Upload
others
View
5
Download
0
Embed Size (px)
Citation preview
RELATIVE PAY FOR RELATIVE PERFORMANCE
Ivan E. Bricka, Darius Paliaa, and Chia-Jane Wangb
This version: August 12, 2013
Abstract
Using a large panel data set, this paper examines industry effects in CEO pay levels
and pay-performance sensitivities. We find the following results. First, we find
significant industry effects in both CEO pay levels and pay-performance sensitivities.
Second, we find strong evidence that CEO compensation relative to other firms in the
industry is positively correlated to that firm’s relative performance within that industry.
Third, we find a stronger relative pay for relative performance relationship in industries
where more industry-specific talent is required. Fourth, our results are robust to both
OLS and dynamic GMM (which controls for dynamic endogeneity of the relationships)
and to using two alternative definitions of the pay-performance sensitivities. These results
suggest that managers are both compensated and evaluated on their talent relative to other
firms in their industry.
______________________________________________________________________________________ aRutgers Business School, and
bManhattan College, respectively. We thank Patrick Bolton, Martijn
Cremers, Valentin Dimitrov, Sirdar Dinc, Mike Fishman, Yuna Heo, Simi Kedia, Oded Palmon, Maya
Vaisman, Kam-Ming Wan and seminar participants at Rutgers, Villanova, the 2012 Triple Crown
Conference and the 2013 City University of Hong Kong International Conference on Corporate Finance
and Financial Markets for helpful comments. Brick (Palia) thanks the David Whitcomb Center for
Financial Services (the Thomas A. Renyi Chair) for partial financial support. All errors remain our
responsibility. Corresponding author: Darius Palia; [email protected]
2
1. Introduction
This paper examines whether CEO’s are paid and evaluated relative to their
industry peers. For example, CEOs of a bank would be paid relative to other banks,
whereas CEOs of a manufacturing company would be paid relative to other
manufacturing companies. They would be a corresponding relative performance
evaluation between that bank and other banks, and a manufacturing company and other
manufacturing companies. We call this hypothesis the relative pay for relative
performance hypothesis.
For the relative pay for relative performance hypothesis to hold as an optimal
equilibrium outcome, one needs two assumptions to hold in the standard principal-agent
model control (Holmstrom, 1979, Holmstrom and Milgrom, 1987, 1992, and Jewitt,
1988). In this model, risk neutral shareholders (or on their behalf, the board of directors),
design a compensation structure based on the manager’s incentive compatible and
participation constraints. The first assumption is that there are varying levels of skill that
affects the reservation utility of managers in the participation constraint. Such an
argument suggests that in order to attract or retain talented CEOs, shareholders have to
offer higher pay (see, for example, Oyer (2004), Murphy and Zabojnik (2007), Gabaix
and Landier (2008), Edmans and Gabaix (2009), Edmans, Gabaix, and Landier (2009))1.
Extending this argument to the type of talent, it seems reasonable that different industries
employ CEOs who have both industry-specific human capital and general management
skills. For example, the CEO of a bank must have general management skills reflective
3
of a CEO of any firm. In addition, the bank CEO must have specific skills such as
managing the loan portfolio and credit risk that are particularly unique to the banking
industry. In order to attract the appropriate managerial talent, firms have to offer these
executives wages that are commensurate with their ability. If the level of managerial
skills varies by industry, we would expect to find an industry effect in the determination
of CEO compensation. Therefore, boards might optimally design a relative compensation
contract for the CEO.
The second assumption that is that the optimal sharing rule (i.e., the manager’s
pay-performance sensitivity) should reduce the noise in performance measures consistent
with enhanced signal extraction. That is, the manager should be paid based upon the
environment s/he can control. Hence, under these models, industry performance which is
not under the control of a specific CEO should be removed from her performance
evaluation. Therefore boards optimally design a relative performance contract for the
CEO.
Note that we are not arguing that some industries have more talented CEOs than
others. Instead we suggest that some industries have more industry-specific talent which
makes it costly for outsiders to enter and/or harder for insiders to leave. We would also
expect that the relative pay for relative performance sensitivity would depend on the
informativeness of the relative performance measures. Accordingly, the more difficult it
is for the board to identify the source of firm success is due to the CEO’s skill or luck, the
less sensitive is the CEO’s relative pay to relative performance. Parrino (1997) argues
1 See Section 2 for a detailed explanation of these and other related papers.
4
that industries that tend to hire from within are easier to monitor and better able to filter
the effects of industry shocks. He finds that firms are more likely to fire CEOs for poor
relative performance in industries that tend to hire CEOs from within. This suggests the
importance of industry-specific talent. That is, the more important the industry-specific
knowledge for a CEO to be successful, the more difficult it is for boards of directors to
hire from outside (Parrino 1997, Murphy and Zabojnik 2007). This suggests that we
should examine whether industries with high insider hiring have a higher sensitivity of
relative pay to relative performance compared to industries that hire CEOs from outside.
Using a large panel data set of US CEO compensation data for the years 1994 to
2008, we examine both CEO pay levels and pay-performance sensitivities and test for
empirical support of the relative pay for relative performance hypothesis. We find the
following results. First, we find significant industry effects in both CEO pay levels and
pay-performance sensitivities.2 Second, we find strong evidence that CEO compensation
relative to other firms in the industry is positively correlated to that firm’s relative
performance within that industry, namely, support for the relative pay for relative
performance hypothesis. This result holds when we use both OLS with standard errors
clustering at the firm level and the dynamic GMM (Arellano and Bover, 1995) method
which controls for unobservable heterogeneity and the dynamic endogeniety of the
2 Evidence for significant industry effects has been found in the capital structure, dividend, and labor wage
literature. Evidence for significant industry effect in capital structure has been found by Bradley, Jarrell and
Kim (1984), Smith and Watts (1992), MacKay and Phillips (2005), Frank and Goyal (2007) and Leary and
Roberts (2010). Lintner (1953), Michel (1979), McCabe (1979), and Dempsey, Laber and Rozeff (1993)
find significant industry effects in dividend policies. See Section 2 for papers on industry wages.
5
relationships.3 Third, we find a stronger relative pay for relative performance relationship
in industries wherein more industry-specific talent is required. Fourth, consistent with
the prior literature, we find no evidence in support for the relative performance evaluation
(RPE) hypothesis that links absolute pay with relative performance. Finally, the above
results are robust to using two measure of the pay-performance relationship, namely, pay-
performance sensitivity (defined as the increase in CEO wealth for a one-percent increase
in shareholder wealth (Core and Guay 2002)), or the scaled wealth-performance
sensitivity (defined as the dollar change in CEO wealth for a 100-percent change in firm
value divided by annual compensation (Edmans, Gabaix, and Landier 2009)). The latter
definition is independent of firm-size and is derived from a model where effort has a
multiplicative effect on firm value and utility.
Our paper is related to two strands of the literature. First, is the literature that
examines the absolute change in CEO compensation to relative performance, which the
literature denotes as relative performance evaluation (RPE). The empirical papers testing
relative performance evaluation implicitly ignore any industry human capital required in
the participation constraint. We are breaking this assumption in the participation
constraint, by allowing the CEO’s reservation utility to be industry specific. Many
empirical papers4 have found little support for RPE. In order to ensure that there is
3 See Section 3.1 for further details.
4 Antle and Smith (1986), Gibbons and Murphy (1990), Jensen and Murphy (1990), Barro and Barro
(1990), Janakiraman, Lambert and Larker (1992), Aggarwal and Samwick (1999a,b), Joh (1999) and
Murphy (1999) fail to find support for the relative performance evaluation. Albuquerque (1999) and
Lewellen (2013) find support for the relative performance evaluation. However, Lewellen (2013) does not
(next page)
6
nothing strange about our sample, we also performed regressions analogous to these
papers that test the relative performance evaluation hypothesis. We also find no evidence
for the RPE hypothesis. We differ from this literature by focusing on the relative change
in pay, rather than the absolute change in pay.
Our paper is also related to the literature that examines benchmarking in CEO
compensation. On the one hand, Bizjak, Lemmon and Naveen (2008), Bizjack, Lemmon
and Nguyen (2011)5
and Albuquerque, De Franco and Verdi (2012) find that
benchmarking is undertaken for retention of talented human capital. On the other hand,
Faulkender and Yang (2010, 2012) find that benchmarking is undertaken for
opportunistic reasons and increased disclosure has not reduced this behavior. These
papers examine whether CEO pay in one firm is related to the CEO pay of its peers,
which they call relative pay. But none of these papers examine whether relative pay is
correlated with the firm’s relative performance. That said, our results are generally
consistent with those in Bizjak, Lemmon and Naveen (2008), Bizjack, Lemmon and
Nguyen (2011), and Albuquerque, De Franco and Verdi (2012).
Some paper have suggested that CEO pay is not optimally set and is the outcome
of powerful and entrenched CEOs and their crony boards (see, for example, Bertrand and
Mullainathan, 2001, Bebchuk and Freid, 2004, Brick, Palmon and Wald 2006, and
Morse, Nanda, and Seru 2011). We do not suggest that these effects are not present. We
find relative performance evaluation in the sample period covered by Albuquerque (1999), but finds
relative performance evaluation when industries are defined by their primary product market competitors. 5 Bizjack, Lemmon and Nguyen (2011) find an upward bias in CEO pay which is reduced with disclosure.
7
only suggest that some part of CEO pay is based as an optimal equilibrium outcome
under the relative pay for relative performance hypothesis.
This paper proceeds as follows. Section 2 explains the related literature whereas
Section 3 describes our data, variables, and empirical methodology. Our empirical
results are reported in Section 4, and conclusions presented in Section 5.
2. Related literature
3.1 Literature for CEO talent
The standard principal-agent model of Holmstrom (1979), Holmstrom and
Milgrom (1987, 1992) did not have a role for CEO ability or talent. The growing
literature that does include CEO talent has its roots in seminal papers by Lucas (1978)
and Rosen (1981). In these papers, compensation for the most talented managers is high
as they manage large firms where the manager’s marginal product is high. In a
competitive assignment model, Gabaix and Landier (2008) and Tervio (2008) suggest
that more talented CEOs are matched to firms with larger market values and therefore
have to be compensated with a higher level of pay. In these models there is no moral
hazard and therefore no incentive effect of CEO pay.
