Upload
others
View
1
Download
0
Embed Size (px)
Citation preview
Is Chief Executive Officer Power Bad?*
E. Han Kim**Ross School of Business, University of Michigan
Yao LuSchool of Economics and Management, Tsinghua University
Received 6 December 2010; Accepted 23 May 2011
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 perfor-
mance. 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 iden-
tify 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 nega-
tive impact of the abnormal power on firm performance is robust to potential reverse causality.
Keywords Chief executive officer power; Pay for performance sensitivity; Firm performance;
Institutional investor concentration
JEL Classification: G30, G34, J33, L25, M52, G32
1. Introduction
The influence of chief executive officers (CEOs) on corporate policies and perfor-
mance 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
*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 Interna-
tional 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.
**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: [email protected].
Asia-Pacific Journal of Financial Studies (2011) 40, 495–516 doi:10.1111/j.2041-6156.2011.01047.x
� 2011 Korean Securities Association 495
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 share-
holders. 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 execu-
tives hurts firm performance. Bertrand and Mullainathan (2000) demonstrate that in
the absence of adequate monitoring by blockholders, CEOs manipulate the compen-
sation process to pay themselves what they can. Bebchuk and Fried (2004) argue that
powerful CEOs reduce the linkage between CEO compensation and firm perfor-
mance. 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, concen-
tration of power in the CEO office helps to expedite decision-making processes, result-
ing 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 per-
formance 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 inten-
sity 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
E. H. Kim and Y. Lu
496 � 2011 Korean Securities Association
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 repre-
sents a 33% decrease in PPS relative to the PPS for the CEO with the least abnor-
mal 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 sub-
ject 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 conclu-
sion 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
Is Chief Executive Officer Power Bad?
� 2011 Korean Securities Association 497
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 invest-
ments 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 adjust-
ment 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 unreason-
ably 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 cap-
turing 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 effec-
tive oversight by shareholders, powerful CEOs capture the compensation process and
set their own pay. This view assumes that most CEOs would rather tilt the compen-
sation 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, owner-
ship power, expert power, and prestige power. Structural power is based on organi-
zational structure and hierarchical authority, both formal and informal.1 The
aforementioned studies demonstrating risks of CEO power focus on this dimension
of power.2 The focus of the present paper is also on CEO structural power.
1Ownership 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 rep-
utation, 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 contribu-
tions 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.2Managerial 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.
E. H. Kim and Y. Lu
498 � 2011 Korean Securities Association
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 compen-
sation for stronger performance and want to maintain the high level of compensa-
tion when performance is weak. To achieve this, CEOs might rig incentive contracts
by altering the benchmark when performance is weak (Morse et al., 2010). Tweak-
ing 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 exec-
utives 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.
Is Chief Executive Officer Power Bad?
� 2011 Korean Securities Association 499
and measure the abnormal fraction hired by CEOs with residuals of the following
regression:
Fracit ¼ a0 þ a1CEOTENit þ a2EXECSENit þ a3OUTSIDEit þ a4KNOWNit
þ a5FRAC 1Yit þ Yeart þ eit : ð1Þ
Here, Frac is the fraction of executives hired by the CEO; CEOTEN, CEO’s tenure
(in years);3 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.
The residuals in the regression may be viewed as abnormal fractions of execu-
tives hired by CEOs. This is our proxy for abnormal CEO structural power over
other top executives.4 Table 2 shows the regression result. Consistent with Landier
et al. (2008), the fraction of executives appointed during a CEO’s tenure is posi-
tively 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.
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,
3Previous 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.4Other 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 cen-
trality, 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.
E. H. Kim and Y. Lu
500 � 2011 Korean Securities Association
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. Insti-
tutional ownership data is obtained from the CAS Spectrum database.
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
Is Chief Executive Officer Power Bad?
� 2011 Korean Securities Association 501
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 obser-
vations 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.5 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 depend-
ing on the availability of data to construct dependent and independent variables.
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, indi-
cate 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
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 (Frac) 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
5We 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.
E. H. Kim and Y. Lu
502 � 2011 Korean Securities Association
outlier problems. The difference in the number of observations across variables is
due to missing variables.6
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:
DTotal Compensationit ¼ b0 þ b1DShareholder Valueit þ b2DShareholder Valueit
� CEO Powerit�1 þ b3CEO Powerit�1
þX6
j¼4
bjDShareholder Valueit � Control Variablesijt�1
þX9
j¼7
bjControl Variablesijt�1 þ Firmi ðor IndustryiÞ
þ Yeart þ eit : ð2Þ
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, respec-
tively. 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
6We also exclude observations with negative market-to-book value ratios.
Is Chief Executive Officer Power Bad?
� 2011 Korean Securities Association 503
Here, DTotal 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. DShareholder Valueit is changes in share-
holder 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,
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
E. H. Kim and Y. Lu
504 � 2011 Korean Securities Association
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 DTotal 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 cumula-
tive 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 b1
and b2 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.7 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 robust to the presence of large
outliers than mean regressions. This robustness is important because summary sta-
tistics in Table 4 indicate the presence of large outliers in changes of CEO compen-
sation flow and of shareholder value even after winsorizing. The precision of
estimates from a median regression is higher, because the median regression is a
more robust estimate of central tendency than the mean regression.8
4.1. Full Sample results
Table 5 reports the regression estimates relating abnormal CEO power to PPS. The
coefficients of interest are those on the interaction terms, especially the interaction
between shareholder value changes and the power variable. A positive coefficient
means abnormal CEO power increases PPS; a negative coefficient means abnormal
power decreases PPS.
