pay duration does not capture all forms of compensation that serves a long-term purpose.
18 According to Cadman and Sunder (2014) assigning a zero duration to salaries and bonuses implicitly assumes
that these components are granted for short-term firm performance of the executive. However, this assumption may not hold. For instance, if an executive attain a (undisclosed) long-term firm purpose, a bonus may be paid. In fact, this bonus served a long-term firm goal, implying the calculated pay duration is biased downwards. Because firms do not disclose details about such cash components, the problem is not addressed.
19 For example, in 2004 Abbott Laboratories (ABT) granted their CEO Miles D. White 51,475 restricted stocks,
which cliff vested at t=5. This means a restricted stock duration of (1/51,475)*(5*51,475) = 5. Further, ABT granted their CEO 404,340 options, vesting immediately after their first anniversary in three equal installments. Thus, (1/404,340)*(1*134,780+2*134,780+3*134,780) = 2. Having identified the restricted stock and option duration respectively (tit), the pay duration can be computed as follows: ((1551,846+2700)*0+2401,3*5+
3.3. Independent variables 3.3.1. Institutional investors
The variable institutional investor shareholding is operationalized as the ratio of shares owned by institutional investors to the total outstanding shares, both measured in the quarter prior fiscal year ends20. Consistent with the SEC definition, institutional investors are any natural person or entity, who/which has obtained discretion over at least $100 million of securities. This measure is similar to previous studies from Cheng et al. (2015) and Cadman et al. (2013). In accordance with Hartzell and Starks (2003) a lagged value of institutional shareholding is included for two reasons. First, it is assumed that institutional investors monitor the firm accurately during the year. Because institutional investors have incentives to hold their shares for longer periods of time, it is likely that both the shareholding as well as the number of institutional investors does not significantly change in relative short periods of time, i.e., during the year (Bushee, 1998). Second, equity-based compensation components, among other things, are reviewed by institutional investors in detail (Hartzell and Starks, 2003), since these components affect future firm decisions (Brisley, 2006; Edmans et al., 2015). In addition, Edmans et al. (2015) claim that equity-based compensation components are commonly determined at the end of the (fiscal) year. A potential intervention in the pay-setting process is thus expected to be observed around fiscal year ending. This intervention is expected to be performed by institutional investors who own shares from the beginning of that particular fiscal year. Therefore, the lagged shareholding of institutional ownership is used in this analysis.
3.3.2. CEO power
To capture CEO power, four proxies are included. A first measure of CEO power is CEO duality. A CEO who is serving a dual role in the governance of firms, that is, acting as a CEO and chairman of the board possesses power to influence the pay-setting process (Yermack, 1996; Van Essen et al., 2015; Shin, 2016). For example, CEO duality may be overwhelming to other board member because the CEO performs the two highest ranks in a firm (Ungson and Steers, 1984)21. Following Hou, Priem and Goranova (2014), CEO duality is obtained from the annual title of the CEO provided by ExecuComp. More generally, if a CEO also serves as chairman, this is marked as 1, and 0 otherwise.
A second proxy to measure CEO power is CEO tenure. Finkelstein (1992) claims that it takes time to build up expertise and reputation in boards. Consequently, after several years the power of CEOs in boards increases. This is supported by Van Essen et al. (2015), who find evidence that longer tenured
20 The calendar quarterly reported institutional investor holdings are matched precisely with the fiscal year end
month obtained from Compustat in order to ensure consistency in reference dates. While references dates matched in most cases, it is important to note that fiscal year ending do not necessarily have to be similar to calendar quarters. For example, Campell Soup’s fiscal year ended on 31th of July, while the institutional investor holdings were reported on either 30th of June or 31th of September. In these cases, the nearest upcoming quarter ending is assigned, that is, 31th of September.
CEOs influence compensation policies of firms. In accordance with Hou et al. (2014), CEO tenure is determined by counting the number of years that have passed since he or she became CEO. Following this approach, negative tenures are presented by twenty-six firms due to input errors of ExecuComp. In order to be conservative these negative CEO tenures were set to 3 years, which is consistent with Morse et al. (2011).