In a moral hazard framework, Himmelberg and Hubbard (2000) find that CEOs of
large firms (namely, more talented CEOs) have a positive relationship with aggregate
stock market returns. This is because the supply of these CEOs is relatively inelastic, and
aggregate productivity shocks will simultaneously raise firm value and the marginal
value of CEO services.
8
Inserting moral hazard into a talent assignment model, Edmans, Gabaix, and
Landier (2009)) derive CEO pay levels and pay-performance sensitivity wherein effort
has a multiplicative effect on firm value and utility. With this multiplicative production
function, effort has a multiplicative effect on firm value resulting in their scaled pay-
performance sensitivity measure (defined as the dollar change in CEO wealth for a 100-
percent change in firm value divided by annual compensation). Such talented CEOs have
a higher level of CEO pay. Endogenizing for firm size, Baranchuk, MacDonald, and
Yang (2011) employ a ‘superstars’ model wherein each firm chooses its size, the
manager and her compensation. In equilibrium, they suggest that higher-ability CEOs
will be matched to larger firms with higher pay levels and pay-performance sensitivities.
Bertrand and Mullainathan (2001) find that CEO pay varies with shocks that are
not under the CEO’s control, a phenomenon they call ‘pay for luck’. However, Oyer
(2004), Falato, Li, and Milbourn (2009), and Eisfeldt, and Kuhnen (2012) show that
industry or market shocks can be correlated with the value of the outside option to CEO
talent. In order to retain talented CEOs within the firm, one would have to offer a higher
pay.
Murphy and Zabojnik (2003, 2007) find that CEO salary and bonus is higher
when CEOs are hired from the outside and for industries where outside hiring is more
prevalent. They attribute this finding to the trend of firms valuing general skills more
than firm-specific skills. Frydman (2007) finds that executives with more general skills
are more likely to have a higher level of pay in the 1990s than in the 1960s than
executives with firm-specific skills. Using detailed biographical data, Custodio, Ferreira,
9
and Matos (2011) find that CEOs with generalist skills (defined as the first principal
component of number of industries she worked for, was a prior CEO before, and had a
general business degree) earned higher pay levels and incentive pay than CEO who have
specialist skills. Cremers and Grinstein (2011) find that benchmarking is more prevalent
and CEO pay is more related to industry performance when successor CEOs come from
outside their industry. They find no evidence for a weaker relationship between
compensation and firm size in industries with more firm-specific skills, a finding contrary
to firm size proxying for talent (as in Gabaix and Landier 2008). Note that these papers
examine either the level of pay or change in pay but not how pay varies within the
industry and how it changes with relative performance.
2.2 Literature on industry compensation
There is a large literature in labor economics that has shown that different
industries pay their employees differently. Specifically, they regress the logarithm of the
wage rate for each employee on individual characteristics such as age, education,
occupation, gender, race, union membership, and so on. When they include industry
dummies, they find them to be strongly statistically significant. These inter-industry wage
differentials were first found by Slichter (1950) during 1923 and 1946. More recent
evidence for inter-industry wage differentials has been found by Krueger and Summers
(1987, 1988), Murphy and Toppel (1987), Thaler (1989), Gibbons and Katz (1992), and
Goux and Maurin (1999). Interestingly, Krueger and Summers (1987) find that the high
10
wage industries in the earlier period correlate highly with the high wage industries in the
1980s.
Some studies have examined whether CEOs with different skills and talent are
compensated differently. Due to political constraints on CEO pay, Joskow, Rose and
Shepard (1993) and Joskow, Rose and Wolfram (1996) find that regulated utilities have
low pay-performance sensitivities and levels. Palia (2000) finds that regulated utilities
attract CEOs with a lower-quality education (proxied by the ranking of educational
institution s/he graduated from) and are offered lower pay levels and sensitivities than
CEOs of manufacturing firms.
3. Methodology, data and variables
3.1 Methodology
In this section we describe the methodology employed to ascertain the
relationship between pay and performance. We begin by testing the relative pay for
relative performance hypothesis using the following empirical specification for CEOs:
(firm pay – industry pay)it = α + β(firm perf.– industry perf.)it + ΩXit + µt + εit (1)
where Xit are a comprehensive set of control variables and µt are year-dummies to control
for general macroeconomic factors. We regress the difference between CEO pay and the
average industry pay against the difference between the firm’s performance and industry
performance. We estimate (1) above using two measures of CEO pay, namely, CEO pay
11
levels and CEO pay-performance sensitivities. If one has to find support for the relative
for relative pay for relative performance to hold in (1) above, one should find a positive
and statistically significant relationship on β.
In contrast, the relative performance hypothesis (RPE) examines the absolute
change in CEO compensation to relative performance. More specifically,
∆(firm pay) it = α + β(firm perf.)it + (industry perf.)it + ΩXit + µt + εit (2)
where Xit are the same set of control variables as in (1) and µt the year-dummies. If one
has to find support for the relative performance hypothesis in (2) above, one should find a
negative and statistically significant relationship for .
Equation (1) is first estimated using OLS. The standard errors are corrected for
clustering at the firm-level so as to take into account that the observations within firms
maybe non-independent.
Given that we have panel data, we use panel data estimation techniques to avoid
any unobserved omitted variable problem. Since we cannot be sure whether the
unobserved effects are correlated with the explanatory variables, firm-level fixed effects
estimation is chosen over random effects estimation.6 But it is possible that certain
unobservable firm and/or CEO characteristics are omitted and cause a spurious
correlation between pay and performance. The fixed effects methodology has the
6A Hausman and Taylor (1981) specification test shows that the fixed effects estimator is preferred over the
random effects estimator in our sample.
12
advantage that it captures any unobservable firm-level heterogeneity that was shown to
be important in Murphy (1985) and Himmelberg, Hubbard, and Palia (1999).7
But it is possible that the relationship between CEO pay and firm performance
suffers from a simultaneity or endogeneity bias (Roberts and Whited forthcoming). For
example, the CEO’s risk aversion parameter should enter her tolerance for risk and would
adversely affect the amount of incentive pay she could bear. The normal way to solve
such omitted variable or endogeniety concerns is to find an instrumental variable. But in
this context, all potential instrumental variables are unable to satisfy the exclusion
restriction for instrumental variable identification as they are likely to be related to both
firm performance and CEO pay.
The simultaneity bias affecting the regression coefficients could be due to static
endogeneity (Palia 2001) and/or dynamic endogeneity (Wintoki, Linck, and Netter 2012).
In static endogeneity, we do not know whether CEO pay in this fiscal year causes firm
performance to rise in the same fiscal year, or vice versa. In dynamic endogeneity, past
values of CEO pay might have caused current firm performance to rise, or vice versa.
One way to check the robustness of our OLS results is to use lagged values of CEO pay
using the dynamic GMM method of Arellano and Bover (1995).8
Arellano and Bover (1995) first check which lags are uncorrelated with the first-
differenced residuals under the null hypothesis of no serial correlation. Consistent with
7Murphy (1985) argues that the firm-level fixed effects model is the optimal estimation methodology.
“Absent a theory indicating the relevant variables, and data on these variables, these cross-sectional models
are inherently subject to a serious omitted variables problem.” (page 12) 8 The dynamic GMM method has also been used by Wintoki, Linck, and Netter (2012) to examine the
relationship between board structure and firm performance.
13
their paper, for each regression specification we conduct an autoregression test and find
that three-year or four-year lags are what we need as valid instrument variables. Further,
the Arellano and Bover method uses these instrumental variables in both levels and
differences and uses GMM to estimate the pay-performance relationship.
3.2 Data and variables used
We use a large panel data set of US CEO compensation data obtained from
ExecuComp for the years 1994 to 2008. We look at both CEO compensation levels and
CEO pay-performance sensitivities. We use data from CRSP to obtain individual stock
returns and accounting data from Standard and Poor’s Compustat. Table 1 defines each
of the variables used in our study.
*** Table 1***
We use ExecuComp’s tdc1 variable as our proxy for CEO pay levels; defined as
the sum of salary, bonus, value of restricted stock granted, total value of stock options
granted (using Black-Scholes), and long-term incentive payouts. To control for skewness,
we take the natural logarithm of tdc1 as our variable of interest. We denote this variable
as ltdc1.
We use two measures to examine the CEO’s pay-performance relationship. The
first measure follows the procedure by Core and Guay (2002) to calculate pay-for-
performance sensitivity, which we denote as totpps. It is defined as the change in the
CEO’s wealth for a one-percent change in equity value. Consistent with the prior
literature, we use this measure of totpps because it is the change in total wealth invested
14
in the firm that incentivizes the CEO. The procedure we use to estimate totpps sums the
pay-for-performance measures of current option grants, options outstanding that were
granted prior to the current fiscal year, and shares of stock that are held by the CEO. We
exclude any years in which the CEO only served for part of the fiscal year as we want to
match firm performance to the particular compensation package of one individual.
To calculate the performance sensitivity of current option granted, let numsecur
denote the number of newly options granted to the CEO with a given maturity m and
exercise price X. Then the performance sensitivity of options granted is given by pps =
e-dm
(N(D1))0.01*P* numsecur,where d is the dividend yield of the stock, P is the price of
the stock at the end of the fiscal year, and N(.) is the standard normal cumulative density
function. D1 is given by D1 = {log(P/X) + m[(r-d) + (2/2)]}/m0.5
where r is the yield
to maturity of the seven-year treasury security at the fiscal year end and is the Black-
Scholes volatility measure of the underlying stock as given by ExecuComp. In a given
fiscal year the firm may give the CEO several grants of options with different m and X.
We evaluate each grant and sum the values to obtain the performance sensitivity for all
newly granted options in that fiscal year.
