[job market paper] - nyupages.stern.nyu.edu/~hchung2/jobmarketpaper1127_haejinchung.pdf[job market...
Post on 15-Mar-2020
2 Views
Preview:
TRANSCRIPT
[Job Market Paper]
Board Independence and CEO Incentives
Hae Jin Chung∗ 27 November 2008
Abstract
Contrary to a commonly-held view in the corporate governance literature, I argue theoretically that the optimal pay-performance sensitivity (PPS) should be smaller in the presence of board monitoring for a risk-averse CEO. My model is based on a simple adaptation of Holmstrom and Milgrom (Econometrica 1987). I show that board monitoring and PPS should be substitutes and the relative weights placed on board monitoring and PPS should depend on firm transparency (ease of monitoring). It is a prediction of my model that if firms that were relying on incentives (PPS) are mandated to strengthen monitoring, their constrained optimal PPS would be smaller.
Using the percentage of outside directors as a proxy for board monitoring, I find empirical evidences consistent with these predictions. In 2002, following the adoption of the Sarbanes-Oxley Act of 2002, major U.S. exchanges began to require the boards of listed firms to have more than 50% of outside directors. In the case of firms affected by this requirement, their CEO pay-performance sensitivity decreased significantly relative to the control group, especially in the case of large firms which are more transparent and thus easier to monitor than small firms. The decrease was a result of the CEOs aggressively selling their stocks while the boards allowed the option sensitivity to remain the same.
∗ Stern School of Business, New York University. E-mail: hchung2@stern.nyu.edu. I thank my committee, Kose John, Jennifer N. Carpenter, Enrichetta Ravina, and Daniel Wolfenzon, and David Yermack for their helpful comments and support. I also thank Rene Stulz, Craig M. Lewis, and Lemma Senbet , and the participants of FMA Doctoral Consortium for their comments. I am benefited from the discussions with Jaewon Choi, Koren Jo, Jihay Ellie Kwack, Sadi Ozelge, Rik Sen, and Andre de Souza. All errors are mine.
1
I. Introduction
Boards with greater outsider representation are perceived as evidence of good
corporate governance (Weisbach (1988)), as is higher pay-performance sensitivity
(Jensen and Murphy (1990)); Yermack (1996); and Starks and Hartzell (2003)).
However, better board governance does not necessarily lead to higher pay-performance
sensitivity. In a standard principal-agent setting following Holmstrom and Milgrom
(1987) and Heinrich (2000), I show that board monitoring and pay-performance
sensitivity should be substitutes. If better boards can acquire a more precise signal on the
CEO’s effort than stock performance, optimal CEO compensation should rely less on
stock performance and more on the additional signal, lowering pay-performance
sensitivity. Firms compare the risk premium they need to pay their risk-averse CEOs to
keep high pay-performance sensitivity with the monitoring cost to acquire a more precise
signal on CEO effort and determine the optimal level of pay-performance sensitivity and
monitoring. The relative weights placed on pay-performance sensitivity and monitoring
should depend on firm transparency (ease of monitoring) since more transparent firms
could generate more precise signal for a given cost. It is a prediction of my model that if
firms that were relying on pay-performance sensitivity are mandated to strengthen
monitoring, their constrained optimal pay-performance sensitivity would be smaller.
Using the adoption of the new rule in 2002 that regulates the outsider representation of
boards regardless of firm characteristics as an exogenous shock to the board monitoring
intensity, I find the firms, especially the large firms (which are more transparent),
affected by the new rule and improved board governance accordingly lowered their CEO
pay-performance sensitivity relative to the firms that are not affected by the new rule.
This supports the substitution between board monitoring and pay-performance sensitivity
predicted by my model.
To my knowledge, this is the first study to examine how exogenous change in
board independence affects pay-performance sensitivity, which gives a partial answer to a
more fundamental question, whether different governance mechanisms are substitutes or
compliments. The negative relationship between board independence and CEO
ownership, which consists large part of pay-performance sensitivity, are documented in
2
many previous studies. However, most of the studies investigated an alternative
hypothesis arguing reverse causality that CEO ownership determines board independence
(Boone et al. (2007); and Link et al. (2008)) or common factors that determine a firm’s
simultaneous choice of the CEO pay-performance sensitivity and board independence
(Cai et al. (2007); and Gillan et al. (2003)). I use the adoption of the majority board
independence requirement by major exchanges in the United States in 2002 as a natural
experiment to observe how the board independence requirement affects CEO incentives.
Following differences-in-differences approach, this paper finds decrease in CEO pay-
performance sensitivity as a consequence of the new rule requiring listed firms in the
United States to have more than 50% of independent directors.
I constructed a panel of 942 CEO-firm pairs that existed during the periods 2000-
2001 (prior to the adoption of the requirement in 2002) and 2004-2005 (after the
requirement became effective) and compared changes in CEO pay-performance
sensitivity of firms that had less than a majority of outside directors at the fiscal year-end
of either 2000 or 2001 to those that had a majority of outside directors (differences-in-
differences approach, hereafter “diff-in-diff approach”). Diff-in-diff approach controls
for time trend around the adoption of the requirement and any differences in firm
characteristics between members of the two groups that existed before the requirement.
Other factors known to influence CEO incentives such as firm risk, size, R&D intensity,
CEO tenure, board size, and past stock performance are also controlled. I find that the
decrease of CEO pay-performance sensitivity is robust in the case of firms that complied
with the requirement by 2004 fiscal year-end and those that failed to comply by then.
The decrease is more evident, however, for the firms that had not yet complied. These
firms failed to comply with the requirement by 2004 had lower percentage of outside
directors thus were relying less on monitoring prior to 2002 than the firms complied by
2004.
Because CEOs are able to alter their stock holdings through the stock market
(Ofek and Yermack (2000)) while there is no such market for their option holdings, it is
also meaningful to analyze CEO’s stock holdings sensitivity and option holdings
sensitivity separately. I measured CEO pay-performance sensitivity using the Black-
Scholes delta of the CEO’s stock and option holdings portfolio, which represents his
3
firm-related wealth change per $1000 in shareholder wealth change. In the data, CEOs’
stock holdings make up a larger part of overall sensitivity than do option holdings.
Therefore, overall pay-performance sensitivity could drop because of CEOs’ aggressive
stock selling even though boards might be acting against CEOs by increasing option
sensitivity, which boards have greater control over. From the separate analyses of stock
holdings sensitivity and option holdings sensitivity, I observed that the overall decrease
in CEO incentives in firms affected by the board independence requirement is driven by a
significant decrease in the sensitivity of CEO stock holdings, but I found no evidence of
boards acting against the CEOs’ selling. Boards of the affected firms do not increase the
sensitivity of CEO option holdings, allowing the overall sensitivity to drop.
I further investigate whether the increase in board independence, a proxy I use for
monitoring intensity in my model, is related to increase in board monitoring intensity as
my model requires and whether the difference in firm opaqueness had differential effect
of the new rule on pay-performance sensitivity in the case of affected firms as my model
predicts. To answer the first question, I examine the change in the number of board
meetings of the affected firms relative to the control firms, and find the increase in the
number of board meetings of affected firms after the rule adoption relative to the control
group. If the frequency of board meetings is another good proxy for board monitoring
intensity (Vafeas (1999)), this evidence shows that increase in board independence
increased monitoring intensity. I investigate the second question by using firm size as a
measure of firm transparency. I sort my sample by size and examined how the change in
pay-performance sensitivity in the case of affected firms differed by size. The large
affected firms, which are more transparent and thus more cost efficient in monitoring,
showed larger decrease in pay-performance sensitivity than the small affected firms as
my model predicted.
