Is Chief Executive Officer Power Bad?*
E. Han Kim,Yao Lu
DOI: https://doi.org/10.1111/j.2041-6156.2011.01047.x
IF: 1.463
2011-01-01
Asia-Pacific Journal of Financial Studies
Abstract:Asia-Pacific Journal of Financial StudiesVolume 40, Issue 4 p. 495-516 Free Access Is Chief Executive Officer Power Bad?* E. Han Kim, Corresponding Author E. Han Kim Ross School of Business, University of Michigan Corresponding author: E. Han Kim, Ross School of Business, University of Michigan, Ann Arbor, MI 48109, USA. Tel: 734-764-2282, Fax: 734-936-6631, email: ehkim@umich.edu.Search for more papers by this authorYao Lu, Yao Lu School of Economics and Management, Tsinghua UniversitySearch for more papers by this author E. Han Kim, Corresponding Author E. Han Kim Ross School of Business, University of Michigan Corresponding author: E. Han Kim, Ross School of Business, University of Michigan, Ann Arbor, MI 48109, USA. Tel: 734-764-2282, Fax: 734-936-6631, email: ehkim@umich.edu.Search for more papers by this authorYao Lu, Yao Lu School of Economics and Management, Tsinghua UniversitySearch for more papers by this author First published: 15 August 2011 https://doi.org/10.1111/j.2041-6156.2011.01047.xCitations: 6 † Acknowledgments: We have benefitted from helpful comments and suggestions from Sugato Bhattacharyya, J. B. Chay, Amy Dittmar, Luo Jiang, Jin-Mo Kim, John McConnell, Adair Morse, Amiyatosh Purnannandam, and participants of finance seminars at Rutgers University and the State University of New York at Buffalo, the University of Michigan, the 2008 China International Conference in Finance, a corporate strategy seminar at the University of Michigan, and the Fifth International Conference on Asia-Pacific Financial Markets. We acknowledge financial support from the Mitsui Life Financial Research Center at the University of Michigan. AboutSectionsPDF ToolsRequest permissionExport citationAdd to favoritesTrack citation ShareShare Give accessShare full text accessShare full-text accessPlease review our Terms and Conditions of Use and check box below to share full-text version of article.I have read and accept the Wiley Online Library Terms and Conditions of UseShareable LinkUse the link below to share a full-text version of this article with your friends and colleagues. Learn more.Copy URL Share a linkShare onFacebookTwitterLinked InRedditWechat Abstract This paper focuses on abnormal chief executive officer (CEO) structural power over top executives and examines its impacts on CEO pay for performance sensitivity and firm performance. We find that greater abnormal power is associated with weaker firm performance, but the relation is significant only when monitoring by external shareholders is weak. We also identify a channel through which the power adversely impacts firm performance: CEOs' capture of the compensation process. Greater abnormal CEO power lowers CEOs' pay for performance sensitivity, but again the relation is driven by observations under weak external monitoring. External monitoring is measured by institutional ownership concentration; the abnormal power, by residuals of a regression relating CEO structural power to its likely determinants. The negative impact of the abnormal power on firm performance is robust to potential reverse causality. 1. Introduction The influence of chief executive officers (CEOs) on corporate policies and performance has been well documented. Bertrand and Schoar (2003) find that CEO fixed effects matter for a wide range of firm policies; Bennedsen et al. (2006) document that CEO deaths are strongly negatively correlated with firm profitability and growth; Cronqvist et al. (2009) show that differences in corporate financial leverage can be traced to CEOs' personal leverage; and Jenter and Lewellen (2011) provide evidence that CEO age approaching retirement has an important impact on the likelihood of their firms being taken over and on the takeover premiums that their shareholders receive. These CEO impacts should be greater when CEOs have more power, where power is defined as the capacity to exert one's own will on corporate decisions. Recent evidence on CEO power suggests that powerful CEOs are bad news for shareholders. Bebchuk et al. (2011) argue that concentration of power in the CEO reduces firm performance. Landier et al. (2008) show that strong CEO power over top executives hurts firm performance. Bertrand and Mullainathan (2000) demonstrate that in the absence of adequate monitoring by blockholders, CEOs manipulate the compensation process to pay themselves what they can. Bebchuk and Fried (2004) argue that powerful CEOs reduce the linkage between CEO compensation and firm performance. Morse et al. (2010) show that powerful CEOs rig incentive contracts. With all these negative impacts of CEO power, why do firms grant power to CEOs? In an ideal world, shareholders