We also calculate the performance sensitivity of previously granted options. We
assume the maturity of previously granted options to be three years less than the average
maturity of the current option grants if these options are immediately exercisable. We
further assume that the maturity of previously granted options that are not immediately
exercisable is one year less than the average maturity of the current option granted. In
both cases, we use the yield to maturity of a five-year treasury security prevailing at the
15
fiscal year end. We define the exercise price for the previously granted options that are
exercisable to be equal to the current stock price less the ratio of the intrinsic value of
unexercised but exercisable previously granted options (inmonex) to the number of
previously granted unexercised but exercisable options (uexnumex). We define the
exercise price for the previously granted options that are not exercisable to be equal to the
current stock price less the ratio of the intrinsic value of unexercisable granted options
(inmonun) less the intrinsic value of the current options granted to the number of
previously granted unexercised and unexercisable options (uexnumun - soptgrnt).
Finally, by definition the pps for stock held by the CEO equals one-percent of the value
of equity held by the CEO. To account for possible non-linearities we take the logarithm
of totpps as our variable of interest. We denote this variable as lpps.
The second measure is the scaled wealth-performance sensitivity of Edmans,
Gabaix, and Landier (2009). These authors derive an optimal sharing rule in a principal-
agent model with talent assignment by considering multiplicative specifications for the
CEO’s utility and production function. The optimal incentive structure is derived as the
dollar change in CEO wealth for a 100-percent change in firm value divided by annual
compensation. Two advantages of this definition are that it is independent of firm-size
and is more effective at solving agency problems that have multiplicative impacts on firm
value. We obtain this measure from http://finance.wharton.upenn.edu/~aedmans/
data.html.
To examine if firms generally reward CEO for relative performance, we need to
calculate the relative pay of the CEO against the relative performance of the firm. The
16
relative pay in year t is defined as the difference between firm i’s CEO compensation
variable (tdc1, ltdc1, pps and lpps) of fiscal year t and the equal-weighted average of all
firms in the industry for that fiscal year. We denote relative pay as dtdc1, dltdc1, dpps
and dlpps, respectively. The relative performance of firm j of industry i is defined as the
difference between firm i’s performance variable (tobinq, roa, and crawret12) and the
equal-weighted average industry performance (mtobinq, indret, and mroa). The equal
industry returns are obtained from Kenneth French’s website http://mba.tuck.dartmouth
.edu/ pages/faculty/ken.french/data_library.html). In order to ensure that we do not have
any spurious correlation between the dependent variable of relative pay and independent
variable of relative performance, we adjust the average of industry performance by
excluding the own firm’s contribution to the industry average. We denote the relative
performance as dtobinq, droa, and dcrawret12.
We use three performance measures of the firm. The first proxy we use is
Tobin’s Q (tobinq) and is the one generally used by a large part of the previous literature.
As in Smith and Watts (1992), Shin and Stulz (1998), and Palia (2001), Tobin’s Q is
calculated as the market value of equity (price of common shares times the number of
shares outstanding), minus the book value of equity, plus the book value of assets, over
book value of assets. The second proxy is the firm’s return-on-assets (roa) and is
calculated as earnings before interest, taxes, and depreciation over book value of assets.
The performance proxy roa has also been used by Gompers, Ishii, Metrick (2003) and
Core, Guay, and Rusticus (2006) as alternative measures of firm performance. The third
proxy is the holding period return of firm’s equity for the fiscal year, denoted as
17
crawret12, used by Antle and Smith (1986), Murphy (1999) and Gong, Li and Shin
(2011).
We include several control variables in our regressions. We use R&D expenses
scaled by total assets (Berger, Ofek, Yermack 1997, Titman and Wessels 1988, Jung,
Kim, and Stulz 1996) to proxy for firm growth. We denote this variable by rd. Given
that the transparency of managerial actions will decline in growth opportunities, thereby
reducing the effectiveness of monitoring by board and large blockholders, the optimal
pay and sensitivity should be higher to encourage taking valuable risk-increasing projects
(Guay 1999). We set rd equal to zero .when Compustat records a missing value. We also
use Tobin’s Q to proxy for the present value of growth opportunities. We expect
compensation may vary with firm size. For this reason, size is included to avoid the
possibility that our explanatory variables might proxy for the size. Size is measure by the
log of sales, denoted by lsales. Since there may be a non-linear relationship between size
and compensation, we include the square term of lsales, which we denote as lsalessq.
We further expect that pay is related to leverage as shown by John and John (1993).
Consequently, we include leverage, denoted as lev as a control variable, and lev is
defined as total debt divided by total assets. Dividend payments may be used to reduce
manager’s control of resources and subject the firm to the external monitoring by the
market (Jensen 1986). To control for the impact of dividends, we include a control
variable, divyield, defined as the ratio of the total dividends paid to the common stock
capitalization at the end of the fiscal year. Finally, we include two CEO characteristics
18
that should impact compensation, the number of years the CEO has served in office
(tenure), and the age of the CEO (age).
Table 2 provides summary statistics for each of these variables. We have
complete compensation information to calculate total dollar compensation for 13,251
firm-year observations. The mean tdc1 is $4.51 million while the median level of
compensation is $2.39 million. To control for extreme values affecting our results, we
take the natural logarithm of pay levels (ltdc1) which has a mean value of 7.84 and a
median value of 7.78. The mean (median) totpps for our sample is $766,542 ($223,090).
To control for extreme values affecting our pay-performance results, we also take the
natural logarithm of pay-performance sensitivities denoted as lpps. The mean (median)
lpps for our sample is 12.36 (12.32), suggesting a much more normal distribution than
when we use totpps.
We now examine the relative pay measures dltdc1, dtotpps, and dlpps. The mean
(median) of each of these variables are -0.014 (-0.031), $6,536.5 ($-221,909), and -0.002
(-0.038), respectively. It seems like the distribution of relative pay to the industry is
skewed to the left. The natural logarithm helps mitigate this problem.
The means and medians of our performance variables are as follows. The mean
(median) Tobin’s Q is 1.989 (1.501). The mean (median) roa is 0.13 (0.13), and the
mean (median) annual stock returns is 15.3% (8.61%). The average size as measured by
annual sales is $4.7 billion, and firms have a mean leverage ratio of 22%. Our sample has
an average dividend yield of 1.3%. The average tenure of a CEO is 7.77 years and the
average age of the CEO is 56 years.
19
***Table 2***
4. Empirical Results
4.1 Are there industry effects in CEO pay?
We begin by examining if different industries have different compensation levels
and sensitivities. It seems reasonable that different industries employ CEOs who have
both industry-relevant human capital and general management skills. In order to attract
appropriate managerial talent, firms have to offer these executives wages that are
commensurate with their ability. If the level of managerial skills varies by industry, we
would expect to find an industry effect in the determination of CEO compensation. Table
3 reports the mean values of our four compensation variables, tdc1, ltdc1, pps and lpps,
of firms grouped by the 48 Fama-French industry definitions. To be included in the
analysis, we required that an industry included firms with pay data for at least four firms
in any given year. The dollar level of total compensation, tdc1, ranges from a low of
$1.19 million (Fabricated Products) to a high of $17.13 million (Beer). The pay-for-
performance sensitivity, totpps, range from a low of $0.84 million (Fabricated Products)
to a high of $2.70 million (Beer). To test the statistical significance of the observed
differences in the mean compensation variables, we perform an analysis of variance
(ANOVA) test across industries. Both the F-test and Chi-square statistic indicate that the
distributions of these variables across industries are not identical, providing evidence of
an industry effect for managerial compensation.
*** Table 3***
20
4.2 Is mean industry pay related to mean industry performance?
While above section shows that different industries have significantly different
mean pay, we now examine whether this mean industry pay is correlated with mean
industry performance. We use equally weighted portfolios, so as to not let our results be
biased towards the largest firms. Table 4 shows these regression results with year
dummies and control variables which not presented for brevity. In all three specifications,
we find industry pay and industry performance is correlated. It is reasonable to assume
that firm performance is partially driven by the overall industry performance and
managerial skill. To retain this managerial talent, the successful CEO should be paid
more than her peers in the industry, but only to the extent that firm performance is not
driven by industry performance. That is the optimal sharing rule should reduce the noise
in performance measures consistent with enhanced signal extraction.9 Consequently, we
would expect a positive relationship between relative pay and relative performance.
*** Table 4***
4.3 Testing the relative pay for relative performance hypothesis using OLS
If managerial skills vary with industry then we would expect that CEO of a firm
in industry i should be evaluated relative to CEOs of other firms in that industry i.
Therefore, we would expect that the relative pay of the CEO of firm in industry i should
9 In a dynamic principal-agent model, Demarzo, Fishman, He and Wang (2012) suggest that the managerial
compensation contract may depend on risks that are beyond the manager’s control.
21
be correlated with the relative performance of her firm. We use OLS with standard errors
clustered at the firm level.
Table 5 provides the results for the pooled panel regressions of CEO’s log of the
relative total compensation, dltdc1, on relative firm performance. The table presents six
specifications. Specifications (1) and (2) use dtobinq as the proxy for relative firm
performance. Specifications (3) and (4) uses droa as the proxy for relative firm
performance and specifications (5) and (6) uses dcrawret12 as the proxy for relative firm
performance. Specifications (1), (3) and (5) do not include control variables whiles
specifications (2), (4) and (6) include our control variables. All estimations include year
dummy variables (not reported), and the t-statistics are computed based on robust
standard errors that incorporate firm-level clustering. Note that for each specification,
there is a positive and statistically significant relationship between relative total
compensation and relative firm performance.10
*** Table 5***
Tables 6 and 7 are analogous to that of Table 5 except that Table 6 uses CEO’s
relative pay performance sensitivity, dtotpps, as the dependent variable while Table 7
uses the logarithm of the relative pay performance sensitivity, dlpps, as the dependent
variable. For both tables, we see a statistically significant relationship between relative
pay performance sensitivity and relative firm performance, except when we use dtotpps
as the dependent variable and droa as a proxy for firm performance. The results are
supportive of the relative pay for relative performance hypothesis. All three tables
22
indicate that the sensitivity to relative performance decreases with dividend yield. This
result is consistent with the notion that dividends play a role in reducing the agency costs
of free cash flows (Jensen, 1986) and consequently the role of compensation as a means
of reducing agency costs can be diminished. In all three tables, size as proxied by the
level of firm sales is also positively related to the sensitivity of compensation to firm
relative performance.