My findings are consistent with previous empirical studies on the relationship
between CEO compensation and board structure and on the effect of rules of corporate
governance introduced around 2002, including the Sarbanes-Oxley Act. I confirm the
negative relationship between board independence and CEO stock ownership or pay-
performance sensitivity in cross-section that has been documented in studies such as
Boone et al. (2007), Linck et al. (2008), and Cai et al. (2007). My research also adds to
4
Chhaochharia and Grinstein (2007), which showed that board independence reduced the
overall level of CEO pay, especially incentive pay. All these findings imply that board
monitoring and CEO incentives are substitutes, as predicted by the theoretical work of
Holmstrom and Milgrom (1987), Heinrich (2000) and Prendergast (2002).
The rest of this paper proceeds as follows. Section II briefly reviews existing
literature on theoretical models and empirical studies pertaining to pay-performance
sensitivity and board effectiveness. Section III derives substitution between board
monitoring and pay-performance sensitivity in a model adapted Holmstrom and Milgrom
(1987) and develops an empirical hypothesis. Section IV describes the data I use in the
analysis and discusses how I construct the treatment group and the control group for diff-
in-diff approach using the board independence requirement adopted by major U.S.
exchanges in 2002. Section V documents the analysis and results. Section VI concludes.
II. Literature Review
I derive substitution between board monitoring and CEO pay-performance
sensitivity by adapting Holmstrom and Milgrom (1987)’s model. Holmstrom and
Milgrom (1987) derive the optimal compensation scheme in a principal-agent setting in
which the agent has exponential utility and controls the mean of a multivariate normal
distribution but not the variance. The principal can observe and contract only on the
outcome (stock performance) of the agent’s action. In their setting, a linear
compensation contract that varies with the observed outcome is optimal, and sensitivity
declines as the variance of outcome increases. Heinrich (2000) extends Holmstrom and
Milgrom’s (1987) model and defines monitoring as acquiring an additional signal on
CEO’s effort that is uncorrelated with the disturbance of the original signal. The optimal
compensation is a linear combination of the two signals, and the sensitivity on the
original signal, stock performance, decreases with the precision of the additional signal.
In this model, better monitoring generates the signal of higher precision, delivering
substitution between CEO pay-performance sensitivity and monitoring.
In a variation of Holmstrom and Milgrom’s (1987) model, Prendergast (1999,
2002) also argues the substitution between direct monitoring on CEO action and pay-
5
performance sensitivity. In addition, he provides a survey of empirical studies regarding
monitoring, trade-off between risk and incentive, and relative performance evaluation.
He brings issues of multitasking and the availability of direct monitoring of inputs instead
of outputs, and shows that if the principal can monitor inputs directly, optimal pay-
performance sensitivity of the risk-neutral agent is lower, even if the output signal is not
noisy. In contrast, if direct monitoring of inputs is impossible and the decisions are
delegated to agent’s discretion, incentive pay based on outcome is higher, even if the
output signal is noisy. This results in a positive relationship between risk and incentive,
unless the availability of direct monitoring is correlated with multitasking. Prendergast’s
theory predicts that if independent boards enhance the monitoring of CEO inputs, CEO
pay-performance sensitivity will decrease, which is consistent with my empirical
findings.
My study also adds to the empirical literature that examines the link between
board structure and CEO compensation or CEO ownership. Core, Holthausen, and
Larcker (1999) examine the effects of various measures of board governance, including
board independence, on compensation level. After controlling for performance and firm
size, they find excessive compensation in firms with weak board governance.
Chhaochharia and Grinstein (2006) analyze how the adoption by major U.S. exchanges in
2002 of the new rule that promotes board independence affected the level of total
compensation and its components for listed U.S. firms. They find the requirement
significantly decreased the compensation levels (especially the levels of the option
component) of the firms that had not met the requirements before the rule change but did
meet it afterward. They show that among the new requirements, independence of a
majority of board members had the most significant impact.
Those studies relate board independence to the level of compensation but not to
the sensitivity of compensation. Yermack (1996) examined the relationship between
board size and pay-performance sensitivity and found that larger boards had lower pay-
performance sensitivity. In their firm-level analysis, Gillan, Hartzell, and Starks (2003)
used a board index consisting of various aspects of board structure, including board
independence; they found a significant positive relationship between the proportion of
incentive pay and board index. Cai, Liu, and Qian (2007) focused on the effect of
6
information asymmetry between insiders and outsiders of a firm, which is measured by
size, R&D expenditures, Tobin’s Q, the number of analysts following the firm and their
earnings forecast errors, and the number of shareholders. They found lower board index,
higher pay-performance sensitivity, and less anti-takeover provisions (Reverse G index)
when there is more information asymmetry. Coles, Lemmon, and Yang (2008) model the
marginal productivity of physical assets and human capital of CEO and outside directors
and derive firm’s optimal choice of CEO pay-performance sensitivity and board
independence and their contribution to firm productivity. The firm’s optimal choice
shows negative relationship between CEO pay-performance sensitivity and board
independence. Boone, Field, Karpoff, and Raheja (2007) and Linck, Netter, and Yang
(2008) studied the determinants of board structure and found board independence and
CEO ownership had a negative relationship, although the relationship weakened after the
enactment of the Sarbanes-Oxley Act.
III. The Model and Hypothesis
Principal-Agent Problem
The substitution between CEO pay-performance sensitivity and board monitoring
can be derived in a standard principal-agent setting like those found in Holmstrom and
Milgrom (1987) and Heinrich (2000). I assume risk-neutral shareholders as principal and
a risk-averse CEO with constant absolute risk aversion as agent. A CEO chooses effort
at a cost of C e , which yields a profit π e, ε that is the sum of CEO’s effort and a
random shock following normal distribution: π e ε, ε~ N 0, σ . Without board
monitoring, shareholders could observe only the profit and contract to pay to CEO
based on but not on e. Shareholders set wage contract that induces desired level of
CEO effort e maximizing the payoff π w.
The CEO chooses effort e that maximizes an exponential utility, V 1
exp r w C e , given the compensation contract , where is the constant
coefficient of absolute risk aversion. The reservation utility for the CEO is assumed to be
0. The cost of the CEO’s effort C e is quadratic: e . Holmstrom and
7
Milgorm(1987) showed that under these assumptions, a linear contract is optimal and the
optimal compensation contract is as follows:
(1) w β β π
where the pay-performance sensitivity is β 1/ 1 rσ ,
and the fixed salary is β 1 rσ /2 1 rσ ,
with the CEO effort of e β 1/ 1 rσ .