would grant an optimal level of power, weighing various costs and benefits specific to a firm's characteristics. For some firms, concentration of power in the CEO office helps to expedite decision-making processes, resulting in more timely and efficient reaction to internal and external problems or pro-action to anticipated changes in market conditions. Such benefits are evident in Adams et al. (2005), who find that powerful CEOs are associated with the best and the worst performances. We argue that the deleterious effects of CEO power arise from deviations from the optimal level. However, the deviation, abnormal CEO power, might not be bad; its desirability depends on how the power is used and whether its use is properly monitored and guided to protect and enhance shareholder value. In the present paper, we focus on the deviation; namely, the abnormal power beyond the customary CEO power given CEO and firm characteristics. Specifically, we measure the abnormal CEO power by residuals of a regression relating a power measure to its likely determinants. This measure of abnormal power is related to a channel through which power might impact firm performance: CEOs' pay for performance sensitivity (PPS). Our purpose is to examine whether power is used to capture the compensation process. We also consider whether the likelihood of the capture is affected by the intensity of monitoring by external shareholders. Because any capture of the compensation process is likely to adversely affect shareholder value, the final phase of our investigation focuses on the relation between the abnormal CEO power and firm performance, which is measured by Tobin's Q or return on assets (ROA). We find that abnormal CEO power is negatively related to PPS, and the negative relation is driven by observations under weak external monitoring (EM). The intensity of monitoring by external shareholders is measured by institutional ownership concentration (IOC). When IOC is high, CEO power has no effect on PPS or firm performance. Only when monitoring by the external shareholders is weak does the abnormal power reduce PPS and firm performance. When IOC is below the sample median, the median PPS for CEOs with the most abnormal power is lower by $0.164 per $1000 change in shareholder value than the median PPS for the CEO with the least abnormal power. Although $0.164 seems small in magnitude, it represents a 33% decrease in PPS relative to the PPS for the CEO with the least abnormal power. As for firm performance, when IOC is below the sample median, a one standard deviation increase in CEO abnormal power leads to a 0.342% decrease in Tobin's Q and a 1.216% decrease in ROA. The sample mean Q and ROA are 2.148 and 4.098%, respectively. These relations concerning firm performance and the abnormal power are subject to reverse causality, which we address by estimating three-stage least square regressions. Although the estimation results do not rule out the possibility that our measures of firm performance affect the abnormal power or IOC, our main conclusion is robust; namely, abnormal CEO power is bad for shareholder value when firms are subject to weak monitoring by external shareholders. This study adds to the emerging literature on CEO impact and power. We use the measure of abnormal CEO structural power of Landier et al. (2008) and identify a negative relation between the power and firm performance that is robust to firm fixed characteristics and possible reverse causality. Furthermore, we demonstrate that the negative relation is at work only when monitoring by external shareholders is weak. When monitoring is strong, the abnormal CEO power seems benign. In addition, we reveal that CEOs' PPS is a channel through which the power affects performance. The rest of the paper is organized as follows. Section 2 develops hypotheses and defines the abnormal CEO power and EM. Section 3 describes our empirical design and data. Section 4 relates the abnormal power to PPS; and Section 5, to firm performance. Section 6 addresses endogeneity issues. Section 7 concludes. 2. Hypotheses Development This section presents hypotheses on how abnormal CEO power affects CEO pay for performance sensitivity (PPS) and firm performance. It also describes our measures of the abnormal power and the strength of monitoring by external shareholders. We begin with two non-mutually exclusive views of CEO power, the contracting and the capturing view. 2.1. Contracting and Capturing Views In the contracting view, shareholders determine the optimal level of CEO power by trading off efficiency gains from having a more centralized CEO office against risks of abusing the power for private benefits. Shareholders also hold CEOs accountable for their actions through incentive contracts linking compensation to performance. However, asymmetric information and transaction costs