A third proxy to capture CEO power is board size22. Larger boards are less effective in preventing CEO power due to problems as free riding of directors. In addition, it is more complicated to generate firm decision agreements when board size increases (Zahra and Pearce, 1989; Jensen, 1993; Yermack, 1996). As a consequence, it is more likely that a CEO is able to dominate the board when board size increases. Board size is measured by counting the number of directors who served the board (Cheng, 2008).23
A fourth proxy to measure CEO power is through board independence. According to Cheng (2008), a director is characterized as independent if three conditions are met. First, a director has not been employed previously as employee or executive at the same company. Second, an independent director has not been connected as family to a current executive of the company. Finally, an independent director has not been involved in any business transactions either personally or through associated companies with the firm. Board independence is calculated by the number of independent directors divided by the total amount of directors.
In order to use these four proxies (i.e., CEO duality, CEO tenure, board independence and board size) simultaneously, an individual CEO power measure is constructed24. As documented before, the CEO power proxies possess an underlying structure to describe CEO power. This is an important requirement for factor analysis (Hair and Black, 2009). After performing an exploratory factor analysis, the four CEO power proxies are loaded into one factor. Then, a principal component analysis is used to create a factor score (CEO power) which summarizes the four CEO power proxies. This procedure is consistent with Abernethy et al. (2015). When a CEO is more powerful, the value of the constructed variable improves.
22 The reference data provided by IRRC are on annual meeting date, whilst Compustat and ExecuComp provide
their reference data on fiscal year end. Because both datasets do not match precisely, it is assumed that annual meeting take place after the fiscal year end. Eyeballing the data it seems reasonable that the annual meeting date are approximately fifth month after fiscal year ending. Therefore, ExecuComp fiscal year end dates are corrected for fifth months and then merged. All doubtful cases, that it, dates around November, December and January are verified by hand in order to ensure proper matching of years.
23 Cheng (2008) claims that in some cases IRRC database is incorrect in determining board size. To identify
potential errors, all board sizes below four directors must be corrected. After verifying my data, I safely conclude that there is no indication that board size is incorrect in my sample.
3.4. Control variables
Control variables affecting CEO’s compensation packages are included to improve the accuracy of the analysis. According to Chen (2004) it is likely that compensation packages are changed when a CEO retires in the near-term. Gopalan et al. (2014) provide evidence that horizon incentives are shorter, when executives are older. The idea is that an older CEO has built a reputation, which is likely to be at stake if short-term value is preferred at the cost of long-term results. Consequently, there exists a self-control for older CEOs and, thus, less need for longer pay durations. To self-control for this effect, a proxy for retirement is considered. More precisely, the current age of the CEO according to the proxy statement is included and adjusted following the procedure described in Appendix A25.
Furthermore, prior literature has highlighted the influence of five firm characteristics in compensation contract design. First, Carter, Ittner and Zechman (2009) claim that firm size captures many potential excluded variables, such as organizational complexity. Consistent with previous studies about pay duration, the natural log of total assets is included. Second, pay duration is positively related to Research and Development (R&D) intensity, because potential benefits of R&D expenses are observable after several years (Cadman and Sunder, 2014; Gopalan et al., 2014). To control for this effect, R&D expenses scaled by total assets is included. Third, Cadman et al. (2013) and Gopalan et al. (2014) state that firms with higher growth potential emphasize more long-term incentives in compensation packages to executives. Therefore, market-to-book value is included to control for this. Fourth, Gopalan et al. (2014) observe that more profitable firms provide longer pay duration to their executives. To address this effect, earnings before interest and tax (EBIT) to sales is included. Finally, debt holders monitor firm’s compensation contract design (Jensen and Meckling, 1976; Jensen, 1986). To control for this effect, financial leverage is taken into account, that is, debt to total assets is added (Gopalan et al., 2014)26.