Column (1), using the median regression, indicates that abnormal power has a
negative and significant impact on PPS. The OLS estimation result in Column (2)
also shows a negative but insignificant coefficient. The overall evidence suggests that
abnormal power helps CEOs capture the compensation process; however, the statis-
tical significance is sensitive to the choice of model specification.
Results on control variables are consistent with previous findings: PPS is
negatively related to firm size (Schaefer, 1998; Baker and Hall, 2004), and the med-
ian regression shows a positive relation between PPS and risk, which is consistent
with Rajgopal et al. (2006) and Morse et al. (2010). According to Lafontaine and
7Stata does not allow estimation of median regressions with firm fixed effects. OLS mean
regressions are estimated to check the robustness to firm fixed effects.8See Koenker and Hallock (2001) for more in-depth discussion on quantile regression.
Is Chief Executive Officer Power Bad?
� 2011 Korean Securities Association 505
Bhattacharyya (1995) and Prendergast (2002), firms with low uncertainty are easier
to monitor and, hence, are less reliant on pay incentives.
4.2. Interactive Effects with External Monitoring
The results based on the full sample mask important heterogeneity. The relation
between abnormal CEO power and PPS might be stronger for firms under weaker
EM, which makes it easier for CEOs to use their power to influence the compensa-
tion process. Strong EM, in contrast, might make it difficult to capture the compen-
sation process. To test this hypothesis, we separate firm-year observations into
strong and weak EM subsamples. An observation is considered to be under strong
(weak) EM if its IOC is above (below) the sample median. Separate estimation for
each subsample allows for coefficients of the independent variables and fixed effects
to vary across strong and weak EM regimes.
Table 6 reports the subsample results. The results reveal no relation between
abnormal power and PPS when EM is strong (odd numbered columns), but a sig-
nificant negative relation when EM is weak (even numbered columns). This is true
regardless of whether we use the median regression specification or OLS with firm
Table 5 Relation between chief executive officer (CEO) power and pay-for-performance
sensitivity
Regression in column (1) is estimated by using the median regressions with year and industry fixed
effects. Industries are defined as in Fama and French (1997). The regression in column (2) is estimated
by using the OLS regression with year and firm fixed effects. Definitions of all variables are given in
Table 1. The sample is constructed by excluding the year of CEO turnover and the following year. Stan-
dard errors are reported in parentheses. Coefficients marked with ***, **, and * are significant at 1, 5,
and 10%, respectively.
Ch_Flow_CEO_Com ($K)
Median regression
(1)
OLS regression
(2)
Ch_sv ($MM) 0.694*** (0.047) 0.870*** (0.178)
Ch_sv ($MM) · CDF(Power)t)1 )0.040* (0.021) )0.056 (0.082)
CDF(Power)t)1 46.861* (27.017) )78.968 (133.716)
Ch_sv ($MM) · CDF(CEO Age)t)1 )0.039* (0.023) )0.135 (0.093)
CDF(CEO Age)t)1 )63.310** (29.178) )138.121 (158.593)
Ch_sv ($MM) · CDF(Firm Size)t)1 )0.597*** (0.045) )0.637*** (0.171)
CDF(Firm Size)t)1 194.600*** (37.328) )1282.532*** (436.178)
Ch_sv ($MM) · CDF(Risk)t)1 0.164*** (0.026) )0.017 (0.097)
CDF(Risk)t)1 66.017 (40.592) 88.457 (243.197)
Constant )312.524 (265.055) 1031.348*** (315.396)
Firm fixed effects and year fixed effects N Y
Industry fixed effects and year fixed effects Y N
Observations 9649 9649
Adjusted R2 (Pseudo R2) (0.0226) )0.06
E. H. Kim and Y. Lu
506 � 2011 Korean Securities Association
Tab
le6
Rel
atio
nb
etw
een
chie
fex
ecu
tive
offi
cer
(CE
O)
po
wer
and
pay
-fo
r-p
erfo
rman
cese
nsi
tivi
ty:
hig
han
dlo
win
stit
uti
on
alo
wn
ersh
ipco
nce
ntr
atio
n
Th
ista
ble
rep
ort
sre
gres
sio
nes
tim
ates
sep
arat
ely
for
firm
year
ob
serv
atio
ns
wit
hh
igh
and
low
inst
itu
tio
nal
ow
ner
ship
con
cen
trat
ion
(IO
C).