in the market for CEOs might lead to deviations from optimal contracting. Some CEO characteristics are unknown to their employers at the time of hiring. When subsequent realizations reveal that the power granted to a CEO exceeds the optimal level, re-contracting is costly because the CEO is likely to resist relinquishing the abnormal power and replacing a CEO is expensive. These re-contracting costs arise because of investments required to acquire firm-specific CEO skills, uncertainty about the quality and availability of other suitable CEO candidates, and search costs, among others. It is possible to reduce these costs ex-ante by initially giving a CEO a very low level of power and later adjusting it when his or her true characteristics are revealed. However, this strategy imposes other costs: interim efficiency losses during the adjustment process. When a firm is run by a CEO with abnormally weak influence, it might function like a firm run by committees. Such firms might not be able to efficiently react to or proactively manage changes in internal and external business environments in a timely manner. Furthermore, a qualified candidate with other comparable employment opportunities will not choose a CEO position that comes with unreasonably weak power, making the ex-ante solution untenable. Therefore, asymmetric information, re-contracting costs, interim efficiency, and competition between firms may lead to situations in which a CEO's power deviates from the optimal level. The positive deviations from the optimal level of CEO power allow for the capturing view, which is an extension of the skimming view articulated by Bertrand and Mullainathan (2000) and others (e.g. Crystal, 1991; Milbourn, 2003; Morse et al., 2010). The skimming view posits that when ownership is diffuse and without effective oversight by shareholders, powerful CEOs capture the compensation process and set their own pay. This view assumes that most CEOs would rather tilt the compensation process in their favor at the expense of firm performance. At any given point in time, the PPS reflects what remains after CEOs capture the governance process. We define power as the capacity to exert one's own will, as in Finkelstein (1992), who defines four dimensions of managerial power: structural power, ownership power, expert power, and prestige power. Structural power is based on organizational structure and hierarchical authority, both formal and informal.11 Ownership power stems from equity ownership. In another paper, we analyze how CEO ownership affects firm performance and risk-taking (Kim and Lu, 2011). Expert power arises from the ability to contribute to organizational success by influencing a particular strategic choice through functional expertise. Prestige power represents personal prestige, status of reputation, and others' perception of CEO influence through contacts and qualifications. Expert and prestige power are similar in that both arise from personal abilities to make contributions to firm performance. These ability-based powers are unlikely to harm firm performance. In an attempt to measure ability-based powers, Milbourn (2003) and Rajgopal et al. (2006) use the number of times a CEO is mentioned in the press as a proxy for CEO reputation and outside employment opportunity. The aforementioned studies demonstrating risks of CEO power focus on this dimension of power.22 Managerial power has been a subject of much research in the management literature. See Adams et al. (2005) for a brief review from the finance/governance perspective. The focus of the present paper is also on CEO structural power. 2.2. Hypotheses Our predictions are about how abnormal CEO structural power affects PPS and firm performance. We assume that CEOs prefer low PPS: they want more compensation for stronger performance and want to maintain the high level of compensation when performance is weak. To achieve this, CEOs might rig incentive contracts by altering the benchmark when performance is weak (Morse et al., 2010). Tweaking the performance benchmark would require other top executives' support, which is easier to obtain when a CEO has strong structural power over his or her top executives. Hence, we predict: Hypothesis 1. Abnormal CEO power is negatively related to PPS. We interpret the negative relation between abnormal CEO power and PPS as evidence of the abnormal power helping CEOs capture the governing process. When governance is captured by CEOs, their private benefits will receive higher priority than shareholder value maximization. Hence, we hypothesize: Hypothesis 2. Abnormal CEO power is negatively related to firm performance. We also predict that the effects of CEO power on both PPS and firm performance depend on the strength of monitoring by external shareholders. Strong EM helps to curb the negative effects of abnormal CEO power. In our paper on CEO ownership (Kim and Lu, 2011), we find that both the benefits and deleterious effects of CEO ownership power are prevalent only when EM is weak. Therefore, we predict: Hypothesis 3. Effects of abnormal CEO power on PPS and firm performance are most pronounced when EM is too weak to preempt or restrain the negative effects of CEO power. Conversely, when EM is strong, the effects on PPS and performance are less noticeable. 