3.5. Regression equation
This research estimates the coefficients of interests by performing a panel data regression analysis. The empirical model is as follows:
𝑃𝑎𝑦 𝑑𝑢𝑟𝑎𝑡𝑖𝑜𝑛 = 𝛼 + 𝛽1∗ (𝐼𝑛𝑠𝑡𝑖𝑡𝑢𝑡𝑖𝑜𝑛𝑎𝑙 𝑖𝑛𝑣𝑒𝑠𝑡𝑜𝑟 𝑠ℎ𝑎𝑟𝑒ℎ𝑜𝑙𝑑𝑖𝑛𝑔)𝑖𝑡−1
+ 𝛽2∗ (𝐶𝐸𝑂 𝑝𝑜𝑤𝑒𝑟)𝑖𝑡
+ 𝛽3∗ (𝐼𝑛𝑠𝑡𝑖𝑡𝑢𝑡𝑖𝑜𝑛𝑎𝑙 𝑖𝑛𝑣𝑒𝑠𝑡𝑜𝑟 𝑠ℎ𝑎𝑟𝑒ℎ𝑜𝑙𝑑𝑖𝑛𝑔𝑖𝑡−1 × 𝐶𝐸𝑂 𝑝𝑜𝑤𝑒𝑟𝑖𝑡)𝑖𝑡
+ ∑6𝑗=4𝛾𝑗∗ (𝐶𝑜𝑛𝑡𝑟𝑜𝑙 𝑣𝑎𝑟𝑖𝑎𝑏𝑙𝑒𝑗)𝑖𝑡+ 𝜀𝑖𝑡 (3)
where i denotes a particular firm measured at time t. Furthermore, j denotes a control variable.
25 After applying this procedure, CEO is less left skewed.
26 According to the SEC definition of institutional investors, convertible debt must also be reported in 13F fillings.
Since institutional investor shareholding is explicitly observed in the empirical analysis, all convertible debt are excluded in order to prevent potentially double counting.
4. R
ESULTS4.1
. Descriptive statisticsTable 2 provides descriptive statistics for all variables. Due to the presence of some large outliers, the variables are winsorized by following the procedure of Keller (2011)27. Moreover, Appendix A reports corresponding definitions of all variables in this analysis.
From table 2, it is apparent that the average pay duration of CEOs is 1.420 years with a corresponding standard deviation (st. dev.) of 0.941 years. The scores for this variable are similar to the pay duration statistics reported by Gopalan et al. (2014), despite a different time period28. The documented pay duration of 1.420 years appears short in comparison with the reported mean of CEO tenure (5.5 years). However, there are two reasons for this observation. First, components as salary and bonuses are considered to be paid immediately29, for example, to provide for the liquidity needs of CEOs (Walker, 2010). This assumption results in a downward bias of pay duration. Second, equity-based compensation components as restricted stock and option grants mainly vest within five years (Zhang et al., 2008).
Further analysis of the independent variables shows that the mean institutional investors shareholding is 0.683 (st. dev. 0.148). More precisely, institutional investors own on average 68.3% of firm’s common shares in the sample. This average is fairly consistent with the finding of Cadman et al. (2013)30. The mean observation of this sample supports the statement of Parrino, Sias and Starks (2002), that is, the largest listed firms in the United States are pre-dominantly owned by institutional owners31. Next, table 2 also reports the separate statistics of the CEO power proxies. Boards are on average dominated by independent directors, with a mean of 0.742 (st. dev. 0.141). Firm’s boards are thus comprised of 74.2% independent directors. These boards further consist on average of 10.2 board members (st. dev. 2.3). In addition, most of the CEOs serve a dual role in firm (0.669). Specifically, 66.9% of the CEOs is also chairman of the board of directors. Finally, an appointed CEO performs, on average, his or her tasks for 5.5 years. In summary, these four CEO power proxy statistics are roughly similar to prior literature who have used these variables (e.g., Almeida et al., 2005; Cheng, 2008; Hou et al. 2014)32.
27 Variables are bounded by the original mean plus and minus three times the original standard deviation of that
particular variable.
28 Gopalan et al. (2014, p. 2793) report on average a CEO pay duration of 1.44 and a standard deviation of 1.045. 29 See also footnote 18.
30 Cadman et al. (2013) document an average an institutional investor shareholding of 61.6% and a median of
68.9%. Unfortunately, a standard deviation is unreported.
31 An interesting graph about the increasing presence of institutional ownership in United States’ stock markets is
depicted in Gillian and Starks (2007, p. 57).
32 Please note that factor scores (here: CEO power) cannot be compared between studies because the underlying
Finally, summary statistics for control variables are comparable to the firm statistics provided in Gopalan et al. (2014)33.