Co
lum
ns
(1)
and
(3)
rep
ort
the
resu
lts
esti
mat
edfo
rth
eH
IOC
sub
sam
ple
,co
lum
ns
(2)
and
(4)
rep
ort
the
resu
lts
esti
mat
edfo
rth
eL
IOC
sub
sam
ple
.H
IOC
and
LIO
Car
ed
efin
edas
abo
ve
and
bel
ow
the
sam
ple
med
ian
IOC
.R
egre
ssio
ns
inco
lum
ns
(1)
and
(2)
are
esti
mat
edb
yu
sin
gm
edia
nre
gres
sio
ns
wit
hye
aran
din
du
stry
fixe
def
fect
s.In
du
stri
esar
e
defi
ned
asin
Fam
aan
dF
ren
ch(1
997)
.R
egre
ssio
ns
inco
lum
ns
(3)
and
(4)
are
esti
mat
edb
yu
sin
gth
eO
LS
regr
essi
on
wit
hye
aran
dfi
rmfi
xed
effe
cts.
Defi
nit
ion
so
f
all
vari
able
sar
egi
ven
inT
able
1.A
llre
gres
sio
ns
con
tro
lfo
rfi
rman
dye
arfi
xed
effe
cts.
Ro
bu
stst
and
ard
erro
rsar
ere
po
rted
inp
aren
thes
es.
Co
effi
cien
tsm
arke
dw
ith
***,
**,
and
*ar
esi
gnifi
can
tat
1,5,
and
10%
,re
spec
tive
ly.
Ch
_F
low
_C
EO
_C
om($
K)
Med
ian
regr
essi
on
OL
Sre
gres
sio
n
HIO
C
(1)
LIO
C
(2)
HIO
C
(3)
LIO
C
(4)
Ch
_sv
($M
M)
0.93
1***
(0.1
11)
0.49
7***
(0.0
69)
1.50
6***
(0.3
47)
0.62
6**
(0.2
55)
Ch
_sv
($M
M)
·C
DF
(Po
wer
) t)
10.
008
(0.0
58)
)0.
164*
**(0
.030
))
0.07
4(0
.186
))
0.25
0**
(0.1
10)
CD
F(P
ower
) t)
1)
21.0
81(5
1.72
0)62
.916
(48.
225)
)22
8.73
6(2
23.5
68)
)64
.690
(240
.737
)
Ch
_sv
($M
M)
·C
DF
(CE
OA
ge) t
)1
)0.
215*
**(0
.061
)0.
043
(0.0
33)
)0.
413*
*(0
.186
))
0.08
0(0
.137
)
CD
F(C
EO
Age
) t)
1)
83.3
89(5
7.35
6)3.
837
(50.
594)
)60
7.14
8**
(278
.996
)31
.904
(270
.143
)
Ch
_sv
($M
M)
·C
DF
(Fir
mSi
ze) t
)1
)0.
783*
**(0
.102
))
0.33
9***
(0.0
64)
)1.
150*
**(0
.323
))
0.27
5(0
.246
)
CD
F(F
irm
Size
) t)
163
.311
(74.
107)
193.
368*
**(6
5.17
9))
1808
.376
**(7
49.6
24)
)12
80.0
67(8
36.9
75)
Ch
_sv
($M
M)
·C
DF
(Ris
k)t)
10.
375*
**(0
.069
))
0.01
3(0
.036
)0.
100
(0.2
08)
)0.
262*
(0.1
36)
CD
F(R
isk)
t)1
79.6
88(7
9.37
1))
45.8
11(7
3.14
8)59
2.50
2(4
11.8
06)
)25
.631
(458
.024
)
Co
nst
ant
)62
2.87
4(4
73.3
33)
156.
463
(343
.789
)68
7.20
1(4
27.0
38)
572.
442
(497
.382
)
Fir
mfi
xed
effe
cts
and
year
fixe
def
fect
sN
NY
Y
Ind
ust
ryfi
xed
effe
cts
and
year
fixe
def
fect
sY
YN
N
Ob
serv
atio
ns
4281
3608
4281
3608
Ad
just
edR
2(P
seu
do
R2)
(0.0
344)
(0.0
233)
)0.
05)
0.00
Is Chief Executive Officer Power Bad?
� 2011 Korean Securities Association 507
fixed effects. The negative relation between abnormal power and PPS is driven by
observations under weak EM.
The median regression estimation result in Column (2) implies that when EM is
weak, the median pay sensitivity for the CEO with the most abnormal power is
lower than that for the CEO with the least abnormal power by $0.164 per $1000
change in shareholder value. Considering that the PPS for the CEO with the least
power is $0.497 per $1000 change in shareholder value, the $0.164 decrease repre-
sents a 33% decrease in PPS.
Our estimated pay sensitivities are smaller than those reported by Milbourn
(2003) and Rajgopal et al. (2006), who use similar empirical methodology. The dif-
ference is in the dependent variable. Their dependent variable includes changes in
the value of all stocks and stock options held by CEOs and, hence, measures CEOs’
firm-related wealth change, whereas our dependent variable includes changes only
in compensation flows. When Hartzell and Starks (2003) estimate executives’ PPS
of salaries and bonuses, they show only $0.032 per $1000 change in shareholder
value. Our estimates are much greater than theirs because we include the value of
new grants of stocks and stock options in our measure of CEO compensation flows.
5. Firm Performance
How does the effect of abnormal CEO power on the compensation process show
up on the bottom line? If a CEO uses the power to capture the compensation
process and lowers his or her PPS, the alignment of his or her incentive to share-
holder value is reduced to below the optimal level. Such deviations from optimal
incentive contracts might hurt firm performance. In this section we test this
hypothesis by examining how firm performance is related to abnormal CEO
power.