3. Empirical Design and Data 3.1. Proxy for Chief Executive Power We measure the abnormal CEO structural power by the excess fraction of top executives hired by a CEO, as in Landier et al. (2008). CEOs are heavily involved in recruiting their top lieutenants; consequently, top executives during a CEO's tenure are more likely to share similar preferences and to be more loyal to the CEO. These executives are less likely to dissent, whereas executives accustomed to working with a previous CEO are more likely to challenge orders received from a new CEO. Therefore, we assume the greater the fraction of top executives hired by a CEO, the greater the CEO's informal hierarchical authority and, hence, the stronger is his or her structural power over the executives. The fraction of top executives hired during a CEO's tenure tends to be related to a number of factors, such as the tenure of the CEO and of other top executives and whether the CEO is recruited from outside. Therefore, we follow Landier et al. and measure the abnormal fraction hired by CEOs with residuals of the following regression: (1) Here, Frac is the fraction of executives hired by the CEO; CEOTEN, CEO's tenure (in years);33 Previous studies relying on ExecuComp to obtain CEO tenure report non-trivial numbers of negative CEO tenure. We trace the negative tenure to the practice of ExecuComp reporting a CEO starting year only for the latest appointment. Thus, if a CEO leaves the position and returns later, relying on an ExecuComp start date will give a negative tenure. We correct for this problem by backtracking the previous appointment year using the CEO name. EXECSEN, average non-CEO executive tenure; OUTSIDE, an indicator variable equal to one if the CEO comes from outside the firm; KNOWN, the fraction of executives for which tenure is reported in the data; and FRAC_1Y, the fraction of executives who arrived within a year of the CEO's nomination. The regression also controls for year fixed effects to account for macroeconomic factors affecting top executive hiring decisions. Table 1 contains definitions of all variables. Table 1. Definitions of variables used in performance and pay-for-performance-sensitivity analyses Variable Panel A: Variables used in performance analyses Power Abnormal fraction of executives hired by the firm during the current CEO's tenure. The internal rank of executives is based on the sum of salaries and bonuses Tobin's Q The market value of common equity plus the book value of total liabilities divided by the book value of total assets ROA Return on total assets FirmAge Log of one plus the listing age of a firm measured by the number of years from the firm's initial public offering as reported in CRSP or the number of years since its first appearance in CRSP LNS Log (Sales) K/S The ratio of property, plant, and equipment to sales I/K The ratio of capital expenditures to property, plant, and equipment Female Indicator equal to one if a CEO is female, and zero otherwise CEOAge Log(CEO age) CEO_Chair Indicator equal to one if a CEO also chairs the board, and zero otherwise IOC The sum of percentage share ownership of the top five institutional investors IOC_IY(−i) The mean value of IOC of all firms in the same industry in a given year, excluding firm i itself Panel B: Variables used in pay-for-performance-sensitivity analyses Ch_Flow_CEO_Com ($K) The change in total direct CEO compensation flow (including salaries, bonuses, options, stock grants, and other compensation) in 2000 dollars from year t − 1 to year t Ch_sv ($MM) The change in shareholder value, measured by the product of shareholder value in 2000 dollars at year t − 2 and the geometric mean of shareholder rate of returns from t − 2 to t CDF(Power) The cumulative density function of the power variable CDF(CEO Age) The cumulative density function of the current age of the CEO CDF(Firm Size) The cumulative density function of log total assets CDF(Risk) The cumulative density function of the variance of daily stock returns during the current year The residuals in the regression may be viewed as abnormal fractions of executives hired by CEOs. This is our proxy for abnormal CEO structural power over other top executives.44 Other proxies used to estimate CEOs' structural power include CEO centrality as measured by the pay gap between CEO and other top executives (Bebchuk et al., 2008; Morse et al., 2010) and CEO being the