Monitoring
Assume that shareholders can set a board with monitoring ability and acquire
another signal on CEO’s effort such that s e u, u~N 0, σ where the precision
σ and is uncorrelated with . The constant represents the ease of monitoring
that could vary with firm characteristics. The degree of information asymmetry between
CEO and shareholders, or opaqueness of firms, is one of the firm characteristics closely
related to the easiness of monitoring (Demsetz and Lehn (1985); Gillan, Hartzell and
Starks (2003); Cai, Liu, and Qian (2007)). Therefore, I denote the opaqueness
coefficient hereafter. Assume that the cost of monitoring ) increases in the
monitoring intensity , being . The higher the , the less the precision σ is
available for a given monitoring intensity and its cost ). Shareholders need to set
up a board with higher monitoring intensity if the firm is more opaque (higher ) or
they desire more precise signal (smaller σ ).
When monitoring is available, shareholders choose monitoring intensity in
addition to a CEO compensation contract w to maximize their payoff π
w M. The board with monitoring intensity costs and acquires an
additional signal of precision . Given the monitoring intensity and the
compensation contract w , the CEO chooses his effort level that maximizes his
utility .
8
Optimal Solution—CEO Pay-Performance Sensitivity and Monitoring Intensity
If shareholders could access an additional signal on the CEO effort of precision
σ , the optimal compensation contract between CEO and shareholders given the
precision σ is a linear contract depending on both and s (see Heinrich(2000)):
(2) w β β π β s
where the pay-performance sensitivity is β ,
the weight on the additional signal is β ,
and the fixed salary is β ,
with the CEO effort of e β β .
The pay-performance sensitivity is therefore, β
β , i.e., given the firm risk σ , the pay-performance sensitivity is lower when
some monitoring takes place, and shareholders can save the risk premium paid to the
CEO. In addition, shareholders benefit from increased expected profit because the
CEO effort e in (2) is larger than the no-monitoring CEO effort e in (1).
Given the benefit and the cost of monitoring, monitoring is optimal—that is,
shareholders benefit from collecting additional signal —if and only if
(3) .
Therefore, monitoring takes place when the opaqueness coefficient is not too
high that acquiring additional signal is too costly and the firm risk σ is not too low that
the benefit of additional signal is negligible.
If the condition (3) is met, the optimal level of monitoring is
9
(4) σ √ .
The optimal precision σ is increasing in the opaqueness coefficient and
decreasing in the firm risk σ . The more expensive the marginal precision of the
additional signal is (the higher is), the less precision σ shareholders can afford. As
the benefit of precise monitoring is higher for high-risk firms, shareholders of the firms
with high σ are willing to bear the greater cost of obtaining a higher precision signal.
Collecting the signal of the optimal precision, ~ , σ , requires shareholders to set a
board with the optimal monitoring intensity .
Hypothesis Development
I assume that the monitoring intensity of boards depends on board
independence (Weisbach (1988); Hermalin and Weisbach (1988); Coles, Lemmon, and
Wang (2008)). Shareholders can set up a board that consists of % of outside directors
that has monitoring intensity of at a cost of . Specifically, let
be a continuous and monotonic increasing function of the percentage of
independent directors, . Given the opaqueness coefficient m, the bounds of
determine the bounds of precision σ . Precision σ is continuous and monotonically
decreasing in the percentage of independent directors . If boards are unable to generate
a useful signal that reduces the noisiness of stock performance signal, it could be
interpreted as σ being close to infinity. Suppose that this is the case if boards are
insider-dominated, having no more than 50% of outside directors while boards with more
than 50% of outsider representation generate a signal effectively reduces the noisiness
of the stock performance signal.
In cross-section, the opaqueness coefficient varies with firm characteristics.
Therefore, given firm risk σ , a firm may or may not monitor, i.e., the condition (3) is
met or not, depending on its value of . If monitoring is optimal and the firm
accordingly hires n (50, 100]% of outside directors, the firm’s CEO pay-performance
10
sensitivity is β of the compensation contract (2), which is lower than β of the
compensation contract (1), which is the CEO pay-performance sensitivity of the firms
with the same firm risk σ that do not generate precise enough signal because of too
high. This leads to the following cross-sectional prediction:
(H1) For given firm risk , the outsider-dominated boards that observe and contract on
the additional signal have lower CEO pay-performance sensitivity than do the insider-
dominated boards.
Suppose there is a regulation that requires that more than 50% of board members
be outside directors and that this imposes a cost to firms of at least 50 . In return,
firms now have monitoring intensity of 50 . If the payoff to the shareholders
50 is nonnegative, those firms optimally had the percentage of
independent directors 50% and had not collected the additional signal prior to the
regulation will start to collect the additional signal of precision σ 50 .
After board monitoring is in place, the compensation contract will change from of (1)
to of (2), and the pay-performance sensitivity will decrease from β 1/ 1
rσ to β 50 . Accordingly:
(H2) When a regulation requiring more than 50% of outside directors on boards is
enforced, the CEO pay-performance sensitivity of the firms that had not met the
requirement prior to the regulation decreases.
Once monitoring is enforced, the difference in the opaqueness coefficient
among firms that had not monitored prior to the regulation results in the difference in
precision they can achieve, σ 50 , per the monitoring intensity 50 .
Ceteris paribus, the firms with low that are able to monitor more efficiently and
acquire more precise signals will decrease CEO pay-performance sensitivity more than
the firms with high . Firm size is known to be a good proxy for the opaqueness
coefficient —large firms whose business are more established, transparent, and under
11
scrutiny of market and regulators have lower than do small firms (Diamond and
Verrecchia (1991); Harris (1994); and Cai, Liu, and Qian (2007)). Therefore, empirically
I predict:
(H3) For given firm risk , large firms, which more efficiently monitor and generate
more precise 50 ; , will lower CEO pay-performance sensitivity more after the
majority independent directors regulation than small firms, which less efficiently monitor
and have less precise 50 ; .
An isomorphic signaling equilibrium where there is information asymmetry
between the CEO and the market like Leland and Pyle (1977) results in similar empirical
predictions as the principal-agent model I derived above. If a risk-averse CEO has
private information about the firm’s value V that cannot be transferred directly to the
market, the CEO can signal the firm’s prospects to the market by tying his wealth to the
firm value, by maintaining high pay-performance sensitivity. The signaling is possible
through a cost, the high risk-premium paid to the CEO. Alternatively, firms can set up at
some cost an independent board that can verify the CEO’s private information about V
and disclose it to the market. If the work of the independent board reduces the
information difference between the CEO and the market, the CEO does not need to keep
high pay-performance sensitivity to signal the market. The CEO is allowed to have low
pay-performance sensitivity and receive a lower risk premium. Therefore, like (H1),
firms with boards dominated by outsiders should have lower pay-performance sensitivity
than firms with boards dominated by insiders.
When there is no restriction on a firm's choice of board structure, firms choosing a
way to relieve the information asymmetry will compare the cost of high CEO pay-
performance sensitivity to the cost of an independent board and choose the mechanism
that costs less. If a regulation requiring a majority of outside directors is enforced,
however, the firms that had chosen to rely on high CEO pay-performance sensitivity
rather than on a board dominated by outsiders will reduce their CEO’s pay-performance
sensitivity as the now outsider-dominated board narrows the information gap between the
CEO and the market ((H2)). The effectiveness of outsider-dominated boards in reducing
12
information differences may vary with firm characteristics. Using firm size as a proxy
for this effectiveness, among the firms that changed their board composition to comply
with the regulation, I expect large firms decrease CEO pay-performance more than the
small firms do ((H3)).