*** Tables 6 and 7***
4.4 Testing which industries have different relative pay for relative performance
sensitivities
We would expect that the relative pay for relative performance sensitivity would
depend on the informativeness of the relative performance measures. We acknowledge
that the hiring and retention of CEOs should be directly related to CEO talent for all
industries. However, the more difficult it is for the board to identify the source of firm
success is due to the CEO’s skill or luck, the less sensitive is the CEO’s relative pay to
relative performance. Parrino (1997) argues that industries that tend to hire from within
are easier to monitor and better able to filter the effects of industry shocks. He finds that
firms are more likely to fire CEOs for poor relative performance in industries that tend to
hire CEOs from within. This suggests the importance of industry-specific talent. That is,
the more important the industry-specific knowledge for a CEO to be successful, the more
difficult it is for boards of directors to hire from outside (Parrino 1997, Murphy and
Zabojnik 2007). This suggests that we should examine whether industries with high
10
Our results do not change when we remove the industry means from each of our control variables.
23
insider hiring have a higher sensitivity of relative pay to relative performance compared
to industries that hire CEOs from outside.
We proxy for industry-specific talent by using the percentage of insider CEOs in
the industry in which the firm operates (also used by Parrino 1997, and Cremers and
Grinstein 2011). They suggest that industries with a lot of insider CEOs have a lot of
industry- and firm-specific capital for which they are more highly compensated. Using
the same logic we would expect that industries with lot of inside hires would have higher
relative pay sensitivity to relative performance than industries with more outside hires.
We obtain the percentage of insider CEOs in the industry from Table 3 of Cremers and
Grinstein (2011). We split our sample into terciles of this variable. We classify industries
with the lowest (highest) tercile percentage of insider CEOs as those with the lowest
(highest) level of industry-specific talent. We rerun our regressions on the relative pay for
relative performance on these two extreme terciles, the results of which are given in
Table 8. For brevity we do not present the results on year dummies and the control
variables. In five of the six specifications, we find a significantly greater sensitivity of
relative pay to relative performance in industries where there is more insider hiring than
in industries where there is more outsider hiring. For example, banking (one of the
industries with high insider hiring) has a much greater sensitivity of relative pay to
relative performance than food products (one of the industries with low insider hiring).
These results are consistent with the informativeness of performance measures and our
relative pay for relative performance hypothesis.
*** Table 8***
24
Aggarwal and Samick (1999) suggest that competitive industries will exhibit a
greater pay for relative performance sensitivity than less competitive industries. We first
test for differences in competitiveness between the industries with more insider hiring
and industries with less insider hiring. We calculate the Herfindahl Index (HI), defined as
the sum of the squared market share of each firm in the industry for each fiscal year. The
median HI for industries with more insider hiring is greater than the HI for industries
with less insider hiring, a finding opposite of what their hypothesis requires. We also
reran the regressions in Tables 5-7 while including HI and the interaction of HI and the
relative performance variables. If competitive industries exhibit a greater pay for relative
performance sensitivity than less competitive industries, we would expect the interaction
term to be negative. None of the specification shows such a negative relation (results not
reported).
4.5 Testing the relative pay for relative performance hypothesis using GMM
Tables 9, 10 and 11 report the results from two-step GMM panel regression to
control for dynamic endogeneity. Once again, all estimations include year dummies, the
results of which are not reported. Table 9 summarizes the results when the dependent
variable is dltdc1. Note that for this regression, the relative pay is not positively
associated only if we use dcrawret12 as our independent variable. Table 10 reports the
results when dtopps is the dependent variable. In this case, the relative pay performance
sensitivity is positively related to all of the relative performance measures. Finally, Table
11 reports the results when dltopps is the dependent variable. Once again, there is a
25
positive association between relative pay and relative performance for each of our
performance proxies. Note that in these tables both size and dividend yield play a
diminished role in impacting the sensitivity of compensation to relative firm performance
as compared to the pooled regression results.
*** Tables 9, 10 and 11***
In summary, the above results show strong evidence in support for the relative
pay for relative performance hypothesis using both OLS (Tables 6-8) and dynamic GMM
(Tables 9-11).
4.6 Testing the relative performance evaluation (RPE) hypothesis
In order to ensure that our results are not being driven by the relative performance
evaluation hypothesis (RPE), we also conducted regressions analogous to those
performed by Gibbons and Murphy (1990), Janakiraman, Lambert and Larker (1992),
Barro and Barro (1990), John (1999), and Antle and Smith (1986). These regressions are
reported in Tables 12 and 13. Each table reports the pooled panel regressions and each
table has two panels. In Table 12, the first panel reports the results when the dependent
variable is ∆tdc1 and the second panel reports when the dependent variable is ∆ltdc1.
When we examine all three performance measures we find that the industry effect is
insignificantly related in four of the six specifications (mtobinq and indret) and only
negatively related in the profitability measure (mroa). Table 12 reports the results when
the dependent variable is ∆pps and ∆lpps. In five of the six specifications, the industry
performance measure is not negatively related to pay levels and pay-performance
26
sensitivities. Consistent with the prior literature of Antle and Smith (1986), Gibbons and
Murphy (1990), Barro and Barro (1990), Janakiraman, Lambert and Larker (1992), John
(1999), and Murphy (1999), the results of Tables 12 and 13 are generally not supportive
of the relative performance evaluation hypothesis.
*** Tables 12 and 13***
4.7 Testing the relative pay for relative performance hypothesis using the scaled wealth-performance sensitivity measure
We now examine whether our support for the relative pay for relative
performance hypothesis is dependent on the definition of the pay-performance sensitivity.
Accordingly, we repeat our regressions using the scaled wealth-performance sensitivity
(wpps) of Edmans, Gabaix, and Landier (2009). Note that the optimal incentive structure
here is derived as the dollar change in CEO wealth for a 100-percent change in firm value
divided by annual compensation. We find the mean value of the wpps 25.34 and a median
value of 7.59. We also used the natural logarithms of this scaled wealth-performance
sensitivity (lwpps) which has a mean value of 2.16 and a median value of 2.03. One
again, we calculate the relative scaled wealth-performance sensitivity to the industry
mean (with industry defined as firms not including that firm) for each of these measures.
We call these dependent variables dwpps and dlpps. The results of such an analysis are
given in Tables 14 and 15.11
In examining the regression coefficients across these tables,
we find a positive and statistically significant relationship in 10 out of 12 specifications.
This shows strong support for the relative pay for relative performance hypothesis and
11
We find similar results (not reported) when we use dynamic GMM.
27
suggests that our previous results are not dependent on the definition of pay-performance
sensitivity.
*** Tables 14 and 15***
In summary, we find that the change in relative compensation is positively related
to the relative performance of the industry. These results suggest that managers are both
compensated and evaluated on their talent relative to other firms in their industry.
5. Conclusions
It seems reasonable that different industries employ CEOs who have both
industry-relevant human capital and general management skills. In order to attract
appropriate managerial talent, firms have to offer these executives wages that are
commensurate with their ability. Accordingly, we would expect that the CEO of a bank
should be evaluated relative to CEOs of other banks. If the level of managerial skills
varies by industry, we would expect to find an industry effect in the determination of
CEO compensation. More specifically, managerial pay levels and pay-performance
relationships should vary by industry. In addition, the optimal sharing rule should reduce
the noise in performance measures consistent with enhanced signal extraction. That is,
the manager should be paid based upon the environment s/he can control. Hence, under
these models, industry performance which is not under the control of a specific CEO
should be removed from her performance evaluation. Using these two arguments, this
paper examines the relative pay for relative performance hypothesis, wherein one would
28
expect that the relative pay of the CEO should be correlated with the relative performance
of her firm.
Using a large panel data set of US CEO compensation data for the years 1994 to
2008 this paper tests the relative pay for relative performance hypothesis. We look at
both CEO total compensation levels and CEO pay-performance sensitivities. We find
significant evidence for an industry effect in both CEO pay levels and CEO pay
performance sensitivity. We also find strong evidence in support of the relative pay for
performance hypothesis, using OLS with standard errors clustering at the firm level and
dynamic GMM (Arellano and Bover, 1995) which controls for dynamic endogeneity of
the relationships. We find a stronger relative pay for relative performance relationship in
industries where the performance measures are more informative. We follow Parrino
(1997) in identifying such industries where more industry-specific talent is important.
Finally, the above results are robust to using two measures of the pay-performance
relationship, namely, pay-performance sensitivity or the scaled wealth-performance
sensitivity. These results suggest that managers are both compensated and evaluated on
their talent relative to other firms in their industry.
29
REFERENCES
Aggarwal, R. and A. Samwick, 1999, Executive compensation, strategic competition, and
relative performance evaluation: Theory and Evidence, Journal of Finance 54, 1999-
2043.
Antle, R. and A. Smith, 1986, An empirical analysis of the relative performance
evaluation of corporate executives, Journal of Accounting Research, 24, 1-39.
Albuquerque, A., 2009, Peer firms in relative performance evaluation, Journal of
Accounting and Economics, 48, 69-89.
Albuquerque, A., De Franco, G., and R. Verdi, 2012, Peer choice in CEO compensation,
Journal of Financial Economics, forthcoming.
Arellano, M. and O. Bover, 1995, Another look at the instrumental variable estimation of
error-components models, Journal of Econometrics, 68 (1), 29-51.
Baranchuk, N., MacDonald, G., and J. Yang, 2011, The economics of super managers,
Review of Financial Studies 24, 3323-3368.
Barro, J. and R. Barro, 1990, Pay, performance, and turnover of bank CEOs, Journal of
Labor Economics, 8 (4), 448-481.
Berger, P., Ofek, E., Yermack, D. 1997, Managerial entrenchment and capital structure
decisions, Journal of Finance, 52, 1411-38.