only inside member of the board (Adams et al., 2005; Bebchuk et al., 2008). We do not use these proxies because they are multidimensional. For CEO centrality, the contracting view suggests that compensation differences between a CEO and his or her top executives should reflect not only the concentration of his or her structural power but also differences in expert and prestige power, which is likely to have a favorable impact on firm performance. As for the CEO being the only member of the board, it has a duality: the CEO might have more structural power over other executives, but he might have less influence over the board because it has more outside directors. Finally, CEO tenure is not used as a proxy for CEO power because it contains both structural power obtained over time and ability-based power as reflected by the ability to hold on to the job. Table 2 shows the regression result. Consistent with Landier et al. (2008), the fraction of executives appointed during a CEO's tenure is positively related to the length of CEO tenure and negatively related to the average non-CEO executive tenure. A substantial fraction of new hiring seems to be done during a CEO's first year in office, and a CEO hired from outside tends to hire fewer top executives during his or her tenure. Table 2. Regression to construct abnormal fraction (Frac) of top-executives hired during a chief executive officer's (CEO's) tenure This table reports regression estimates to construct the abnormal fraction of top executives hired during a CEO's tenure. Power variable is the residual of the regression in which the dependent variable is the percentage of top executives hired by the firm during a CEO's tenure. The regression controls for year fixed effects. Robust standard errors are in parentheses. Coefficients marked with ***, **, and * are significant at 1, 5, and 10%, respectively. Frac (1) CEOTEN 0.002*** (0.000) EXECSEN −0.001*** (0.000) OUTSIDE −0.011*** (0.002) KNOWN −0.007* (0.004) FRAC_1Y 0.835*** (0.005) Constant −0.012*** (0.003) Year fixed effects Y Observations 20 730 Adjusted R2 0.83 3.2. Strength of External Monitoring Our proxy for the strength of EM is IOC. Previous researchers demonstrate the important monitoring role of institutional investors and block holders in shaping corporate governance (e.g. Shleifer and Vishny, 1986; Bertrand and Mullainathan, 2000, 2001; Hartzell and Starks, 2003; Cremers and Nair, 2005; Del Guercio et al., 2008; Edmans, 2009). We follow Hartzell and Starks (2003) and estimate IOC by the percentage of institutional holdings by top five institutions. The results are robust to measuring IOC by the Herfindahl Index of institutional ownership. Institutional ownership data is obtained from the CAS Spectrum database. 3.3. Data and Summary Statistics We use panel data from 1994 through 2006, constructed by merging the executive data in ExecuComp with accounting data in Compustat and stock return data in CRSP. We drop firm-year observations in which a new CEO's first year in office overlaps with the last year of the previous CEO. Our total sample consists of 11 474 firm-year observations associated with 1397 unique firms over the period 1994–2006. Table 3 shows the number of observations by year for the full sample, and separately for high IOC and low IOC. High IOC (HIOC) and Low IOC (LIOC) are defined as those with above and below the sample median IOC.55 We use a pooled sample median as the demarcation point rather than the yearly median because of the time trend. IOC has been increasing steadily over time: the mean IOC is 18% in 1992, 24% in 1999, and 30% in 2006. Therefore, classifying 2006 firm observations into high and low IOC by the 2006 median might wrongly classify a firm-year into low IOC when it has a high IOC relative to the entire sample. The number of LIOC observations is greater (smaller) than HIOC observations in the earlier (later) years because of the steady increase in IOC over time. Sample size for individual regressions varies depending on the availability of data to construct dependent and independent variables. Table 3. Number of observations by year This table shows the number of observations by year. Column (2) reports the number of firms in the full sample by year. Columns (3) and (4) report the number of firms in the HIOC and LIOC sample, respectively. IOC is defined as the sum of percentage share ownership of the top 5 institutional investors. HIOC and LIOC are defined as above and below the sample median IOC. Year (1) Full (2) HIOC (3) LIOC (4) 1994 577 154 283 1995 610 189 293 1996 645 191 327 1997 692 253 309 1998 765 286 343 1999 867 339 382 2000 930 389 398 2001 928 406 401 2002 963 463 398 2003 1031 457 458 2004 1082 614 361 2005 1130 691 336 2006 1254 755 307 Total 11 474 5187 4596 Table 4 provides summary statistics for key variables. The average fraction