IV. Data
My study uses accounting and CEO compensation data from ExecuComp, which
covers firms in the S&P 1500 index. The sample size reduces, though, to around 1,350
firms per year for which board data from Thomson Financials (formerly, the Investor
Responsibility Research Center) are also available. To acquire additional accounting data
beyond the coverage of ExecuComp and for stock market data, respectively, I use
COMPUSTAT and CRSP data. To investigate the effect of the adoption of the majority
independent directors rule in 2002, I choose firms that existed in 2000-2001 period and
that survived at least until 2004. This survived sample, which is the focus of main
analysis, produces 942 CEO-firm pairs and 3,388 observations during the sample period
2000-2005.
I define the overall CEO pay-performance sensitivity (PPS), the dependent
variable, as the sensitivity of a CEO’s stock and option holdings to $1,000 shareholder
wealth change. It is the weighted sum of the delta of stock holdings, which is 1, and the
delta of option holdings by Black-Scholes formula. Each is weighted by the proportion
of the number of each holding in thousands to the number of shares outstanding in
millions. This measure captures the year-to-year variation of ex ante sensitivity of the
CEO’s firm-tied wealth to shareholder wealth change. The sensitivity from CEO’s stock
and option holdings accounts for more than 90% of CEO wealth sensitivity to
shareholder wealth change (Jensen and Murphy (1990); Aggarwal and Samwick (1999)).
Other forms of a CEO’s firm-related wealth (such as salary and bonus), however, are also
sensitive to shareholder wealth change.
Because the delta of stock holdings to stock price is 1 by definition, I need only
the number of CEO’s stock holdings excluding options and the number of shares
outstanding to derive the sensitivity from CEO stock holdings. The sensitivity, the
13
change of CEO stock holding values per $1,000 shareholder wealth change, is the former
in thousands divided by the latter in millions. The sensitivity from CEO option holdings,
however, involves more variables than just the number of options because the Black-
Scholes delta of an option to stock price is affected by stock price, exercise price, risk-
free rate, dividend yield, stock volatility, and time to maturity at the time of evaluation.
The number of grants, stock price at the fiscal year-end, exercise price, and time
to maturity of the current year’s grants are available in ExecuComp. Because
ExecuComp does not provide this information on existing options, however, it is
approximated using the method suggested by Core and Guay (2002). Core and Guay
show the delta using the approximation method lies within the 1% range of the exact
delta. The seven-year U.S. treasury bond interest rate, prior three-year average dividend
yield, and 60-months prior historical stock volatility are used as estimates for the risk-free
rate, dividend yield, and stock volatility, respectively.
I also separately analyzed the effect of change in board composition on a CEOs’
option holdings sensitivity (OptionPPS) and stock holdings sensitivity (StockPPS). PPS
represents the CEO’s total incentives; boards effectively control option pay-performance
sensitivity and have less control over stock pay-performance sensitivity. Most of the
CEOs’ stock holdings, which constitute a large part of PPS, are parts of their personal
disposable wealth which they can buy or sell through the market. On the other hand,
CEOs’ option holdings are not as tradable as their stocks nor are they easy to hedge once
granted by boards. Moreover, boards can control the OptionPPS not only by adjusting
the number of option holdings through new grants but also by adjusting the strike price of
existing holdings. Boards can lower the strike price of deep out-of-the-money options
that provide little incentives to CEOs and thereby boost OptionPPS.
Accordingly, it is important to see whether StockPPS and OptionPPS move in the
same direction. If they do not, it is evidence that boards are working against their CEOs.
I can further examine how effective boards are in unwinding CEOs’ action by comparing
the magnitudes of change in OptionPPS and StockPPS.
Board independence, which is the object of the new rule, is defined as the
percentage of a board’s directors who are independent. Control variables that affect PPS
include firm risk, firm size, R&D intensity, CEO tenure, board size, and past stock
14
performance. The theoretical work by Holmstrom and Milgrom (1987) expects a
negative relationship between firm risk and PPS. Empirical works such as Aggarwal and
Samwick (1999) and Jin (2002) confirm this relationship. Following their works, I
include the annual rank percentile of the product of the previous 60-month stock return
volatility and the market value at the beginning of the estimation period (rank of dollar
return variance) to control the firm risk.
Finding a negative relationship between firm size and PPS, Demsetz and Lehn
(1985) and many other empirical studies argue that firm size is a proxy for firm
complexity, firm risk, or growth opportunity. I use log of sales (ln(Sales)) to measure
firm size. R&D expenditure divided by assets (R&D/Assets) serves as a proxy for
growth opportunity or for information asymmetry between firm insiders and outsiders.
For either reason, I expect high R&D/Assets firms to have high PPS. I also include a
dummy (MissingR&D) that has value 1 if the data on R&D/Assets is not available and 0
otherwise.
CEOs tend to accumulate stock and options during their service. I, therefore,
include the number of years served as a CEO (CEO tenure) to control for such a positive
relationship. Yermack (1990) and other studies document that board size (the number of
directors on a board) is negatively related to PPS; accordingly, I control board size.
CEOs are contrarians who sell their stocks when the prices rise above the price at the
time they are awarded and hold otherwise (Seyhun (1986); and Lakonishok and Lee
(2001)); consequently, I include the three-year total returns to shareholders to control for
this disposition effect of CEOs which causes a negative relationship between PPS
(especially StockPPS) and past stock performance.
NYSE, NASDAQ, and AMEX required a majority of independent directors on
corporate boards after the Enron scandal in 2002. Listed firms had to comply with this
rule by the end of 2004 (by the end of 2005 for firms with staggered boards), and they
started to adjust their boards in 2003 (Chhaochharia and Grinstein (2006)). Among those
firms that existed during 2000-2001 period and survived until 2004, I define the treatment
group as the group of firms that had less than 50% of independent directors in either 2000
or 2001 and therefore were affected by the new rule. The control group consists of the
firms that were not affected by the rule as well as related rules regulating committee
15
independence. The new rules regarding committees that were adopted along with the
board majority independence rule require that 1) the compensation committee, the
nominating committee, and the audit committee must consist of 100% of independent
directors and 2) the audit committee must have more than three members. Members of
the control group must be clean of both requirements because firms whose boards
included a majority of independent directors but that did not meet the committee
requirements before 2002 may hire additional independent directors to fulfill these
requirements. Therefore, I exclude from the control group firms that had outsider-
dominated boards prior to 2002 but did not meet the committee independence
requirements.
Table 1 reports the summary statistics of the treatment group and the control
group. To see whether survival bias affects my sample, in an unreported table I compare
the characteristics of all the firms in 1996-2005 that had no more than 50% of
independent directors to my treatment group. I also compare the characteristics of all the
firms in 1996-2005 that had more than 50% of independent directors (regardless of
committee independence) to my control group. The treatment group and the control
group showed similar characteristics as their counterparts of the entire sample, suggesting
little influence of survival bias.
The treatment group has much higher PPS (mean 53.07 and median 19.59) than
the control group (mean 23.08 and median 11.12) due to the difference in StockPPS. The
average percentage of independent directors of the treatment group is 52.01%, and that of
the control group is 77.66%. The treatment group is of similar firm size (market value
and sales) and firm risk (dollar return variance) as the control group on average but
smaller on median; the treatment group has greater variation. The treatment group
showed greater variation in total direct compensation to CEOs (total compensation
including salary, bonus, stock and option grants, long-term incentive plan payouts, and
other compensation both annual and one-time payments) than the control group; the
average total direct compensation was similar for both groups.