Table 2: Descriptive statistics
Variable Number of
observation s
Mean Standard
deviation
Minimum Median Maximum
Pay duration 695 1.420 0.941 0 1.390 4.385 Institutional investor shareholding 664 0.683 0.148 0.232 0.698 1 CEO power 624 0 1 -2.264 -0.100 3.004 Board independence 624 0.742 0.141 0.311 0.769 1 Board size 625 10.226 2.307 5 10 17.161 CEO age 695 55.148 7.032 39 55 77 CEO duality 695 0.669 0.471 0 1 1 CEO tenure 695 5.531 4.742 0 4 21.636
Debt to total assets 674 0.500 0.210 0 0.510 1.131
EBIT to sales 674 0.140 0.150 -0.772 0.124 0.596
Market to book 674 3.882 2.317 -2.733 3.437 10.378
R&D to total assets 674 0.040 0.046 0 0.025 0.201
Total assets (in millions ($))
674 20,828.653 59,848.666 182,743 6,276.381 750,507
4.2. Correlation matrix
The Pearson correlation matrix is presented in table 3. According to Blumberg, Cooper and Schindler (2014) multicollinearity may be an issue when a correlation between independent variables is equal to or (lower) higher than (-)0.7. After verifying the correlation matrix, there are no correlations exceeding these thresholds. This implies that there is no indication of potential multicollinearity despite significant and relatively high correlations between total assets and CEO power; R&D to total assets and debt to total assets (0.583, -0.436, respectively, with corresponding p<0.01 values)34.
Interestingly, the dependent variable pay duration has no high correlation with any of the variables of interest for this research. Consistent with Gopalan et al. (2014), pay duration is negatively related to CEO age and debt to total assets, whilst it is positively correlated to market to book and R&D to total assets. In contrast to Gopalan et al. (2014) and Cadman and Sunder (2014), total assets is negatively correlated to pay duration.
33 For example, the average age of a CEO is 55 years (st. dev. 7) and the market value of equity is on average 3.882
times higher than the book value of equity (st. dev. 2.317).
34 To further indicate that multicollinearity is not an issue, a variance inflation factor is performed after including
time-specific dummies. The highest VIF value is 4.280, which is below the critical threshold of 10 (Neter, Wasserman, and Kutner, 2004). In addition, similar results in the analysis are obtained by inserting the variables one by one, which may also indicate that multicollinearity is not an issue in this sample.
Table 3: Pearson correlation matrix Pay duration Institutional investor shareholding (ISS)
CEO power CEO age Debt to total
assets EBIT to sales Market to book R&D to total assets Total assets Pay duration 1 IIS -0.021 1 CEO power 0.046 -0.246*** 1 CEO age -0.092** -0.041 0.021 1
Debt to total assets -0.132*** -0.005 0.412*** 0.099** 1
EBIT to sales 0.006 0.036 -0.002 -0.106*** 0.079** 1
Market to book 0.086** -0.006 -0.053 0.054 -0.038 0.181*** 1
R&D to total assets 0.101*** -0.095** -0.280*** 0.085** -0.436*** -0.134*** 0.245*** 1
Total assets 0.082* -0.194*** 0.583*** 0.061 0.373*** -0.166*** 0.165*** -0.242*** 1
* Correlation is significant the following levels *** p<0.01.
** p<0.05. * p<0.1.
4.3. Regression results
Before the regression is estimated, a Hausman test is performed to assess whether a fixed-effects or random-effects model is more appropriate (Hausman, 1978). Because the Hausman hypothesis is rejected (χ2 = 4.130, p=0.765), a random-effects model is used to estimate the coefficients35. Furthermore, the standard errors are robust for heteroscedasticity by using Huber-White standard errors. The regression results are provided in table 4. More precisely, Model I presents the control variables, whereas Model II and Model III respectively report the direct effects of institutional investor shareholding and CEO power. Finally, Model IV adds the interaction term.
Based on the view that institutional investors can influence the pay process of firms and focus on maximizing long-term value, hypothesis 1 states that institutional ownership is positively associated with a longer CEOs pay duration. According to the hypothesis, a positive and significant coefficient is expected. Although the coefficient is positive, it is insignificant (β=0.129, p=0.319). Therefore, there is no evidence that institutional investors are associated with a longer CEOs pay duration. Hypothesis 1 is thus not accepted.