The relation between firm performance and abnormal CEO power is estimated
with the following specification:
Performit ¼ b0þb1CEO Powerit þX8
j¼2
bjControl Variablesijt þ FirmiþYeart þ eit ; ð3Þ
where Performit is measured by either Tobin’s Q or ROA. Tobin’s Q is equal to the
book value of assets plus the market value of common stock minus the sum of
book value of common stock divided by the book value of assets. ROA is return on
assets. Summary statistics on Q and ROA are reported in Table 4.
The control variables include the standard controls used in Q regressions (e.g.
Himmelberg et al., 1999; Kim and Lu, 2011), plus variables measuring CEO charac-
teristics. We control for firm size, measured by the log of sales (LNS);9 growth and
9We do not measure firm size by total assets in performance regressions because the calcula-
tions of both Tobin’s Q and ROA include total assets.
E. H. Kim and Y. Lu
508 � 2011 Korean Securities Association
discretionary investment opportunities, measured by the ratio of capital expendi-
tures to property, plant, and equipment (I ⁄ K); tangibility of assets, measured by the
ratio of property, plant, and equipment to sales (K ⁄ S); firm age, measured as the
log of one plus the number of years from a firm’s initial public offerings as reported
in CRSP or the number of years since its first appearance in CRSP (FirmAge); an
indicator for CEO gender (Female); CEO age (CEOAge); and an indicator for a
CEO chairing the board (CEO_Chair). In addition, we control for firm and year
fixed effects. All regressions are estimated using pooled panel data with robust stan-
dard errors.
5.1. Full Sample Results
Table 7 reports regression estimates for Tobin’s Q and ROA. The results are consis-
tent with those for PPS. Firm performance, whether measured by Tobin’s Q or
ROA, is negatively related to the measure of abnormal CEO power. The coefficients
on Power are negative and significant for both Q and ROA. The coefficients imply
that a one standard deviation increase in the abnormal power is associated with a
0.136% lower Tobin’s Q and a 0.485% lower ROA.
Control variables indicate that Q is negatively correlated with size, firm age, tan-
gibility of assets, and CEO age, but positively correlated with I ⁄ K. All of these
results are consistent with previous studies (e.g., Himmelberg et al., 1999; Kim and
Lu, 2011). ROA is also negatively related to firm age, but, unlike Q, is positively
related to size.
Table 7 Relation between chief executive officer (CEO) power and firm performance
Regression in column (1) estimates the relation CEO power and Tobin’s Q with the full sample over the
entire sample period. Regression in column (2) estimates the relation between CEO power and ROA.
Definitions of all variables are given in Table 1. All regressions control for firm- and year fixed effects.
Robust standard errors are reported in parentheses. Coefficients marked with ***, **, and * are signifi-
cant at 1, 5, and 10%, respectively.
Tobin’s Q
(1)
ROA
(2)
Power )1.286*** (0.302) )4.575** (1.800)
FirmAge )0.521*** (0.076) )2.293*** (0.454)
LNS )0.917*** (0.061) 2.738*** (0.367)
K ⁄ S )0.033*** (0.009) 0.045 (0.053)
I ⁄ K 5.716*** (0.291) 0.792 (1.738)
Female 0.188 (0.328) )1.735 (1.956)
CEOAge )0.592** (0.258) )0.680 (1.541)
CEO_Chair 0.165* (0.087) 0.542 (0.519)
Constant 11.100*** (1.074) )4.432 (6.410)
Firm fixed effects and year fixed effects Y Y
Observations 9855 9855
Adjusted R2 0.39 0.40
Is Chief Executive Officer Power Bad?
� 2011 Korean Securities Association 509
5.2. Interactive Effects with External Monitoring
Because the effects of abnormal CEO power on PPS depend on the strength of EM,
if abnormal power indeed affects firm performance, the performance effect should
also depend on EM. That is, the negative effects on firm performance should be
most pronounced when EM is too weak to deter the negative effects of abnormal
power. (See Durnev and Kim (2005) for interactive effects of internal and external
governance in a cross-country context.) To test this hypothesis, we estimate the
interactive effects of EM and abnormal power on firm performance.
Table 8 reports the estimation results. Columns (1), (2), (4), and (5) contain
separate estimation results for strong and weak EM subsamples for Q or ROA. The
results are remarkably consistent with the PPS results. Abnormal power and firm
performance are unrelated for observations under strong EM. The negative relation
for the full sample is driven by observations under weak EM. This is true regardless
of whether performance is measured by Tobin’s Q (Column 2) or ROA (Column 5).
The estimated coefficients for the low IOC sample indicate that one standard devia-
tion increase in the abnormal CEO power is associated with a 0.342% decrease in
Tobin’s Q and a 1.216% decrease in ROA.