The treatment group experienced higher total returns (mean 10.47%) in the past
three years than the control group (mean 8.53%). It has slightly lower R&D/Assets,
which is not observed for 52% of the treatment group and is not observed for the 40% of
16
the control group. The average CEO tenure is longer (mean 9.99 years) in the treatment
group than in the control group (mean 6.91 years), suggesting weaker monitoring of
insider-dominated boards. Board size and CEO age show little difference between the
two groups.
I examine the treatment group further to determine whether the firms that
complied with the rule by 2004 are different from the firms that failed to comply by 2004
(not reported). Compared to the compliant firms in the treatment group, the firms that
failed to comply with the rule by 2004 have higher PPS due to higher StockPPS while
having lower percentages of independent board directors. They are smaller and more
diverse in terms of sales and dollar risk, and have slightly lower R&D/Assets. The total
direct CEO compensation is also smaller. Their boards have, on average, one more
director.
V. Empirical Analysis and Results
A. Cross-Sectional Analysis
Before investigating the effect of adopting the majority independence rule in 2002, I
analyze the relationship between PPS and board independence in cross-section ((H1)). I
use the entire sample from 1996-2005 for this analysis. I regress
(5) %
I control for the other variables that affect PPS, such as firm’s dollar risk, ln(Sales),
R&D/Assets, dummy for missing R&D intensity, CEO tenure, and board size, and three-
year total returns to shareholders. Column (1) in Table 2 shows the analysis with CEO-
firm pair fixed effects and year dummies. Column (2) includes industry dummies based
on the first two digits of SIC code and year dummies.
Using CEO-firm pair fixed effects is more appropriate than using CEO fixed effect or
firm fixed effect (alone or combined) because PPS depends not only on CEO
compensation but also on CEO’s personal wealth invested in a firm. For example, insider
17
CEOs accumulate their respective firm’s stocks and options before becoming CEOs
while outsider CEOs have much lower ownerships of their respective firms when they
start working as CEOs. If an insider CEO switches firms to become an outsider CEO,
CEO fixed effects do not reflect the difference between the CEO’s ownership of the
previous firm and the new firm. If a firm replaces an insider CEO with an outsider CEO,
firm fixed effects do not reflect the difference between the firm ownership of the
incumbent insider CEO and the firm ownership of the new outsider CEO. Unless each
firm always hires insiders or always hires outsiders and each CEO always serves,
respectively, as an insider CEO or an outsider CEO, using both the CEO fixed effect and
firm fixed effect only partially solves the problem. In contrast, CEO-firm pair fixed
effects allow each CEO to have different ownership in different firms he serves and
allows each firm to hire CEOs who have different ownership interests in the firm.
Analyzing the deviation from the CEO-firm pair average, the CEO-firm pair fixed effect
model takes care of the different average PPS of the incumbent CEO and the new CEO
when there is a CEO turnover.
As predicted in (H1), a negative and significant relationship between PPS and board
independence is observed in both columns. This result is consistent with the empirical
findings of Linck et al. (2008) and Cai et al. (2007). The 10 percent point increase in
board independence is related to a PPS decrease of $0.93 when CEO-firm pairs fixed
effects are controlled and a PPS decrease of $4.78 per $1,000 shareholder wealth change
when only industry is controlled.
The StockPPS has a significantly negative relationship with board independence
while OptionPPS shows mixed evidence depending on controls. The 10 percent point
increase in board independence is associated with a $0.55 decrease in StockPPS if CEO-
firm pair fixed effect is included (Column (3)) and with a $5.21 decrease when industry is
controlled (Column (4)). After controlling for the CEO-firm pair fixed effects, a 10
percent point increase of board independence decreases OptionPPS by $0.38 per $1,000
shareholder wealth change (Column (5)), which is about 5% of average OptionPPS.
OptionPPS increases by $0.43 per $1,000 shareholder wealth change (Column (6)),
however, as board independence increases by 10 percent point when only industry is
controlled.
18
The rank of dollar return variance and ln(Sales) have negative coefficients, which is
as expected. R&D/Assets is negative and significant for PPS and StockPPS but positive
and significant for OptionPPS when controlling only industry fixed effects but not CEO-
firm fixed effects. All the negative effects of R&D/Assets on PPS and StockPPS
disappear after controlling for the CEO-firm fixed effects. The coefficient on CEO
tenure is positive and significant while board size is negatively related. The coefficient
on three-year total returns to shareholders is seldom significant and does not support the
disposition effect hypothesis.
B. Time Series of Board Independence and CEO Pay-Performance Sensitivity
Before analyzing the effect of the adoption of the majority independence rule in 2002
on CEO incentives, I report the time series trend of board independence and CEO
incentives in Table 3. Panel A shows that the average board independence of the control
group stayed around 78% of independent directors during 2000-2005 as required, while
that of the treatment group increased significantly from less than 46% before 2002 to
more than 59% after 2003.
The average and median of PPS, StockPPS, and OptionPPS of the control group
remain at the same level before the adoption of the new rule in 2002 and after the rule
became effective in 2004, peaking in the transition period of 2002-2003. In contrast,
when comparing pre-2002 figures with post-2003 figures, the average PPS and StockPPS
of the treatment group decrease by one-third and the median by a half. An examination
of the raw data shows that these decreases in PPS and StockPPS are caused by the CEOs
aggressively selling their stocks rather than by increased numbers of outstanding shares.
It is also worth noting that the PPS and StockPPS standard deviations of the treatment
group are twice as large as the control group’s. Nevertheless, the treatment group’s
OptionPPS shows similar time-series patterns and magnitude as the control group’s, with
slightly higher standard deviation.
Figure 1 plots the annual median PPS, StockPPS, and OptionPPS of the treatment and
control groups. Panel A shows that as the percentage of independent directors increased
due to the new rule, the median PPS of the treatment group decreased significantly;
19
meanwhile, the control group that did not experience any change in board structure shows
little change in the median PPS. Panel B shows that the StockPPS drives the different
PPS behaviors of the two groups. The OptionPPS shows little difference between the
two groups, suggesting that the treatment group boards are not acting against their CEOs,
who are effectively lowering their PPS by selling their stocks.
Though not controlled for other control variables, this simple comparison of medians
supports (H2) which holds that, for firms affected by the new requirement, the forced
improvement in board independence leads to lower CEO pay-performance sensitivity.
This is because enhanced board monitoring substitutes for the CEO incentives provided
by high CEO pay-performance sensitivity.
Figure 2 presents the distribution of PPS of the year 2001, the last year prior to
the rule change, and the year 2004, the deadline for compliance, of both the treatment
group and control group. PPS has highly skewed distribution in both the groups. The
PPS of the control group shows little difference between year 2001 and year 2004 while
the distribution of PPS of the treatment group shifts toward left becoming much denser at
the bottom. Figure 2 implies the decrease in PPS of the firms affected by the majority
independence rule is unlikely to be driven by outliers but by general decrease in the
group.