The results presented in column III provide a direct test for hypothesis 2. Based on the MPT, if equity-based components are provided to a CEO, a powerful CEO is likely to shorten the corresponding vesting schemes of his equity-based components. As a result, hypothesis 2 proposes that more powerful CEOs are negatively associated with pay duration. Thus, a negative and significant coefficient is expected. However, the CEO power coefficient is positive and insignificant (β=0.042, p=0.233). Therefore, there is no evidence that more powerful CEOs are negatively associated with her or his pay duration. To conclude, hypothesis 2 is thus not accepted.
Hypothesis 3 predicts that the positive association of institutional ownership on pay duration is weakened by the presence of a more powerful CEO. A significant negative interaction term would indicate that the effect of institutional investors on pay duration decreases as CEO power increases. As provided in Model IV, the interaction term is negative and significant at a 10% level (β=-0.403, p=0.050). Thus, this indicates that there is weak evidence that the influence of institutional ownership on pay duration decreases, when the CEO is more powerful. However, as put forward by Model II and Model IV, institutional investors are not significantly associated with lengthened pay duration in compensation contracts of CEOs. In fact, based on Model IV, the negative (insignificant) association of institutional investors is further strengthened by the interaction term. Therefore, the hypothesis is not accepted.
35 From an economic perspective, the usage of a random-effects model is an appropriate choice, because this study
primary focuses on arbitrary drawn U.S. largest firms that have certain characteristics. See Wooldridge (2014, p. 351) for an extensive discussion about the usage of random effects and fixed effects.
To conclude, the statistics in this paragraph indicate unexpected results. The next paragraphs, therefore, move on to verify and discuss the outcomes.
Table 4: Random-effects regression output for pay duration36
Variables: Model I Model II Model III Model IV
Independent variables Institutional investor shareholding (ISS) 0.129 (0.319) -0.128 (0.325) CEO power 0.042 0.314** (0.058) (0.168)
(ISS x CEO power) -0.403*
(0.245)
Control variables
CEO age -0.093 -0.099 -0.125 -0.126
(0.099) (0.100) (0.103) (0.105)
Debt to total assets -0.730** -0.716*** -0.614** -0.664**
(0.305) (0.308) (0.309) (0.312)
EBIT to sales -0.511* -0.496** -0.177 -0.210
(0.268) (0.277) (0.228) (0.225)
Market to book 0.039** 0.038** 0.023* 0.022
(0.019) (0.019) (0.017) (0.017)
R&D to total assets 0.908 1.131 1.935* 1.769
(1.286) (1.293) (1.348) (1.391) Total assets 0.299*** 0.304*** 0.211* 0.208* (0.114) (0.114) (0.149) (0.145) Constant 0.589 0.472 0.863* 0.994** (0.413) (0.483) (0.549) (0.550) Observations 674 664 624 617 F-test 18.270*** 18.980*** 21.040*** 24.23*** R2 0.059 0.060 0.065 0.066
Note: One-tailed tests are reported for the independent variables and the interaction term. Robust standard errors are in parentheses.
*** p<0.01. ** p<0.05. * p<0.1.
4.4. Evidence using similar variables
To verify the outcomes, the analyses are re-run with different variables. Starting with institutional investor shareholding, the main analysis assumes that aggregated holdings of institutional investors provide opportunities to control the firm and its policies collectively (Davis and Thompson, 1994). Due to large monitoring costs, it is however likely that the free-rider problem is present. More precisely, some small institutional investors may benefit from the monitoring actions of large
36 Please note that the number of observations in all analyses differ. Therefore, when comparing statistics it is
institutional investors. Almazan, Hartzell and Starks (2005) support this view in an executive compensation setting process. Prior literature further proposes that the larger the shareholder, the more likely that the influence increases (e.g., Shleifer and Vishny, 1986). In particular, these larger institutional investors may collaborate to suppress their motives (Black, 1992). Consistent with Hartzell and Starks (2003), the influence of institutional ownership may therefore be more appropriate to test with the concentration of top five institutional investor shareholding and a Herfindahl Index of institutional investor ownership concentration37. Using both measures, significant, negative results for institutional investors are found (β=-0.923, p=0.035 and β=-2.484, p=0.074, respectively). Thus, the largest institutional investors possess capabilities to monitor the pay process. However, these results counter the theoretical underpinnings, that is, pay duration is in fact shortened. A potential explanation is that in this sample larger institutional investors may mainly consist of transient institutional investors. These are in fact more short-term oriented (Bushee, 1998). Regarding the interaction term, no significant results are found.