Bizjak, J.M., Lemmon, M.L., and Naveen, L., 2008, Does the use of peer groups
contribute to higher pay andless efficient compensation?, Journal of Financial
Economics 90, 152-168.
Bizjak, J.M., Lemmon, M.L., and Nguyen,T., 2011, Are all CEOs above average? An
empirical analysis of compensation peer groups and pay design, Journal of Financial
Economics, 100, 538-555.
Bradley, M., Jarrell, G. A. and E. H. Kim, 1984, On the existence of an optimal capital
structure: Theory and evidence, The Journal of Finance 39(3), 857-878.
Brick, I., Palmon, O. and J. Wald, 2006, “CEO compensation, director compensation, and
firm performance: Evidence of cronyism?” Journal of Corporate Finance, 12, 403-423.
Cremers, M., and Y. Grinstein, 2011, “Does the Market for CEO Talent Explain
Controversial CEO Pay Practices?”, working paper, Cornell University.
Core, J. and W. Guay, 2002, Estimating the value of employee stock option portfolios and
their sensitivities to price and volatility, Journal of Accounting Research, 40 (3), 613-630.
Core, J., Guay, W., Rusticus, T.,2006, Does weak governance cause weak stock returns?
An examination of firm operating performance and investors’ expectations, Journal of
Finance, 61, 655-687.
30
Custodio, C., Ferreira, M., and P. Matos, 2011, Generalists versus specialists: managerial
skills and ceo pay, working paper, Arizona State.
Dempsey, S. J., Laber G. and M.S. Rozeff, 1993, Dividend policies in practice: Is there
an industry effect? Quarterly Journal of Business and Economics, 32 (4), 3-13.
DeMarzo, P., Fishman, M. He, Z. and N. Wang, 2012, Dynamic agency and the q theory
of investment, Journal of Finance, 67, 2295-2340.
Edmans, A., Gabaix, X. and A. Landier, 2009, A multiplicative model of optimal CEO
incentives in market equilibrium, Review of Financial Studies, 22, 4881-4917.
Esifeldt, A., and C. Kuhnen, 2012, CEO turnover in a competitive assignment
framework, Journal of Financial Economics forthcoming.
Falato, A., Li, D., and T. Milbourn, 2009, Inside the CEO labor market: The role of CEO
talent in pay and turnover decisions, working paper, Federal Reserve Board.
Faulkender, M., and Yang, J., 2010, Inside the black box: The role and composition of
compensation peer groups, Journal of Financial Economics, 96, 257-270.
Frank M.Z. and V.K.Goyal, 2009, Capital structure decisions: Which factors are reliably
important, Financial Management 38 (1), 1-37.
Frydman, C., 2007, Rising through the ranks: The evolution of the market for corporate
executives, 1936-2003, working paper, MIT Sloan.
Gabaix, X. and A. Landier, 2008, Why has CEO pay increased so much? Quarterly
Journal of Economics, 123, 49-100.
Gibbons, R. and L. Katz, 1992, Does unmeasured ability explain inter-industry wage
differentials? The Review of Economics Studies 59 (3), 515-535.
Gibbons, R. and K. Murphy, 1990, Relative performance evaluation of chief executive
officers, Industrial & Labor Relations Review, 43 (3), 30-51.
Gompers, P., Ishii J., A. Metrick, 2003, Corporate governance and equity prices,
Quarterly Journal of Economics, 118, 107-155.
Gong, G., Li, L., and J.Y. Shin, 2011, Relative performance evaluation and related peer
groups in executive compensation contracts, Accounting Review, 86 (3), 1007- 1043.
Goux, D., and E. Maurin, 1999, Persistence of interindustry wage differentials; A
reexamination using firm panel data, Journal of Labor Economics, 17, 492-533.
Guay, W., 1999, The sensitivity of CEO wealth to equity risk: an analysis of the
magnitude and determinants, Journal of Financial Economics, 53, 43-71.
Hausman, J., Taylor, W., 1981, Panel data and unobservable individual effects,
Econometrica, 49, 1377-1398.
31
Himmelberg, C., Hubbard, G., Palia, D.,1999, Understanding the determinants of
managerial ownership and the link between ownership and performance, Journal of
Financial Economics, 53, 353-384.
Himmelberg, C., and G. Hubbard, 2000, Incentive pay and the market for CEOs: An
analysis of pay-for-performance sensitivity, working paper, Columbia Business School.
Holmstrom, B., 1979, Moral hazard and observability, The Bell Journal of Economics,
10, 74-90.
Holmstrom, B. and P. Milgrom, 1987, Multitask principal agent analysis-Incentive
contracts, asset ownership, and job design, Journal of Law Economics and Organization
7, 237-275.
Holmstrom, B. and P. Milgrom, 1992, Aggregation and linearity in the provision of
intertemporal incentives, Econometrica 55, 303-328.
Jewitt, I., 1988, Justifying the first-order approach to principal-agent problems,
Econometrica, 58 (5), 1177-1190.
Janakiraman, S., Lambert, R., and D. Larcker, 1992, An empirical investigation of the
relative performance evaluation hypothesis, Journal of Accounting Research, 30 (1), 53-
69.
Jensen, M. and K. Murphy, 1990, Performance pay and top management incentives,
Journal of Political Economy, 76 (2), 323-329.
Jensen, M., 1986, Agency costs of free cash flow, corporate finance and takeovers,
American Economics Review, 76 (2), 323-329.
Joh, S., 1999, Strategic managerial incentive compensation in Japan: Relative
performance evaluation and product market collusion, Review of Economics and
Statistics, 81 (2), 303-313.
John, K. and John, T., 1993, Top management compensation and capital structure,
Journal of Finance, 48 (3), 949-974.
Joskow, P., Rose, N., and A. Shepard, 1993, Regulatory constraints on ceo compensation,
Brookings Papers on Economic Activity, Microeconomics, 1-72.
Joskow, P., Rose, N., and C. Wolfram, 1996, Political constraints on executive
compensation: Evidence from the electric utility industry, Rand Journal of Economics,
27, 165-182.
Jung, K, Kim, Y., Stulz, R., 1996, Timing, investment opportunities, managerial
discretion, and the security issue decision, Journal of Financial Economics, 42, 159-185.
Krueger, A., and L. Summers, 1987, Reflections on the inter-industry wage structure, in
K. Lang and J. S. Leonard, eds., Unemployment and the Structure of Labor Markets,
Oxford: Blackwell, 1987, pp. 17-47.
32
Krueger, A. and L. Summers, 1988, "Efficiency wages and the inter-industry wage
structure," Econometrica, 56, 259-93.
Lewellen, S., 2013, Executive compensation and peer effects, working paper, Yale
University,
Lintner, J. 1953, The determinants of corporate savings, in Heller et. al (ed), Savings in
the Modern Economy, University of Minnesota Press.
Leary, M., and M. Roberts, 2010, Do peer firms affect corporate financial policy?
Working paper, Rodney L. White Center for Financial Research, University of
Pennsylvania.
Lucas, R., 1978, On the size distribution of firms, Bell Journal of Economics, 9, 508-523.
MacKay , P. and G. Phillips, 2005, How does industry affect firm financial structure?
Review of Financial Studies, 18(4), 1433-1466
McCabe, G., 1979, The empirical relationship between investment and financing: A new
look," Journal of Financial and Quantitative Analysis 14 (1), 119-135.
Michel A., 1979, Industry influence on dividend policy, Financial Management, 8 (3),
22-26.
Murphy, K., 1985, Corporate performance and managerial remuneration: An empirical
analysis, Journal of Accounting and Economics, 7, 11-42.
Murphy, K., 1999, Executive compensation, in Orley Ashenfelter and David Card (eds.),
Handbook of Labor Economics, Vol. 3b, Elsevier Science North Holland, Chapter 38:
2485-2563.
Murphy, K., and R. Topel, 1987, Unemployment risk, and earnings: testing for equalizing
wage differences in the labor market, in K. Lang and J. S. Leonard, eds., Unemployment
and the Structure of Labor Markets, Oxford: Blackwell, 103-40.
Murphy, K., and J. Zabojnik, 2003, CEO pay and appointments: A market-based
explanation for recent trends, AEA Papers and Proceedings, 94, 192-196.
Murphy, K., and J. Zabojnik, 2007, Managerial capital and the market for CEOs, working
paper, University of Southern California.
Oyer, P., 2004, Why do firms use incentives that have no incentive effects? Journal of
Finance, 59, 1619-1649.
Palia, D., 2000, The impact of regulation on ceo labor markets, Rand Journal of
Economics, 31, 165-179.
Palia, D., 2001, The endogeniety of managerial compensation in firm value: a solution,
Review of Financial Studies, 14, 735-64.
Roberts, M. and T. Whited, forthcoming, Endogeneity in empirical corporate finance, in
G. Constantinides and R. Stulz, eds., Handbook of the Economics of Finance, 2, Elsevier.
33
Rosen, S., 1981, The economics of superstars, American Economic Review, 71,845-858.
Shin, H. and R.M. Stulz, 1998, Are internal capital markets efficient? Quarterly Journal
of Economics, 113 (2), 531-552
Smith, C.W. and R. L. Watts, 1992, The investment opportunity set and corporate
financing, dividend, and compensation policies, Journal of Financial Economics, 32,
263-292.
Slichter, S.H., 1950, Notes on structure of wages, Review of Economics and Statistics, 32
(1), 80-91.
Tervio, M, 2008, Difference that CEOs make: An assignment model approach, American
Economic Review, 98, 642-668.
Thaler, R.H., 1989, Anomalies: Inter-industry wage differentials, Journal of Economics
Perspective 3, 181-193.
Titman, S., and R. Wessels, The determinants of capital structure choice, Journal of
Finance, 43, 1-19.
Wintoki, B, Linck, J., and J. Netter (2012), Endogeneity and the dynamics of corporate
governance, Journal of Financial Economics, 105, 581-606.