I further analyze the difference in PPS between year 2001 and year 2004 to see
how much of the decrease is due to CEO turnover. If a long-served CEO with high
ownership is replaced with an outsider CEO with little ownership, PPS drops
significantly as a consequence. Among the CEOs served in 2001, 25% were replaced
with new CEOs by 2004 in the treatment group and 30% were replaced in the control
group, which are usual rate of CEO turnover. The CEOs who served the treatment group
in 2001 but not in 2004 had average PPS of 43.83 and median of PPS 18.40, which are
below those of the entire treatment group. Their successors in 2004 had average PPS of
19.97 and the median of 12.93. The difference in average explains 60% of the 9.88
decrease in average of the entire sample and this difference in average PPS is driven
mostly by the difference in average StockPPS, which was 36.87 in 2001 and 7.68 in
2004. The 75% of CEOs who continued to serve the treatment group lowered PPS by
5.19 on average, which is about 9% of 2001 average. The CEO turnovers in control
20
group also caused decrease in PPS (from average 14.84 and median 6.78 in 2001 to
average 7.70 and median 2.09 in 2004) that explains 70% of the change in average PPS in
the entire sample, but the difference is smaller and StockPPS and OptionPPS equally
contribute to the decrease.
C. Diff-in-Diff Analysis
In this subsection, I apply the diff-in-diff approach to analyze the effect of the new
rule regulating board independence on the CEO pay-performance sensitivity. The
adoption of the majority independence rule in 2002, an exogenous shock in board
structure that applies regardless of firm characteristics, provides a unique opportunity to
analyze how this exogenous shock changes CEO pay-performance sensitivity.
Among the firms that existed both before and after the rule change, I analyze how the
new rule changed CEO incentives of firms that the rule affected; I compare the change in
CEO incentives of these firms to the change in CEO incentives of the control group
(firms unaffected by the rules regarding majority board independence or committee
independence). Taking into account changes in CEO incentives among the control group
removes the effect of common factors that affected both groups around the rule change. I
also account for the variation in other control variables over time. As in the cross-
sectional analysis, rank of dollar return variance, ln(Sales), R&D/Assets, MissingR&D,
CEO tenure, board size, three-year total returns to shareholders, as well as year and either
CEO-firm fixed effects or industry dummies are controlled. Formally, I estimate the
following specification:
(6) 04_05 50% 02
if estimating the CEO-firm fixed effects model that takes care of differences in average
PPS caused by CEO turnover, and the following specification:
21
(7) 04_05 50% 02
50% 02
if only industry dummies are controlled. If one of the purposes of boards in initiating
CEO turnover is to reduce PPS, this specification, which does not control for the effect of
CEO turnover on PPS, is the right measure with which to infer boards’ intentions.
04_05 50% 02 measures the
effect of the adoption of the majority board independence rule. It is an interaction
between 04_05 , a dummy that has value 1 for year 2004 and 2005 and 0
otherwise, and 50% 02 , a dummy that has value 1
for the treatment group firms that had no more than 50% of outside directors prior to the
rule adoption in 2002 and 0 otherwise. 50% 02 is
included in the analysis to capture the difference that remains unexplained (even after
accounting for the variation in other control variables) between the treatment group and
the control group prior to the rule adoption. controls for the time trend
in PPS around the adoption of the rule.
Column (1) of Table 4 reports the estimates of Specification (6), and column (2)
estimates Specification (7). In both columns, the coefficients on 04_05
50% 02 are negative and significant as predicted
in (H2) and shown in the simple comparison of median PPSs before and after the rule
adoption (Section IV.B.). The PPS of the firms affected by the new rule decreased by
$4.74 (almost 10% of the average PPS of treatment group in year 2001) per $1,000
shareholder wealth change, if CEO-firm pair fixed effect and time effect is controlled. If
only industry and time are controlled (and CEO turnover effect is not corrected), the drop
is more severe: the PPS of the treatment group decreased by 8.12, about 14% of the 2001
average PPS of the treatment group.
Consistent with prior empirical studies on PPS, rank of dollar return variance and
ln(sales) are negatively related to PPS. CEO tenure and PPS have a positive relationship,
while board size and PPS have a negative relationship. R&D/Assets and three-year total
returns to shareholders show no significant relationship with PPS. The t-statics of the
22
OLS regressions are reported in the parentheses. The t-statics of Column (2) are based on
robust standard errors. The results are robust to the standard errors clustered by firm or
industry.
I compared the change in PPS of all the firms that adoption of the new rule
affected (regardless of their compliance) to the control group. These firms must
eventually meet the majority independence rule; in expectation of the change in board
independence, they may have started adjusting PPS appropriate to the new board
structure before actually complying with the new rule. To examine the different effects
of board independence on PPS of firms that actually complied with the rule by 2004 and
firms that are expected to comply but had not done so by 2004, I assign a dummy for
each group of firms and compare the change in PPS with the control group following
diff-in-diff approach as in specifications (6) and (7). I find that, regardless of actual
compliance, all firms experience decreased PPS, and most of the decrease is due to CEO
turnover. Estimated by the CEO-firm fixed effects model, firms that failed to comply
with the rule by 2004 had a larger decrease in PPS (6.04) than the firms that complied by
2004 (4.21). These numbers are not reported in the tables.
To ensure that the results are not driven by outliers, I run quantile regressions and
present the estimates in columns (3)-(5) of Table 4. Because PPS data have high standard
deviations and skewness, it is possible that outliers with very high PPS drive the result
and mislead the behavior of the entire sample. I run median, q25, and q75 regressions
(which minimize the according weighted sum of absolute residuals and are thus more
robust to outliers than are ordinary least square regressions) with industry and time
dummies; I find significant decreases in PPS consistent with OLS regression results
throughout the quartiles. In the median regression (Column (3)), the majority
independence rule decreases PPS by an estimated $1.80 per $1,000 shareholder wealth
change; this accounts for about 8% of the 2001 median PPS of the treatment group firms,
when firms affected by the new rule are compared to the control group. The q25
regression estimates the PPS decrease of 1.35 (Column (4)) and the q75 regression of
8.48 (Column (5)).
Because boards have more control over OptionPPS than StockPPS, I investigate
how StockPPS and OptionPPS of affected firms responded to the new rule. From the
23
analysis of only PPS, it is unclear whether CEOs, boards, or both together are causing the
result. If StockPPS and OptionPPS move in different directions, boards are acting
against the decisions of CEOs, and the change in PPS depends on the relative magnitude
of the changes in StockPPS and OptionPPS. If StockPPS and OptionPPS movements are
coordinated, the separate analysis of each sheds light on how much each component
contributes to the overall change in PPS. I conduct the same analysis of specifications
(6)-(7) but with different dependent variables, StockPPS and OptionPPS. Table 5 reports
the results of the analysis on StockPPS. The t-statics of the OLS regressions are reported
in the parentheses. The t-statics of Column (2) are based on robust standard errors. The
results are robust to the standard errors clustered by firm or industry.
Columns (1) and (2), report decreases in StockPPS of the treatment group firms in
similar magnitudes to the decreases in PPS reported in Table 4. The variable
04_05 50% 02 estimates -4.21, about
8.5% less than the 2001 average StockPPS, when CEO-firm pair fixed effects are
controlled (Column (1)), and -8.53, about 17% less, when only industry is controlled
(Column (2)). These results imply that the decrease in StockPPS, or in CEO stock
ownership, causes the decrease in PPS.