Columns (3) and (6) utilize the full sample with an interaction term between
the abnormal power and IOC. This approach has the benefit of fully utilizing the
information in the IOC variable, but at the cost of forcing the coefficients of
other independent variables to be the same for observations under both strong
and weak EM. The estimation result is robust to this alternative specification. The
interaction term is positive and significant, regardless of whether performance is
measured by Q or ROA, implying that stronger monitoring by institutional inves-
tors negates the negative effects that abnormal CEO power has on firm perfor-
mance.10
6. Robustness
In this section we address endogenous issues. There are alternative explanations for
our results. For one, poorly performing firms might give their CEOs’ more abnor-
mal power to make it easier to implement changes, causing a negative correlation
between the power variable and firm performance. However, this does not explain
why the negative relation is observed only when EM is weak. When EM is strong,
the external monitors are more likely to be aware of the need for stronger leader-
ship to implement changes, predicting a stronger negative correlation, which is con-
tradicted by our evidence of no relation under strong EM.
10We do not use this interaction term approach for the PPS analysis because it would require
several triple interaction terms in a single regression, causing a rather severe multicollinearity
problem. It would also make it difficult to interpret the coefficients.
E. H. Kim and Y. Lu
510 � 2011 Korean Securities Association
Tab
le8
Rel
atio
nb
etw
een
chie
fex
ecu
tive
offi
cer
(CE
O)
po
wer
and
firm
per
form
ance
:h
igh
and
low
inst
itu
tio
nal
ow
ner
ship
con
cen
trat
ion
Th
ista
ble
rep
ort
sre
gres
sio
nes
tim
ates
sep
arat
ely
for
firm
year
ob
serv
atio
ns
wit
hh
igh
and
low
inst
itu
tio
nal
ow
ner
ship
con
cen
trat
ion
(IO
C).
Co
lum
ns
(1)
and
(4)
rep
ort
resu
lts
esti
mat
edfo
rth
eH
IOC
sub
sam
ple
,co
lum
ns
(2)
and
(5)
rep
ort
resu
lts
esti
mat
edfo
rth
eL
IOC
sub
sam
ple
.C
olu
mn
s(3
)an
d(6
)re
po
rtth
ere
sult
ses
ti-
mat
edfo
rth
efu
llsa
mp
lew
ith
the
inte
ract
ion
term
bet
wee
nIO
Can
dC
EO
po
wer
.H
IOC
and
LIO
Car
ed
efin
edas
abo
vean
db
elo
wth
esa
mp
lem
edia
nIO
C.
Th
e
dep
end
ent
vari
able
isT
ob
in’s
Qin
colu
mn
s(1
)–(3
)o
rR
OA
inco
lum
ns
(4)–
(6).
Defi
nit
ion
so
fal
lva
riab
les
are
give
nin
Tab
le1.
All
regr
essi
on
sco
ntr
ol
for
firm
-an
d
year
fixe
def
fect
s.R
ob
ust
stan
dar
der
rors
are
rep
ort
edin
par
enth
eses
.C
oef
fici
ents
mar
ked
wit
h**
*,**
,an
d*
are
sign
ifica
nt
at1,
5,an
d10
%,
resp
ecti
vely
.
To
bin
’sQ
RO
A
HIO
C
(1)
LIO
C
(2)
Fu
ll
(3)
HIO
C
(4)
LIO
C
(5)
Fu
ll
(6)
Pow
er0.
122
(0.2
83)
)3.
222*
**(0
.695
))
6.73
0***
(0.7
80)
)1.
620
(1.6
10)
)11
.467
***
(4.0
57)
)17
.036
***
(4.5
21)
Pow
er·
IOC
20.2
96**
*(2
.725
)50
.308
***
(15.
804)
IOC
)1.
743*
**(0
.380
)0.
367
(2.2
03)
Fir
mA
ge)
0.18
9**
(0.0
81)
)1.
815*
**(0
.235
))
0.66
2***
(0.0
95)
0.52
5(0
.459
))
1.52
7(1
.372
))
1.30
4**
(0.5
50)
LN
S)
0.19
6***
(0.0
68)
)1.
860*
**(0
.147
))
1.05
1***
(0.0
71)
2.80
5***
(0.3
86)
2.21
6***
(0.8
58)
2.85
1***
(0.4
13)
K⁄S
)0.
272*
**(0
.074
))
0.06
2***
(0.0
13)
)0.
041*
**(0
.010
))
3.50
6***
(0.4
21)
0.31
6***
(0.0
76)
0.28
4***
(0.0
55)
I⁄K
4.10
5***
(0.3
35)
6.07
5***
(0.5
27)
5.84
7***
(0.3
12)
9.16
0***
(1.9
08)
)5.
996*
(3.0
78)
0.65
8(1
.806
)
Fem
ale
)0.
121
(0.3
05)
0.59
4(0
.964
)0.
177
(0.3
87)
)0.
498
(1.7
33)
)1.
047
(5.6
27)
)0.
623
(2.2
46)
CE
OA
ge)
0.50
3**
(0.2
50)
)0.
472
(0.5
83)
)0.
584*
*(0
.285
)1.
864
(1.4
21)
)3.
958
(3.4
04)
)0.
506
(1.6
53)
CE
O_
Ch
air
0.12
6(0
.084
))
0.05
4(0
.197
)0.
144
(0.0
95)
1.54
4***
(0.4
78)
)1.