34
Table 1: Variable definitions
Variable Definition
Dependent Variables:
tdc1 CEO’s total compensation: salary, bonus and other annual payouts including
granted options, restricted stock and long-term incentive payouts
dltdc1 Relative total compensation: the natural log of tdc1- the industry mean log of tdc1
totpps Pay-performance sensitivity: the dollar change in CEO’s wealth in outstanding
options and stock for one percent change in stock price (Core and Guay 1999)
dtotpps Relative pay-performance sensitivity: totpps - the industry mean totpps
lpps Natural log of the pay-performance sensitivity
dlpps Relative natural log of the pay-performance sensitivity: lpps - the industry mean
lpps
Independent Variables
tobinq Tobin’s Q: the market value of assets over book value of assets
mtobinq Industry mean Tobin's Q
dtobinq Tobin's Q - the industry mean Tobin's Q
crawret12 Stock return: the holding period return for 12 months of the current fiscal year
indret Industry mean stock return: the equal-weighted holding return for the industry for
a given calendar year and month
dcrawret12 Relative stock return: crawret12- indret
roa Return on assets: earnings before interest, tax, depreciation and amortization over
total assets
mroa Industry mean ROA
droa Relative ROA: roa - mroa
rd R&D expense over total assets
lsales Natural log of sales
lev Total debt divided by total assets
divyield Current fiscal year dividend yield
tenure The number of years the CEO is in office
age The age of the CEO
35
Table 2: Descriptive statistics
Variable N Mean Median Std. Dev. Min Max
tdc1a 13,251 $4,507.94 $2,393.33 $8,299.78 $0.00 $369,888.00
ltdc1 13,250 7.839 7.781 1.034 -6.908 12.821
dltdc1 13,250 -0.014 -0.031 0.987 -14.910 4.564
totpps 13,142 $766,541.7 $223,089.6 $3,014,890.0 $331.77 $108,000,778.4
lpps 13,142 12.357 12.315 1.436 5.804 18.498
dtotpps 13,142 $6,536.48 -$221,909.7 $3,053,746 -$6,339,666 $107,000,000
dlpps 13,142 -0.002 -0.038 1.364 -6.601 6.062
tobinq 13,251 1.989 1.501 1.825 0.298 78.565
dtobinq 13,619 0.006 -0.164 1.675 -5.142 73.775
roa 12,826 0.131 0.133 0.112 -2.671 0.807
droa 12,826 0.118 0.090 0.165 -2.360 2.255
crawret12 12,909 1.152875 1.086107 0.6166814 0.0215827 27.19421
dcrawret12 12,909 0.0011802 -0.0375959 0.5608874 -2.385436 24.74578
Salesb 13,038 $4,692.33 $1,239.9 $13,295.2 $0 $402,298
lsales 13,029 7.194 7.125 1.582 -2.364 12.905
lev 12,989 0.220 0.211 0.167 0.000 0.959
divyield 13,034 0.013 0.005 0.027 0.000 1.445
tenure 12,403 7.772 5.671 7.078 0.033 55.033
age 12,758 55.518 56 7.118 29 90
a $thousands, and
b $millions, respectively.
36
Table 3: Analysis of Variance for CEO compensation across industry This table reports the analysis of variance for CEO compensation for 2,113 sample firms over the period 1994-
2008. The four CEO compensation variables, TDC1, LTDC1, TOTPPS and LPPS are as defined in Table 1.
tdc1 ltdc1 totpps lpps
Industry ff Mean Std. Dev. Mean Std.
Dev. Mean Std. Dev. Mean
Std.
Dev.
Food 2 $4,655 $4,489 7.96 1.04 $8,588 $25,391 12.72 1.28
Beer 4 $17,131 $21,267 9.19 1.09 $26,683 $15,166 14.59 0.77
Toys 6 $4,994 $6,170 8.06 0.92 $13,283 $42,691 12.68 1.52
Fun 7 $5,623 $8,095 7.83 1.51 $10,755 $17,795 12.91 1.52
Books 8 $4,213 $3,270 8.06 0.79 $6,440 $9,090 12.63 1.33
Hshld 9 $4,894 $6,174 8.01 0.99 $6,323 $13,312 12.41 1.42
Clths 10 $4,474 $6,339 7.84 1.03 $6,648 $11,262 12.26 1.54
Hlth 11 $4,452 $5,172 7.88 1.03 $6,532 $8,767 12.73 1.16
MedEq 12 $3,066 $4,231 7.58 0.90 $7,674 $19,761 12.54 1.29
Drugs 13 $6,047 $6,842 8.08 1.19 $8,068 $11,643 12.79 1.41
Chems 14 $3,518 $2,872 7.85 0.82 $3,369 $8,019 11.96 1.19
Rubbr 15 $2,524 $3,503 7.41 0.88 $2,672 $2,406 12.17 0.84
Txtls 16 $1,317 $977 7.00 0.59 $6,203 $21,358 11.17 1.84
BldMt 17 $3,348 $3,948 7.68 0.90 $5,010 $10,539 12.05 1.37
Cnstr 18 $6,841 $8,083 8.26 1.10 $10,242 $19,025 12.59 1.62
Steel 19 $2,536 $2,471 7.50 0.80 $1,989 $3,154 11.54 1.18
FabPr 20 $1,187 $645 6.96 0.50 $838 $831 11.08 0.66
Mach 21 $3,239 $3,135 7.73 0.84 $2,842 $3,788 11.98 1.09
ElcEq 22 $4,250 $10,822 7.69 0.98 $4,560 $8,794 12.11 1.25
Autos 23 $3,842 $4,650 7.78 0.95 $4,039 $5,430 12.08 1.36
Aero 24 $6,193 $6,191 8.28 0.99 $8,074 $14,917 12.55 1.42
Ships 25 $3,173 $1,436 7.97 0.46 $2,490 $2,184 12.14 0.74
Guns 26 $4,091 $4,347 7.73 1.19 $3,535 $3,256 12.23 1.20
Gold 27 $1,942 $1,766 7.29 0.71 $1,138 $1,869 10.78 1.33
Mines 28 $5,848 $5,902 8.17 1.08 $3,238 $4,564 12.08 1.09
Oil 30 $4,084 $7,045 7.76 0.99 $3,526 $5,978 12.04 1.22
Util 31 $2,727 $3,098 7.50 0.89 $1,753 $2,708 11.24 1.32
Telcm 32 $10,098 $20,209 8.31 1.31 $19,315 $64,419 12.82 1.67
PerSv 33 $4,906 $11,259 7.73 1.03 $8,310 $15,651 12.59 1.44
BusSv 34 $3,801 $4,762 7.83 0.89 $6,002 $10,049 12.53 1.21
Hardw 35 $6,454 $20,328 8.02 1.22 $11,415 $53,593 12.39 1.60
Softw 36 $6,824 $18,503 7.93 1.32 $10,363 $34,377 12.68 1.36
Chips 37 $4,582 $6,026 7.86 1.07 $7,512 $18,708 12.49 1.38
LabEq 38 $3,182 $4,746 7.61 0.90 $4,624 $7,118 12.32 1.20
Paper 39 $2,616 $2,720 7.50 0.84 $2,109 $2,636 11.66 1.14
Boxes 40 $3,254 $1,639 7.92 0.66 $4,157 $6,612 12.25 1.19
Trans 41 $3,059 $3,412 7.60 0.90 $17,949 $101,103 12.28 1.46
Whlsl 42 $2,670 $2,433 7.58 0.80 $3,787 $6,254 11.92 1.42
Rtail 43 $4,753 $7,432 7.87 1.09 $8,833 $21,669 12.50 1.57
Meals 44 $3,942 $4,552 7.73 1.09 $9,776 $25,288 12.58 1.57
Banks 45 $4,916 $6,406 7.99 0.97 $7,152 $15,413 12.61 1.28
Insur 46 $5,216 $5,411 8.15 0.91 $12,861 $36,297 12.88 1.41
Fin 48 $7,949 $10,464 8.32 1.19 $26,506 $86,204 13.19 1.66
37
Other 49 $5,553 $11,387 7.61 1.32 $12,177 $27,548 12.19 2.28
Total
$4,508 $8,300 7.84 1.03 $7,665 $30,149 12.36 1.44
F-
Statistic 11.30a
15.54a
7.8a
30.49a
Bartlett's
Test for
10,000b
650b
2,000b
412b
Equal
Variances
aAll F-statistics are statistically significant at the 1% level.
bAll Chi-Square tests are with 43 degrees of freedom and are statistically significant at the 1% level.
38
Table 4: Pooled panel regressions of mean industry CEO compensation on mean industry
performance
This table reports the regression results when regressing mean industry compensation variables against mean
performance variables over the period 1994-2008. All estimations include year effects and t-statistics are
computed based on robust standard errors that incorporate firm-level clustering. Control variables are included
but not reported. a, b and c denotes significance at 1%, 5% and 10% level, respectively.
Industry mean log of total compensation
(ltdc1)
Industry mean log of pay-performance
sensitivity (lpps)
Model (1) (2) (3)
(4) (5) (6)
Intercept 8150.573a 8121.950
a 8724.849
a 12.753
a 12.758
a 12.352
a
(10.69) (10.64) (11.25) (96.47) (97.17) (95.83)
Industry performance measure:
Mean Q 59.650b 0.014
b
(2.48) (2.48)
Mean ROA -2728.502a -0.440
(-6.39) (-0.450)
Mean return 623.892a 0.586
a
(4.10) (30.20)
control variables Yes Yes Yes Yes Yes Yes
R2 0.26 0.27 0.25 0.50 0.50 0.39
39
Table 5: Pooled panel regressions of CEO’s relative total compensation on relative firm
performance
This table reports the panel regression results for sample firms over the period 1994-2008. The dependent
variable is dltdc1, the natural log of total compensation relative to the industry mean. All estimations include
year effects and t-statistics are computed based on robust standard errors that incorporate firm-level clustering.
a, b and c denotes significance at 1%, 5% and 10% level, respectively.