Other variables are also estimated to be very similar to the PPS analysis, but the
coefficient on three-year total returns to shareholders is now significant at the 10% level
(Column (2)), indicating the disposition effect in CEOs’ stock selling behavior. The
analysis with the different dummies for the firms that complied with the rule by 2004 and
those that failed to comply also yielded very similar estimates to the analysis on PPS.
I run median, q25, and q75 regressions on StockPPS and the results are similar to
the analysis on PPS for median and q75 regressions, but in q25 regression the decrease of
StockPPS of the treatment group is estimated to be smaller than that of PPS. Columns
(3)-(5) document the estimates of these robust regressions. The decrease of the treatment
group’s StockPPS is significant at the 10% level for q25 and median regression, and at
the 1% level for q75 regression.
In an unreported table, the same sets of analysis on OptionPPS show no
significant difference in OptionPPS of the treatment group before and after the rule
relative to the control group. The coefficients on the main variable 04_05
24
50% 02 are not significant, and the magnitude is
only 0.05. With the exception of R&D/Assets, other variables behave the same as in the
other analyses. R&D/Assets was not significant in the other analyses, but it is positively
related to OptionPPS if industry is controlled for. The effect disappears when CEO-firm
pair fixed effects are also controlled; this suggests that R&D intensity explains the cross-
sectional variation of the OptionPPS, which is part of persistent firm characteristics.
Though not significant, the coefficients on three-year total return to shareholders are
positive and similar in magnitude to those observed in the StockPPS analysis; this
suggests that boards increase OptionPPS to counter the decrease in StockPPS due to the
mechanical stock selling of CEOs when stock price appreciates.
D. Discussion
Board Meeting Frequency
One of the proxies that measure the intensity of board monitoring is the frequency
of board meetings (Vafeas (1999)). If there is little difference among boards in the
amount of work they perform each meeting, the number of meetings is a measure of the
amount of work a board completed in a given year. Assuming that the increase in
percentage of outside directors itself does not change boards’ work capacity per meeting
(a reasonable assumption based on the findings of Vafeas (1999)) I expect the boards
affected by the new rule met more frequently after the rule became effective as they
engage in more monitoring. A simple comparison of the average frequency of board
meetings supports this prediction. The frequency of board meetings of affected firms
increased from 6.42 times in 2000 to 7.71 times in 2005. In comparison, the meeting
frequency of the control group was higher and stable: 7.41 times in 2000 and 7.95 times
in 2005.
More formally, I examine the change in board meeting frequency of the treatment
group using diff-in-diff approach. In addition to the variables in the PPS analysis, I
include a dummy for the CEO’s last year of service, and a dummy for the CEO’s first
year of service to control for any increase in the number of board meetings due to CEO
25
turnover. Three-year total returns to shareholders now control for unfavorable events in
the past. Column (1) of Table 6 shows that, estimated by a CEO-firm pair fixed effects
model, the boards of firms of the treatment group meet 0.40 times more than boards of
the control group after the majority board independence rule became effective. The
increased frequency is larger, 0.58 times more per year, if only industry dummies are
included (Column (2)).
Firm Size and Monitoring Intensity
To test the implications of (H3), I divide the samples by firm size and compare
the change in PPS of large treatment group firms (upper 30 percentile) to large control
group firms and of small treatment group firms (bottom 30 percentile) to small control
group firms. Columns (3)-(6) of Table 6 document the results. As predicted in (H3), I
find the effect of the new rule on PPS differs among firms with different transparencies.
The PPS of the large affected firms decrease significantly (by 6.26 in the CEO-firm pair
fixed effects model and by 10.96 in the industry fixed effects model) after the adoption of
the rule relative to the large control firms. Meanwhile, the PPS decrease of small firms is
smaller and not significant. The large firms, which are more transparent and able to
generate more precise signal on CEO effort, can lower PPS more than the small firms,
which lack transparency.
Placebo event in 2001
My diff-in-diff approach is comparing 2000-2003 sample to 2004-2005, which is
more conservative than comparing 2000-2001 to 2004-2005, assuming most firms met
the deadline at the last minute in 2004. The affected firms started to increase board
independence after 2002 in preparation to the compliance deadline of 2004, and I observe
the decrease in PPS from 2002 accordingly. If I exclude year 2002 and 2003 when the
rule was made but before the compliance deadline, I obtain stronger results because
the gradual decrease in PPS and StockPPS after 2002 works against finding any
difference before and after the rule change.
26
To test whether my result is driven by other event that proceeds the rule change in
2002, I take the same diff-in-diff approach and compare the change in PPS and StockPPS
of the affected firms to the change of the control firms but as if the rule became effective
in 2001. The affected firms did not change and were not expected to change their board
independence at then in 2001. If I find a significant PPS change in this placebo event
analysis, the rule change in board independence may not be the cause of the decrease in
PPS. Using 2-year window of before and after the placebo event (1999-2000 and 2001-
2002), I find no significant decrease in both PPS and StockPPS. I estimate -1.55 (t-
statistics 0.59) with CEO-firm pair fixed effects and -4.59 (t-statistics 0.70) with industry
dummies for PPS. For Stock PPS, the estimates are -1.05 (t-statistics 0.42) with CEO-
firm pair fixed effects and -4.01 (t-statistics 0.61) with industry dummies. From the
median, q25, and q75 regressions, I find unlike my results using real event, only q75
estimates are significant, i.e., the (insignificant) decrease in OLS regressions are driven
likely driven by outliers in this case.
Formal Compensation Contracts
A long-term formal compensation contract for a CEO limits the board’s control of
CEO pay-performance sensitivity because compensation can change only when there is
CEO turnover or when the contract expires. In such a case, enhanced board monitoring
will not affect compensation as long as the existing contract is in effect. Unlike CEOs in
other countries (Germany, for example), however, most U.S. CEOs do not have long-
term formal contracts; boards decide their compensation each year. If there is a formal
contract, it typically lasts at most three years (Gillan, Hartzell, and Parrino (2008)). This
implies that during my sample period of six years, even CEOs with formal contracts
would have different contracts before the new rule is adopted in 2002 and after the rule
became effective in 2004.
27
Staggered Boards
Because firms with staggered boards can replace only a third of their directors at a
time, they were granted an extended deadline, the end of fiscal year 2005, to comply with
the new rule. I examine the effect of this different deadline and the compliance of
staggered boards and non-staggered boards on the change in PPS. I find no evidence that
having boards with staggered or unstaggered terms affects the change in PPS. What
mattered was the percentage of independent directors prior to the rule. Firms that had
close to 50% of outside directors prior to the rule complied earlier and experienced no
significant decreases in PPS, while the firms that had far less than 50% of outside
directors tended to comply later and experienced larger decreases in PPS.
The Effect of Compensation Committee Independence
The compensation committee directly affects CEO compensation. I analyze
whether the new requirements of 100% of independent directors in compensation
committee in 2002 affects changes in PPS. Prior to the rule, about 25% of the firms had
on their compensation committees directors who were not independent. Following diff-
in-diff approach, I compare the PPS, StockPPS, and OptionPPS of the control group
firms to the firms whose boards lacked majorities of independent directors and whose
compensation committee included members who were not independent prior to 2002. I
find no difference from the results of the analysis of the impact of only the majority board
independence rule. Sorting firms further based on compliance with the two rules by
2004, I examine whether timely compliance with the rules had a different effect and
again find no evidence of difference.