092
(1.1
49)
0.51
9(0
.553
)
Co
nst
ant
5.38
5***
(1.0
59)
21.4
64**
*(2
.404
)12
.758
***
(1.1
86)
)22
.016
***
(6.0
20)
11.4
71(1
4.03
5))
8.76
1(6
.879
)
Fir
mfi
xed
effe
cts
and
year
fixe
def
fect
s
YY
YY
YY
Ob
serv
atio
ns
4757
4221
8978
4757
4221
8978
Ad
just
edR
20.
520.
360.
400.
430.
430.
43
Is Chief Executive Officer Power Bad?
� 2011 Korean Securities Association 511
Another possible scenario is that when a firm performs poorly, more top exec-
utives leave, opening more positions for the remaining CEO to fill, thereby
increasing the abnormal fraction of executives hired during the current CEO’s
tenure.11 This story also does not explain why it does not work for observations
under strong EM. If anything, external monitors under strong EM are more likely
to demand a shakeup in executive suites, leading to more replacement and more
new hiring.
Finally, IOC might not be completely exogenous. It could be affected by the
abnormal CEO power, Q, or ROA. For example, institutional investors might be
more attracted to firms with an appropriate level of CEO power, better performing
firms, or firms undervalued by the market. To address these potential issues, we
estimate three-stage least squares simultaneous regressions as follows:
Performit ¼ a0 þ a1Powerit þ a2Powerit � IOCit þ a3IOCit þ a4FirmAgeit þ a5LNSit
þ a6K=Sit þ a7I=Kit þ a8Femaleit þ a9CEOAgeit þ a10CEO Chairit
þ gt þ hi þ eit ; ð4Þ
Powerit ¼ b0 þ b1IOCit þ b2Performit þ b3LNSit þ b4FirmAgeit þ b5Femaleit
þ b6CEOAgeit þ b7CEO Chairit þ gt þ hi þ eit ; and ð5Þ
IOCit ¼ c0 þ c1Powerit þ c2Performit þ c3FirmAgeit þ c4LNSit
þ c5K=Sit þ c6I=Kit þ c7IOC IYð�iÞit þ hi þ eit :ð6Þ
In this system of equations, the performance variables, the power variable, and
IOC are all treated as endogenous. We estimate the system of equations with the
full sample using the interaction terms of the power variable and IOC to estimate
the interactive effects of the power and the strength of EM. In the performance
equation, we include the power variable, IOC, and all control variables measuring
firm characteristics and CEO characteristics used in equation (3). In the CEO power
equation, independent variables include IOC, the performance variable, FirmAgeit,
Femaleit, CEOAgeit, and CEO_Chairit, but not K ⁄ S and I ⁄ K, because the abnormal
CEO power might be related to CEO characteristics and some basic firm character-
istics, such as firm size and firm age, but might be unrelated to the tangibility of
assets or growth and discretionary investment opportunities. In the IOC equation,
independent variables include CEO power, a performance variable, FirmAgeit, K ⁄ Sit,
and I ⁄ Kit, but not Femaleit, CEOAgeit, and CEO_Chair. Although IOC might be
affected by firm characteristics, it is unlikely to be affected by CEO characteristics.
Additionally, we include a predictor of IOC in the IOC equation: IOC of peer firms,
11This possibility was raised by Landier et al. (2008) but was rejected as an explanation for a
negative relation between the abnormal power and Q.
E. H. Kim and Y. Lu
512 � 2011 Korean Securities Association
Tab
le9
Th
ree-
stag
ele
ast
squ
ares
esti
mat
ion
of
firm
per
form
ance
,ch
ief
exec
uti
veo
ffice
r(C
EO
)p
ow
er,
and
inst
itu
tio
nal
ow
ner
ship
con
cen
trat
ion
Reg
ress
ion
sin
colu
mn
s(1
)–(3
)es
tim
ate
the
equ
atio
ns
for
To
bin
’sQ
,C
EO
po
wer
,an
dIO
C.
Reg
ress
ion
sin
colu
mn
s(4
)–(6
)es
tim
ate
the
equ
atio
ns
for
RO
A,
CE
O
po
wer
,an
dIO
C.
Defi
nit
ion
so
fal
lva
riab
les
are
give
nin
Tab
le1.
All
regr
essi
on
sco
ntr
ol
for
ind
ust
ry-
and
year
fixe
def
fect
s.In
du
stri
esar
ed
efin
edas
inF
ama
and
Fre
nch
(199
7).
Ro
bu
stst
and
ard
erro
rsar
ere
po
rted
inp
aren
thes
es.
Co
effi
cien
tsm
arke
dw
ith
***,
**,
and
*ar
esi
gnifi
can
tat
1,5,
and
10%
,re
spec
tive
ly.
3SL
So
fT
ob
in’s
Qre
gres
sio
ns
3SL
So
fR
OA
regr
essi
on
s
To
bin
’sQ
(1)
Po
wer
(2)
IOC
(3)
RO
A
(4)
Po
wer
(5)
IOC
(6)
Pow
er)
4.07
9***
(0.7
18)
)0.
0112
(0.0
0937
))
6.87
8(4
.586
))
0.00
379
(0.0
0941
)
Pow
er·
IOC
11.5
2***
(2.5
64)
19.6
1(1
6.38
)
IOC
)5.