Dependent variable: relative total compensation (dltdc1)
Model (1) (2) (3) (4) (5) (6)
Intercept -0.010 -2.951a
-0.080c
-2.769a -0.003 -2.845
a
(-0.26) (-10.04) (-1.93) (-9.55) (-0.07) (-10.38)
Relative performance measure:
dtobinq 0.065a 0.078
a
(6.25) (8.54)
droa 0.607a 0.646
a
(5.27) (8.04)
dcrawret12 0.081a 0.112
a
(3.67) (3.71)
Control Variables:
rd 1.960a 1.933
a 2.089
a
(5.66) (7.48) (5.96)
lsales 0.355a 0.278
a 0.350
a
(5.20) (3.85) (5.61)
lsalessq 0.003 0.008c 0.003
(0.70) (1.72) (0.82)
divyield -2.004a -1.838
a -1.872
a
(-3.32) (-3.15) (-3.17)
lev 0.115 0.092 0.006
(1.38) (1.11) (0.07)
tenure -0.001 -0.001 -0.001
(-0.31) (-0.31) (-0.23)
age 0.000 0.000 -0.002
(-0.14) (-0.05) (-0.65)
R2 0.01 0.38 0.01 0.37 <0.01 0.37
N 13,033 11,702 12,825 11,520 12,908 11,425
40
Table 6: Pooled panel regressions of CEO’s relative pay-performance
sensitivity on relative firm performance
This table reports the panel regression results for sample firms over the period 1994-2008. The dependent
variable is dtotpps, the CEO’s pay-performance sensitivity relative to the industry mean. All estimations include
year effects and t-statistics are computed based on robust standard errors that incorporate firm-level clustering.
All coefficients are expressed as a fraction of 106. a, b and c denotes significance at 1%, 5% and 10% level,
respectively.
Dependent variable: relative pay-performance sensitivity (dtotpps)
Model (1) (2) (3) (4) (5) (6)
Intercept 0.008 -2.441a -0.035 -2.000
a 0.004 -2.241
a
(0.20) (-3.62) (-0.58) (-3.05) (-0.10) (-3.39)
Relative performance measure:
dtobinq 0.253a
0.274a
(5.80) (6.18)
droa 0.342 0.450
(0.87) (1.46)
dcrawret12 0.313a 0.304
a
(5.44) (5.42)
Control Variables:
rd 0.380 0.584 0.856c
(0.76) (1.33) (1.75)
lsales 0.090 -0.003 0.068
(0.51) (-0.02) (0.40)
lsalessq 0.018 0.024c
0.020
(1.27) (1.83) (1.45)
lev -2.256b
-2.069b -2.320
b
(-2.24) (-2.05) (-2.28)
divyield -0.331 -0.668b -0.699
a
(-1.34) (-2.45) (-2.75)
tenure 0.073a 0.073
a 0.074
a
(3.07) (3.01) (3.06)
age 0.004 0.002 0.002
(0.25) (0.12) (0.13)
R2
0.02 0.09 0.00 0.06 <0.01 0.06
N 12,945 11,630 12,737 11,448 12,898 11,415
41
Table 7: Pooled panel regressions of log of CEO’s relative pay-performance
sensitivity on relative firm performance
This table reports the panel regression results for sample firms over the period 1994-2008. The dependent
variable is dlpps, the log of CEO’s pay-performance sensitivity relative to the industry mean. All estimations
include year effects and t-statistics are computed based on robust standard errors that incorporate firm-level
clustering. a, b and c denotes significance at 1%, 5% and 10% level, respectively.
Dependent Variable: Log of relative pay-performance sensitivity (dlpps)
Model (1) (2) (3) (4) (5) (6)
Intercept 0.032 -3.872a -0.119
c -3.431
a 0.015 -3.737
a
(0.54) (-10.45) (-1.90) (-9.39) (0.24) (-10.73)
Relative performance measure:
dtobinq 0.203a
0.215a
(6.33) (6.46)
droa 1.275a 1.237
a
(8.40) (9.87)
dcrawret12 0.356a 0.366
a
(8.20) (6.85)
Control variables:
rd 1.340a 1.345
a 1.748
a
(3.47) (4.68) (4.53)
lsales 0.492a 0.332
a 0.494
a
(5.85) (3.88) (6.50)
lsalessq -0.003 0.008 -0.003
(-0.56) (1.35) (-0.58)
lev -5.223a -5.101
a -5.105
a
(-4.87) (-4.69) (-4.85)
divyield -0.224c
-0.417a -0.503
a
(-1.65) (-3.09) (-3.69)
tenure 0.068a 0.068
a 0.069
a
(17.93) (17.00) (17.25)
age -0.001 -0.002 -0.004
(-0.41) (-0.40) (-1.01)
R2
0.06 0.41 0.02 0.36 0.02 0.36
N 12,945 11,630 12,737 11,448 12,898 11,415
42
Table 8: Differences in the relative pay for relative performance across industries with low and high insider hiring
This table reports the panel regression results for sample firms over the period 1994-2008. All estimations include year effects and t-statistics are computed based
on robust standard errors that incorporate firm-level clustering. Control variables are included but not reported. a, b and c denotes significance at 1%, 5% and
10% level, respectively.
Industries with low insider hiring
Industries with high insider hiring
t-statistics for differences
Model (1) (2) (3)
(4) (5) (6)
(1) - (4) (2) - (5) (3) - (6)
Q ROA Returns
Q ROA Returns
Q ROA Returns
Panel A: dltdc1 0.076a 0.615
a 0.038 0.117
a 1.017
a 0.179
a -0.041
c -0.402
c -0.141
a
(6.9) (4.56) (1.27) (5.43) (6.44) (5.34) (-1.72) (-1.96) (-3.13)
Control variables Yes Yes Yes Yes Yes Yes Yes Yes Yes
R2 0.402 0.391 0.378 0.455 0.461 0.454
Panel B: dlpps 0.205a 1.256
a 0.271
a 0.392
a 1.329
a 0.476
a -0.187
a -0.073 -0.205
b
(7.88) (6.44) (4.17) (8.69) (4.84) (6.64) (-3.59) (-0.22) (-2.11)
Control variables Yes Yes Yes Yes Yes Yes Yes Yes Yes
R2 0.436 0.373 0.363 0.409 0.372 0.375
43
Table 9: Dynamic panel estimation of relative total compensation
This table reports the results from two-step GMM panel regression for sample firms over the period 1994-2008.
The dependent variable is dltdc1, the natural log of total compensation relative to the industry mean. All
estimations include year effects and Windmeijer (2005) bias-corrected robust estimator z-statistics are reported
in parentheses. a, b and c denotes significance at 1%, 5% and 10% level, respectively.
Dependent variable: relative total compensation (dltdc1)
(1) (2) (3)
dltdct-1 0.140a
0.143a 0.135
a
(3.40) (3.81) (3.71)
Relative performance measure:
dtobinq 0.088a
(3.15)
droa 1.385a
(5.70)
dcrawret12 -0.017
(-0.30)
Control variables:
rd 0.548c
0.183 0.623b
(1.89) (1.01) (2.07)
lsales -0.223a -0.285
a -0.190
a
(-5.49) (-6.87) (-5.01)
lsalessq 0.037a 0.041
a 0.034
a
(10.65) (11.94) (10.82)
lev -0.040 0.039 -0.199b
(-0.43) (0.43) (-2.23)
divyield -1.453b
-0.625 -1.490b
(-2.33) (-1.15) (-2.10)
tenure 0.002 0.000 0.003
(0.94) (0.15) (1.13)
age -0.009a -0.008
a -0.010
a
(-3.59) (-3.38) (-4.45)
AR(2) (Prob > z) 0.20 0.48 0.52
N 8,816 8,697 8,666
Note: 1. AR(2) is test for second-order serial correlation in the first-differenced residuals under the null
hypothesis of no serial correlation.
2. The instruments used are the explanatory variables lagged by three years and higher.
44
Table 10: Dynamic panel estimation of relative pay-performance sensitivity
This table reports the results from two-step GMM panel regression for sample firms over the period 1994-2008.
The dependent variable is dtotpps, the total pay-performance sensitivity relative to the industry mean. All
estimations include year effects and Windmeijer (2005) bias-corrected robust estimator z-statistics are reported
in parentheses. All coefficients are expressed as a fraction of 105. a, b and c denotes significance at 1%, 5% and
10% level, respectively.
Dependent Variable: relative pay-performance sensitivity (dtotpps)
(1) (2) (3)
dtotppst-1 0.000a
0.000a 0.000
a
(10.45) (17.55) (21.34)
Relative performance measure:
dtobinq 2.534a
(4.45)
droa 8.622b
(2.50)
dcrawret12 6.430a
(5.24)
Control variables:
rd -2.271 -1.295 0.950
(-1.16) (-0.64) (0.49)
lsales -1.383a -2.001
a -0.980
b
(-3.44) (-4.21) (-2.53)
lsalessq 0.165a 0.213
a 0.130
a
(5.34) (5.96) (4.67)
lev 0.119 -1.318 -1.130
(0.10) (-1.07) (-1.15)
divyield -4.959 -3.708 -1.450
(-1.01) (-0.80) (-0.40)
tenure 0.336a 0.292
a 0.310
a
(5.75) (6.54) (7.20)
age -0.020 -0.009 -0.040
(-0.65) (-0.32) (-1.30)
AR(2) (Prob > z) 0.30 0.29 0.33
N 8,788 8,669 8,660
Note: 1. AR(2) is test for second-order serial correlation in the first-differenced residuals under the null
hypothesis of no serial correlation.
2. The instruments used are the explanatory variables lagged by three years and higher.
45
Table 11: Dynamic panel estimation of relative log pay-performance sensitivity
This table reports the results from two-step GMM panel regression for sample firms over the period 1994-2008.
The dependent variable is dlpps, the log of pay-performance sensitivity relative to the industry mean. All
estimations include year effects and Windmeijer (2005) bias-corrected robust estimatorl. z-statistics are reported
in parentheses. a, b and c denotes significance at 1%, 5% and 10% level, respectively.