To further investigate the outcomes regarding CEO power, all CEO power proxies (i.e., CEO duality, CEO tenure, board size and board independence) are regressed separately. For all variables, insignificant results are obtained. Alternatively, the influence of CEO power may not be precisely described in an unidimensional construct (Henderson, Masli, Richardson and Sanchez, 2010). Therefore, an index is constructed to reflect the four facets of CEO power by following the procedures of previous studies (Grinstein and Hribar, 2004; Henderson et al., 2010; Morse et al., 2008). The procedures are described in Appendix A. In brief, the results are similar to the previously reported outcomes in table 4. Except for the interaction term, the result are mixed, and, in general, insignificant. Appendix Dprovides results for both institutional investor variables as well as CEO tenure and the CEO power index. Because other variables yield the same results, these are not tabulated.
4.5. Evidence using other statistical techniques
To verify the previous outcomes, four alternative statistical procedures are applied38. First, since many pay duration observations are clustered around zero (14.380% of the sample is actually zero) and the variable cannot become negative, a Tobit regression may be an appropriate method to estimate the coefficients. Second, ignoring cross-section and time effects, a simple ordinary least squares (OLS) method is performed. Furthermore, it is likely that pay contracts are correlated within firms during the
37 According to studies of Gopalan (2008), Aggrawal et al. (2015) and Dasgupta and Placentino (2015) the usage
of the top five institutional investors can be questioned, because both the top five’s individual and aggregated holding can be too small for influencing corporate policies. Therefore, these studies argue the usage of a 5% threshold for an institutional investor (institutional block holders). By using institutional block holding, insignificant results are however found in all cases. See also reported statistics in Appendix E.
38 Because the Breusch-Pagan Lagrange Multiplier test rejects the use of pooled OLS (χ2=113.170, p=0.000), this
sample period39. Therefore, OLS with clustered standard errors on firm level are examined. Third, the influence of both institutional investors and CEO power may be observable after several years, therefore the analyses are also run with lags for all of the independent variables. Finally, the analyses are re-estimated including industry and time dummies. In brief, the results do not differ in a material way. Except for the results of the interaction term, which are mixed. In general, all variables of interest are insignificant. The outcomes also hold for the proposed, alternative variables. To demonstrate, results using Tobit and (clustered) OLS for different variables are reported in Appendix E. Because results using different specifications yield the same outcomes, the latter two methods are not tabulated in Appendix E. The same applies to tests with other variables.
4.6. Endogeneity
An important challenge in executive compensation literature is endogeneity (Armstrong, Jagolinzer and Larcker, 2010)40. Although studies of Sánchez-Ballesta and García-Meca (2007) and Abernethy et al. (2015) point out that the endogeneity threat is alleviated by applying a random effects model and lagged independent variables, the effectiveness of both methods are to open debate. Therefore, an Instrumental Variables approach is performed to prevent unbiased outcomes for the main variable institutional investor shareholding.
The endogeneity issue between institutional investor ownership arises when institutional investors’ holdings are not randomly distributed between firms. Especially the increasing presence of institutional investors in U.S.’ largest firms may be determined by other factors related to the willingness to invest in a particular firm41. If some of these factors correlate with a CEO’s pay duration, then the measures of institutional investor shareholding could be correlated with the error term. As a consequence, if one does not adequately control for these factors in the regressions, the estimated coefficients may be biased. For example, it is possible that particular stock market characteristics drive the share price of stocks to their fundamental value, which may encourage firms to tie CEOs’ compensation packages to more stock-based components (Holmström and Tirole, 1993). In addition, these stock market characteristics may also explain the increasing presence of institutional investors.