802*
**(0
.282
))
0.01
77(0
.012
1)19
.59*
**(1
.799
))
0.00
764
(0.0
121)
Tob
in’s
Q)
0.00
177*
**(0
.000
441)
)0.
0082
0***
(0.0
0039
4)
RO
A)
8.47
e-05
(7.1
2e-0
5)0.
0006
72**
*(6
.30e
-05)
Fir
mA
ge)
0.23
6***
(0.0
369)
0.00
173
(0.0
0155
))
0.01
03**
*(0
.001
37)
)0.
0777
(0.2
35)
0.00
242
(0.0
0154
))
0.00
854*
**(0
.001
38)
LN
S)
0.10
4***
(0.0
205)
)0.
0015
5*(0
.000
865)
)0.
0073
9***
(0.0
0075
9)2.
065*
**(0
.130
))
0.00
115
(0.0
0087
9))
0.00
813*
**(0
.000
772)
K⁄S
)0.
0038
6(0
.008
77)
)0.
0011
1***
(0.0
0032
9))
0.64
3***
(0.0
558)
)0.
0006
68**
(0.0
0033
3)
I⁄K
6.81
4***
(0.2
90)
0.00
325
(0.0
112)
3.88
4**
(1.8
43)
)0.
0575
***
(0.0
109)
Fem
ale
1.07
6***
(0.2
41)
0.00
598
(0.0
104)
3.82
9**
(1.5
42)
0.00
438
(0.0
104)
CE
OA
ge)
0.75
2***
(0.2
06)
)0.
0391
***
(0.0
0882
)5.
237*
**(1
.314
))
0.03
68**
*(0
.008
82)
CE
O_
Ch
air
0.13
6**
(0.0
590)
)0.
0002
29(0
.002
54)
)0.
278
(0.3
77)
)0.
0005
09(0
.002
54)
IOC
_IY
()i)
)0.
333*
**(0
.060
6))
0.33
1***
(0.0
612)
Co
nst
ant
0.15
8***
(0.0
536)
)33
.47*
**(8
.085
)0.
0600
(0.0
541)
Ob
serv
atio
ns
8774
8774
8774
8774
8774
8774
Ind
ust
ryfi
xed
effe
cts
and
fixe
def
fect
s
YY
YY
YY
v296
61.9
419
1.65
6929
2.46
997.
2617
6.61
6833
6.96
R2
0.18
90.
020
0.16
10.
091
0.02
00.
160
Is Chief Executive Officer Power Bad?
� 2011 Korean Securities Association 513
IOC_IY()i)it, the mean value of IOC of all firms in the same industry in the same
year, excluding the firm itself. A firm’s IOC might be affected by how much institu-
tional investors are attracted to other firms in the same industry during the same
year. All regressions in the system include firm and year fixed effects.
Table 9 presents the estimation results. They show no relation between power
and IOC; however, the results do not rule out the possibility that firm performance
affects the power measure or IOC. Nevertheless, column (1) demonstrates the
robustness of our conclusions regarding the relation between Q and power: CEO
power has a negative effect on firm value and the negative effect is deterred by high
IOC. The same cannot be said about the relation between ROA and power. ROA is
based on accounting numbers; as such, a CEO with powerful influence over his or
her top executives is in a stronger position to smooth out fluctuations in ROA. We
suspect the weak relation between power and ROA is partly due to earnings man-
agement. Because Q is less subject to such ‘‘management’’ and is a more direct mar-
ket-based measure of shareholder value, we conclude that abnormal CEO power is
bad for shareholder value when external monitoring is not strong enough to
restrain it.
7. Conclusions
This paper focuses on CEO structural power, as measured by an abnormal frac-
tion of top executive hired by the current CEO. The abnormal CEO power seems
to decrease CEOs’ pay for performance sensitivity and to reduce firm perfor-
mance. These results not only suggest that CEOs’ abnormal power is bad for firm
performance but also illustrate that using abnormal power to capture the com-
pensation process is a channel through which CEO power affects firm perfor-
mance. These results are robust to firm fixed characteristics and potential reverse
causality.
The negative effects of CEO power on CEO PPS and firm performance are
observed only when monitoring by institutional investors is weak. CEO power
seems benign when they are subject to close monitoring by institutional investors.
Proper monitoring and guidance by institutional investors appear effective in pre-
venting CEOs from using their power to capture the compensation process and
hurting firm performance. Although unchecked CEO power is dangerous, it can be
benign when sufficient checks and balances against potential abuse of the power
exist.
References
Adams, R. B., H. Almeida, and D. Ferreira, 2005, Powerful CEOs and their impact on
corporate performance, Review of Financial Studies 18, pp. 1403–1432.
Aggarwal, R. K., and A. A. Samwick, 1999, The other side of the trade-off: The impact of
risk on executive compensation, Journal of Political Economy 107, pp. 65–105.
E. H. Kim and Y. Lu
514 � 2011 Korean Securities Association
Baker, G. P., and B. J. Hall, 2004, CEO incentives and firm size, Journal of Labor Economics
22, pp. 767–798.