Dependent Variable: relative log of pay-performance sensitivity
(dlpps)
(1) (2) (3)
dlppst-1 0.371a
0.478a 0.482
a
(7.05) (9.25) (8.90)
Relative performance measure:
dtobinq 0.118a
(5.46)
droa 0.837a
(3.58)
dcrawret12 0.408a
(3.29)
Control variables:
rd -0.185 -0.283 -0.013
(-0.62) (-1.25) (-0.05)
lsales -0.201a -0.211
a -0.139
a
(-4.81) (-4.83) (-3.85)
lsalessq 0.030a 0.028
a 0.023
a
(8.28) (7.72) (7.24)
lev -0.195b
-0.190b -0.214
b
(-2.15) (-2.00) (-2.38)
divyield -3.491a -3.226
a -2.726
a
(-3.92) (-3.95) (-3.59)
tenure 0.043a 0.035
a 0.037
a
(9.51) (7.82) (7.75)
age -0.006b -0.005
c -0.008
a
(-2.35) (-1.78) (-3.23)
AR(2) (Prob > z) 0.22 0.16 0.28
N 8,788 8,669 8,660
Note: 1. AR(2) is test for second-order serial correlation in the first-differenced residuals under the null
hypothesis of no serial correlation.
2. The instruments used are the explanatory variables lagged by three years and higher.
46
Table 12: Tests of the Relative Performance Evaluation (RPE) hypothesis: pooled panel
regressions of change in total compensation on firm performance and industry performance
This table reports the pooled panel regression results for sample firms over the period 1994-2008. The
dependent variable in Panel A is total compensation: tdc1t- tdc1t-1, and the dependent variable in Panel B is
log of total compensation: ltdc1t- ltdc1t-1. All estimations include year effects and t-statistics are computed
based on robust standard errors that incorporate firm-level clustering. a, b and c denotes significance at 1%, 5%
and 10% level, respectively.
Dependent variable Panel A: Change in total compensation Panel B: Change in Log of total compensation
tdc1
ltdc1
Intercept
-847.506 432.115 -841.235 -0.286b
-0.132 -0.376a
(-1.06) (0.48) (-1.02) (-2.90) (-1.38) (-3.74)
Performance measure:
Tobinq
406.314a
0.030a
(3.13)
(3.55)
Mtobinq
51.632
0.004
(1.29)
(0.80)
Roa
2031.091a
0.403a
(3.12)
(5.62)
Mroa
-936.412b
-0.087b
(-2.07)
(-1.93)
crawret12 995.047c
0.159b
(1.85) (2.16)
Indret 406.517 0.102c
(0.96) (1.74)
Control variables:
Rd
-4355.187a -1247.372 -1750.022
c -0.509
a -0.200
c -0.374
a
(-2.66) (-1.24) (-1.86) (-3.95) (-1.72) (-3.49)
Lsales
-106.263 -278.556 -91.997 0.029 -0.013 0.026
(-0.48) (-1.24) (-0.44) (1.40) (-0.58) (1.23)
Lsalessq
12.654 23.758 12.010 -0.001 0.001 -0.001
(0.83) (1.55) (0.81) (-1.04) (0.92) (-0.89)
Lev 668.363 76.677 42.360 -0.015 -0.035 -0.039
(1.63) (0.23) (0.13) (-0.36) (-0.84) (-0.96)
Divyield
1040.717 85.340 1423.564 -0.353 -0.393 -0.128
(0.29) (0.02) (0.39) (-1.11) (-1.18) (-0.47)
Tenure
4.803 5.577 5.402 -0.002a -0.002
a -0.002
a
(0.69) (0.81) (0.78) (-3.34) (-3.25) (-3.31)
Age
-5.842 -9.042 -10.322 0.001 0.001 0.001
(-0.75) (-1.11) (-1.29) (1.59) (1.50) (1.29)
R
2
0.02 0.01 0.02 0.02 0.02 0.02
N
8,817 8,698 8,816 8,817 8,697 8,816
47
Table 13: Tests of the Relative Performance Evaluation (RPE) hypothesis:
pooled panel regressions of change in total compensation on firm performance
and industry performance
This table reports the pooled panel regression results for sample firms over the period 1994-2008. The
dependent variable in Panel A is pay for performance sensitivity: ppst-ppst-1, and the dependent variable in
Panel B is log of pay-performance sensitivity: lppst-lppst-1. All estimations include year effects and t-statistics
are computed based on robust standard errors that incorporate firm-level clustering. The coefficients for pps
are expressed as a fraction of 105. a, b and c denotes significance at 1%, 5% and 10% level, respectively.
Dependent variable Panel A: Change in total compensation Panel B: Change in Log of total compensation
pps
lpps
Intercept
-8.428a
-4.051 -8.686a -1.043
a -0.786
a -1.320
a
(-2.87) (-1.47) (-3.32) (-8.97) (-7.25) (-11.89)
Performance measure:
Tobinq
1.733a
0.063a
(6.31)
(5.88)
Mtobinq
-0.041
-0.003c
(-1.26)
(-1.77)
Roa
3.125a
0.543a
(2.77)
(6.53)
Mroa
0.755
-0.062
(0.84)
(-0.95)
crawret12 4.762a 0.460
a
(6.66) (4.04)
Indret 1.251c
0.222a
(1.80) (3.11)
Control variables:
Rd
-10.220a 2.404 -0.997 -0.539
a 0.004 -0.374
a
(-3.57) (1.38) (-0.61) (-3.86) (0.02) (-3.75)
Lsales
0.824 0.647 0.831 0.080a 0.025 0.067
a
(1.19) (1.06) (1.31) (3.16) (1.02) (3.03)
Lsalessq
-0.055 -0.043 -0.054 -0.004b -0.001 -0.003
b
(-1.15) (-0.99) (-1.20) (-2.49) (-0.41) (-2.28)
Lev 3.091a 0.101 0.456 0.050 -0.015 0.020
(3.46) (0.14) (0.65) (1.16) (-0.36) (0.55)
Divyield
-3.582 -9.375b -1.803 -2.218
a -2.438
a -1.670
a
(-1.18) (-2.59) (-0.66) (-5.26) (-5.26) (-4.61)
Tenure
0.103b
0.109b 0.106
b 0.001 0.002
c 0.001
c
(2.25) (2.37) (2.32) (1.47) (1.77) (1.70)
Age
-0.023 -0.045 -0.043 0.002c
0.002 0.001
(-0.50) (-0.96) (-0.93) (1.99) (1.40) (1.48)
R
2
0.06 0.02 0.06 0.14 0.12 0.30
N
8,788 8,669 8,660 8,788 8,669 8,660
48
Table 14: Pooled panel regressions of CEO’s relative wealth scaled pay-
performance sensitivity on relative firm performance
This table reports the panel regression results for sample firms over the period 1994-2008. The dependent
variable is dwpps, the wealth scaled CEO’s pay-performance sensitivity relative to the industry mean. All
estimations include year effects and t-statistics are computed based on robust standard errors that incorporate
firm-level clustering. a, b and c denotes significance at 1%, 5% and 10% level, respectively.
Dependent variable: Relative wealth scaled pay-performance sensitivity (dwpps)
Model (1) (2) (3) (4) (5) (6)
Intercept 0.522 -42.001c
-1.030 -34.569 0.423 -38.775
(0.40) (-1.72) (-0.60) (-1.46) (0.33) (-1.60)
Relative performance measure:
Dtobinq 5.477a
5.040a
(3.18) (3.01)
Droa 13.571 7.929
(1.45) (0.90)
dcrawret12 7.098a 6.793
b
(2.59) (2.32)
Control variables:
Rd -18.828 -14.647 -8.932
(-0.77) (-0.59) (-0.35)
Lsales 8.958c 7.403 8.638
c
(1.84) (1.63) (1.80)
Lsalessq -0.428 -0.323 -0.394
(-1.39) (-1.15) (-1.31)
Lev -38.713c -34.079 -38.545
c
(-1.78) (-1.56) (-1.69)
Divyield -25.896a -33.275
a -32.546
a
(-2.66) (-3.14) (-3.05)
Tenure 2.788a 2.786
a 2.826
a
(5.14) (5.06) (5.10)
Age -0.254 -0.286 -0.292
(-0.92) (-1.01) (-1.03)
R2
0.01 0.05 <0.01 0.04 <0.01 0.05
N 13,032 11,702 12,822 11,520 12,903 11,425
49
Table 15: Pooled panel regressions of log of CEO’s wealth scaled relative pay-
performance sensitivity on relative firm performance
This table reports the panel regression results for sample firms over the period 1994-2008. The dependent
variable is dlwpps, the log of wealth scaled CEO’s pay-performance sensitivity relative to the industry mean.
All estimations include year effects and t-statistics are computed based on robust standard errors that
incorporate firm-level clustering. a, b and c denotes significance at 1%, 5% and 10% level, respectively.
Dependent Variable: Log of relative wealth scaled pay-performance sensitivity
(dlwpps)
Model (1) (2) (3) (4) (5) (6)
Intercept 0.071 -1.045a -0.787
a 0.044 -0.932
a -0.932
a
(1.21) (-3.65) (-2.63) (0.74) (-3.20) (-3.20)
Relative performance measure:
dtobinq 0.144a
0.140a
(5.08) (4.88)
droa 0.748a 0.655
a
(6.35) (5.65)
dcrawret12 0.287a 0.266
a
(9.29) (8.38)
Control variables:
rd -0.482c
-0.458 -0.239
(-1.80) (-1.51) (-0.90)
lsales 0.164a 0.078 0.155
a
(2.83) (1.27) (2.64)
lsalessq -0.007c -0.001 -0.006
c
(-1.91) (-0.31) (-1.65)
lev -2.888a -2.921
a -2.814
a
(-3.27) (-3.16) (-3.22)
divyield -0.363a -0.535
a -0.535
a
(-3.00) (-4.36) (-4.38)
tenure 0.069a 0.068
a 0.070
a
(16.88) (16.27) (16.47)
age -0.001 -0.001 -0.002
(-0.28) (-0.30) (-0.66)
R2 0.04 0.23 0.01 0.20 0.02 0.20
N 13,032 11,702 12,822 11,520 12,903 11,425