of top four executives hired during a CEO's tenure is 15.6%. Importantly, the mean and median of our proxy for abnormal CEO power are close to zero. A number of variables, especially changes in CEO compensation flow and shareholder value, indicate the presence of large outliers even after winsorizing them at 1 and 99%. We are mindful of these outliers and design our estimation of PPS to mitigate the outlier problems. The difference in the number of observations across variables is due to missing variables.66 We also exclude observations with negative market-to-book value ratios. Table 4. Summary statistics for all variables used in performance and pay-for-performance-sensitivity analyses Definitions of the variables are given in Table 1. Panel A reports the summary statistics for the variables used in performance analyses. Panel B reports the summary statistics for the variables used in pay-for-performance sensitivity analyses. Variable Observation (1) Mean (2) Median (3) SD (4) Minimum (5) Maximum (6) Panel A: Variables used in performance analyses FRAC 11 474 0.156 0.000 0.248 0.000 1.000 Power 11 479 −0.003 0.010 0.106 −0.840 0.191 Tobin's Q 11 479 2.148 1.548 2.560 0.398 105.090 ROA 10 788 4.098 4.896 16.825 −577.850 233.264 FirmAge 11 423 2.880 2.996 0.897 0.000 4.407 LNS 10 783 7.312 7.222 1.564 −2.279 12.578 K/S 10 468 0.883 0.451 2.923 0.000 244.333 I/K 10 343 0.125 0.098 0.103 0.000 4.302 Female 11 479 0.014 0.000 0.119 0.000 1.000 CEOAge 10 920 4.014 4.025 0.138 3.367 4.511 CEO_Chair 11 479 0.646 1.000 0.478 0.000 1.000 IOC 9783 0.259 0.251 0.104 0.000 0.972 IOC_IY(−i) 9565 0.261 0.263 0.041 0.107 0.415 Panel B: Variables used in pay-for-performance sensitivity analyses Ch_Flow_CEO_Com ($K) 10 740 226.424 91.250 2194.579 −9492.284 9514.891 Ch_sv ($MM) 13 091 444.756 113.294 1611.402 −7118.041 13 868.280 CDF(Power) 8593 0.503 0.504 0.293 0.000 1.000 CDF(CEO Age) 13 170 0.551 0.600 0.284 0.000 1.000 CDF(Firm Size) 13 960 0.533 0.544 0.279 0.000 1.000 CDF(Risk) 14 724 0.488 0.481 0.285 0.000 1.000 4. Pay for Performance Sensitivity In this section, we examine whether abnormal CEO power helps CEOs capture the compensation process by relating changes in total CEO compensation flows to changes in shareholder value. Our specification closely resembles those of Hartzell and Starks (2003), Milbourn (2003), and Rajgopal et al. (2006). We estimate: (2) Here, ΔTotal Compensationit is the change in total direct compensation flow (including salaries, bonuses, value of new stocks and stock options granted, and other compensation) to the CEO of firm i in 2000 dollars from year t − 1 to year t. It is winsorized at the 1 and 99 percentile. ΔShareholder Valueit is changes in shareholder value from t − 2 to t, equal to shareholder value in 2000 dollars at t − 2 multiplied by the geometric mean rates of returns from t − 2 to t. We use the mean shareholder returns over 2 years to allow CEO compensation to reflect performance not only during the concurrent year but also during the prior year. Control variables include: firm size, as measured by log of total assets, to account for the well-documented relation between firm size and PPS (e.g. Schaefer, 1998; Baker and Hall, 2004); risk measured by the variance of stock return to account for a possible correlation between the uncertainty surrounding the firm and output based pay (Lafontaine and Bhattacharyya, 1995; Aggarwal and Samwick, 1999; Prendergast, 2002); and CEO age because it is shown to be related to PPS (Gibbons and Murphy, 1992; Milbourn, 2003). Power and control variables are lagged by 1 year. Because ΔTotal Compensation is calculated based on the change in total direct compensation flow of the CEO from t − 1 to year t, we drop firm-year observations in which a new CEO's first year in office overlaps with the last year of the previous CEO's tenure. We follow Aggarwal and Samwick (1999) and Milbourn (2003) and use cumulative density functions for variables interacted with shareholder value change. Using cumulative density functions reduces the importance of outliers by normalizing the variables to the unit interval. It also helps to interpret the estimated coefficients β1 and β2 in an economically meaningful way, revealing how the power affects PPS. We estimate two regressions: equation (1), a median regression with industry fixed effects to account for industry differences in PPS (Murphy, 1999), where industry is defined by the Fama–French 48 industry groupings; and equation (2), an OLS regression with firm fixed effects.77 Stata does not allow estimation of median regressions with firm fixed effects. OLS mean regressions are estimated to check the robustness to firm fixed effects. Both regressions include year fixed effects. Median regressions are also used by Hall and Liebman (1998), Milbourn (2003), and Rajgopal et al. (2006). They are more robu