Bebchuk, L., and J. Fried, 2004, Pay without performance: The unfulfilled promise of
executive compensation (Harvard University Press, Cambridge, MA.).
Bebchuk, L. A., K. J. M. Cremers, and U. C. Peyer, 2011, The CEO pay slice. Journal of
Financial Economics (forthcoming).
Bennedsen, M., F. Perez-Gonzalez, and D. Wolfenzon, 2006, Do CEOs matter? Working
Paper, Columbia University, New York.
Bertrand, M., and S. Mullainathan, 2000, Agents with and without principals, American
Economics Review 90, pp. 203–208.
Bertrand, M., and S. Mullainathan, 2001, Are CEOs rewarded for luck? The ones without
principals are, Quarterly Journal of Economics 116, pp. 901–932.
Bertrand, M., and A. Schoar, 2003, Managing with style: The effect of managers on firm
policies, Quarterly Journal of Economics 118, pp. 1169–1208.
Cremers, K. J. M., and V. B. Nair, 2005, Governance mechanisms and equity prices, Journal
of Finance 60, pp. 2859–2894.
Cronqvist, H., A. K. Makhija, and S. E. Yonker, 2009, What does CEOs’ personal leverage
tell us about corporate leverage? Working Paper Series 2009-4, Ohio State University,
Charles A. Dice Center for Research in Financial Economics, Columbus, OH.
Crystal, G., 1991, In search of excess: The overcompensation of American executives
(W.W. Norton, New York).
Del Guercio, D., L. Seery, and T. Woidtke, 2008, Do boards pay attention when institutional
investor activists ‘‘just vote no’’? Journal of Financial Economics 90, pp. 84–103.
Durnev, A., and E. H. Kim, 2005, To steal or not to steal: Firm attributes, legal environment,
and valuation, Journal of Finance 60, pp. 1461–1495.
Edmans, A., 2009. Blockholder trading, market efficiency, and managerial myopia, Journal of
Finance 64, pp. 2481–2514.
Fama, E., and K. French, 1997, Industry costs of equity, Journal of Financial Economics 43,
pp. 153–193.
Finkelstein, S., 1992, Power in top management teams: Dimensions, measurement, and
validation, Academy of Management Journal 35, pp. 505–538.
Gibbons, R., and K. Murphy, 1992, Optimal incentive contracts in the presence of career
concerns: Theory and evidence, Journal of Political Economy 100, pp. 468–505.
Hall, B. J., and J. B. Liebman, 1998, Are CEOs really paid like bureaucrats? Quarterly Journal
of Economics 111, pp. 653–691.
Hartzell, J., and L. T. Starks, 2003, Institutional investors and executive compensation,
Journal of Finance 58, pp. 2351–2374.
Himmelberg, C. P., R. G. Hubbard, and P. Darius, 1999, Understanding the determinants of
managerial ownership and the link between ownership and performance, Journal of
Financial Economics 53, pp. 353–384.
Jenter, D., and K. Lewellen, 2011, CEO preferences and acquisitions, Working Paper,
Stanford University, Palo Alto, CA.
Kim, E. H., and Y. Lu, 2011, CEO ownership, external governance, and risk-taking, Journal of
Financial Economics (forthcoming).
Koenker, R., and K. F. Hallock, 2001, Quantile regression, Journal of Economic Perspectives
15, pp. 143–156.
Is Chief Executive Officer Power Bad?
� 2011 Korean Securities Association 515
Lafontaine, F., and S. Bhattacharyya, 1995, The role of risk in franchising, Journal of
Corporate Finance 2, pp. 39–74.
Landier, A., D. Sraer, and D. Thesmar, 2008, Bottom-up corporate governance, NYU
Working Paper No. FIN-05-011, New York, NY. Available at SSRN: http://ssrn.com/
abstract=1294147
Milbourn, T. T., 2003, CEO reputation and stock-based compensation, Journal of Financial
Economics 68, pp. 233–263.
Morse, A., V. Nanda, and A. Seru, 2010, Are incentive contracts rigged by powerful CEOs?
Journal of Finance (forthcoming).
Murphy, K. J., 1999, Executive compensation, In O. Ashenfelter, D. Card eds: Handbook of
labor economics, Chapter 38 (Elsevier Science B. V., Amsterdam), pp. 2485–2563.
Nguyen, B. D., and K. M. Nielsen, 2010, What death can tell: Are executives paid for their
contributions to firm value? Working Paper, Chinese University of Hong Kong,
Hong Kong.
Prendergast, C., 2002, The tenuous trade-off between risk and incentives, Journal of Political
Economy 110, pp. 1071–1102.
Rajgopal, S., T. Shevlin, and V. Zamora, 2006, CEOs’ outside employment opportunities and
the lack of relative performance evaluation in compensation contracts, Journal of Finance
61, pp. 1813–1844.
Schaefer, S., 1998, The dependence of pay-performance sensitivity on the size of the firm,
Review of Economics and Statistics 80, pp. 436–443.
Shleifer, A., and R. Vishny, 1986, Large shareholders and corporate control, Journal
of Political Economy 94, pp. 461–488.
E. H. Kim and Y. Lu
516 � 2011 Korean Securities Association