To address the endogeneity issue with respect to the main relation between institutional investors and pay duration, a two-stage least squares analysis is performed. Following Hartzell and Starks (2003), the instrumental variable is share turnover. Share turnover captures the intuition that institutional investors prefer to invest in corporations with higher stock market liquidity (Falkenstein,
39 Gopalan et al. (2014) argue that pay contracts are highly correlated within industries. Therefore, they cluster
standard errors between industries. This assumption is however refuted by Cadman and Carter (2014). Ultimately, the unreported results do not materially differ in comparison to firm clustering.
40 Wooldridge (2014) defines endogeneity as the situation in which the error terms (ε
it) and the dependent variables
are contemporaneously correlated.
41 A firm may also have certain ownership preferences. For example, to prevent potential them from drops in share
price and management replacements (Brickley, Lease, and Smith, 1988; Bushee and Noe, 2000; Hotchkiss and Strickland, 2002). Remember, however, this indirect view is beyond the scope of the study.
1996; Del Guerco, 1996)42. From a statistical point of view, Bennett, Sias and Starks (2003) document a substantial correlation between institutional investor shareholding and share turnover. Therefore, share turnover may be classified as a valid instrument.
In the first stage, institutional investors shareholding is predicted by share turnover, controlling for industry and time43. Then, a normal OLS is performed to estimate the coefficients44. As provided in table 5,the results of the second stage do not significantly differ from the main analysis. By performing a Durbin-Wu-Hausman test, it is proven that the results are not driven by endogeneity (χ2=1.578, p=0.209). As a result, the prior estimated coefficient of institutional investor shareholding is thus not biased and inconsistent45.
Table 5: Regression estimates of both stages by applying the Instrumental Variables approach
Panel A: First stage of 2SLS Model I
Independent variable Share turnover 0.002 (0.027) Control variables CEO age -0.004 (0.012) Debt to assets 0.041 (0.036) EBIT to sales 0.050 (0.040) Market to book -0.002 (0.002)
R&D to total assets -0.346**
(0.163) Total assets -0.069*** (0.011) Constant 0.868*** (0.047) Observations 664 F-test 9.110*** R2 0.111 Adjusted R2 0.099
42 One may debate whether all institutional investors prefer more stock market liquidity. However, it seems to be
more likely that a sale in an illiquid market results in a downward price pressure. This implies that any institutional investors benefit from more liquid markets notwithstanding their horizon intentions.
43 The dummies for industry and time are unreported in the table 5.
44 An OLS is performed, since random effects may already mitigate potential endogeneity by (partly) removing
the heterogeneity in the error term.
45 Please note that a potential endogeneity issue in the interaction term (hypothesis 3) is ignored. I am unfortunately
Panel B: Second stage of 2SLS Model I
Independent variable
Predicted values of institutional investor shareholding -1.209
(1.118) Control variables CEO age -0.230*** (0.081) Debt to assets -0.726*** (0.214) EBIT to sales -0.325 (0.274) Market to book 0.045*** (0.017)
R&D to total assets 0.561
(1.037) Total assets 0.238** (0.010) Constant 1.751 (1.073) Observations 664 F-test 44.610*** R2 0.023 Adjusted R2 0.012
Note: One-tailed tests are reported for the independent variables. Robust standard errors are in parentheses.
*** p<0.01. ** p<0.05. * p<0.1.
4.7. Discussion
The main finding of this study does not support the idea that institutional investors are associated with providing more long-term horizon incentives to a CEOs’ compensation package. In addition, there is no evidence reported that CEO power is significantly associated with shortened pay duration. Furthermore, the prediction that the positive association of an institutional investor is constrained by the presence of a powerful CEO is also rejected. In general, the results hold when using other variables, such as an institutional block holding or a CEO power index. To explain the rather contradictory results, alternative theoretical explanations are provided.
First, studies of Shleifer and Vishny (1997) and La Porta, Lopez-de-Silanes, Shleifer and Vishny (1997; 1999; 2000) describe different characteristics of corporate governance systems around the world. In particular, the Anglo-Saxon model of corporate governance emphasizes, among other elements, a well-developed stock market. Many U.S. firms are publicly held and are characterized by widely dispersed ownership. This implies that there are less large shareholders (i.e., an ownership percentage of 51 or more). In addition, high stock market liquidity provides investors the opportunity to easily ‘exit’ their (small) stakes rather than employing a ‘voice’, if investments do not meet the expectations (Coffee,