New Stock and Option Grants
I examine the changes in new stock and option grants of the treatment group firms
relative to the control group firms to see whether boards are actively adjusting PPS, at
least the parts they have greatest control for. Using diff-in-diff approach to compare the
28
amount and the proportion of new stock and option grants to total compensations, I find
no significant change in new stock and option grants after the adoption of the majority
board independence rule. This result is consistent with the main result that boards do not
act against CEOs who aggressively sell their stocks after the adoption of the rule.
Three-Year Window
Extending the sample period to 1999-2006 includes three-year windows before
the adoption of the rule in 2002 and after it became effective in 2004. This generates
stronger results than what I report in the tables because both the StockPPS and
OptionPPS of the firms affected by the majority board independence rule drop
dramatically relative to the control group in 2006. One should be careful interpreting
these results, however, because most of the OptionPPS of 2006 is based on actual option
holdings and not the Core and Guay (2001) approximation method, which I used to
calculate OptionPPS for other years. This inconsistency is due to a change in reporting
format that year.
V. Conclusion
This paper investigates how board monitoring affects CEO pay-performance
sensitivity. I first show that monitoring and CEO pay-performance sensitivity are
substitutes in a principal-agent model adapting Holmstrom and Milgrom (1987).
Empirically, the adoption of a new law in 2002 that requires firms listed on major U.S.
exchanges to maintain majority-independent boards provides a unique opportunity to
analyze the effect of changes in board independence on CEO pay-performance
sensitivity. Several studies reported a negative relationship between board independence
and CEO pay-performance sensitivity in cross-section. It remained unclear, however,
whether board independence lowered CEO pay-performance sensitivity or it was a
coincidence of a third factor, such as firm size, that caused both high board independence
and low CEO pay-performance sensitivity.
29
Following the diff-in-diff approach, and controlling for time trend and pre-
adoption differences between the affected group and unaffected group, I find that
increased board independence lowered CEO pay-performance sensitivity. The decrease
in CEO overall pay-performance sensitivity is due to lower CEO stock ownership. CEO
option sensitivity does not change to compensate for the decrease in stock sensitivity,
implying that boards do not act to increase overall CEO pay-performance sensitivity.
This finding is consistent with the theoretical predictions that board monitoring and CEO
pay-performance sensitivity are substitutes.
30
References
Aggarwal, Rajesh, and Andrew Samwick, 1999, “The Other Side of the Trade-off: The Impact of Risk on Executive Compensation”, Journal of Political Economy, 107, 65-105. Baker, George, and Brian Hall, 2004, “CEO Incentives and Firm Size”, Journal of Labor Economics, 767-798. Boone, Audra, Laura Field, Jonathan Karpoff, and Charu Raheja, 2007, “The Determinants of Corporate Board Size and Composition: An Empirical Analysis”, Journal of Financial Economics, 85, 66-101. Cai, Jie, Yixin Liu, and Yiming Qian, 2007, “Information Asymmetry and Corporate Governance”, Working Paper. Chhaochharia, Vidhi, and Yaniv Grinstein, 2006, “CEO Compensation and Board Structure”, Working Paper. Coles, Jeffrey, Michael Lemmon, and Yan Wang, 2008, “The Joint Determinants of Managerial Ownership, Board Independence, and Firm Performance”, Working Paper. Core, John, Robert Holthausen, and David Larcker, 1999, “Corporate Governance, Chief Executive Officer Compensation, and Firm Performance”, Journal of Financial Economics, 51, 371-406. Core, John, and Wayne Guay, 2001, “Estimating the Value of Employee Stock Option Portfolios and Their Sensitivities to Price and Volatility”, Journal of Accounting Research, 40, 613-630. Core, John, and Wayne Guay, 2002, “The Other Side of Trade-Off: The Impact of Risk on Executive Compensation A Revised Comment”, Working Paper. Demsetz, Harold, and Kenneth Lehn, 1985, “The Structure of Corporate Ownership: Causes and Consequences”, Journal of Political Economy, 93, 1155-1177. Diamond, Douglas and Robert Verrecchia, 1991, “Disclosure, Liquidity, and the Cost of Capital”, Journal of Finace, 46, 1325-1359. Garen, John, 1994, “Executive Compensation and Principal-Agent Theory”, Journal of Political Economy, 102, 1175-1199. Gillan, Stuart, Jay Hartzell, and Robert Parrino, 2008, “Explicit vs. Implicit Contracts: Evidence from CEO Employment Agreements”, forthcoming Journal of Finance. Gillan, Stuart, Jay Hartzell, and Laura Starks, 2003, “Explaining Corporate Governance: Boards, Bylaws, and Charter Provisions”, Working Paper.
31
Harris, Lawrence, 1994, “Minimum Price Variations, Discrete Bid-ask Spreads, and Quotation Sizes”, Review of Financial Studies, 7, 149-178. Hartzell, Jay and Laura Starks, 2003, “Institutional Investors and Executive Compensation”, Journal of Finance, 58, 2351-2374. Heinrich, Ralph, 2000, "Complementarities in Corporate Governance: Ownership Concentration, Capital Structure, Monitoring and Pecuniary Incentives," Kiel Working Papers 968, Kiel Institute for the World Economy. Hermalin, Benjamin, and Michael Weisbach, 1988, “The Determinants of Board Composition”, Rand Journal of Economics, 19, 589-606. Holmstrom, Bengt, and Paul Milgrom, 1987, “Aggregation and Linearity in the Provision of Intertemporal Incentives”, Econometrica, 55, 303-328. Jensen, Michael, and Kevin Murphy, 1990, “Performance Pay and Top-Management Incentives”, Journal of Political Economy, 98, 225-264. Jin, Li, 2002, “CEO Compensation, Diversification and Incentives”, Journal of Financial Economics, 66, 29-63. Leland, Hayne, and David Pyle, 1977, Information Asymmetries, Financial Structure, and Financial Intermediation” Journal of Finance, 32, 371-387. Lakonishok, Josef, and Inmoo Lee, 2001, “Are Insider Trades Informative?,” Review of Financial Studies, 14, 79-111. Linck, James, Jeffry Netter, and Tina Yang, 2008, “The Determinants of Board Structure”, Journal of Financial Economics, 87, 308-328. Ofek, Eli, and David Yermack, 2000, “Taking Stock: Equity-Based Compensation and the Evolution of Managerial Ownership”, Journal of Finance, 55, 1367-1384. Prendergast, Canice, 1999, “The Provision of Incentives in Firms”, Journal of Economic Literature, 37, 7-63. Prendergast, Canice, 2002, “The Tenuous Trade-off between Risk and Incentives”, Journal of Political Economy, 110, 1071-1102. Ravina, Enrichetta, and Paola Sapienza, 2008, “What Do Independent Directors Know? Evidence from their Trading”, Working Paper. Seyhun, H.Nejat, 1986, “Insiders’ profits, costs of trading, and market efficiency”, Journal of Financial Economics”, 16, 189-212.
32
Vafeas, Nicos, 1999, “Board Meeting Frequency and Firm Performance”, Journal of Financial Economics, 53, 113-142. Weisbach, Michael, 1988, “Outside Directors and CEO Turnover”, Journal of Financial Economics, 20, 431-460. Yermack, David, 1996, “Higher market valuation of companies with a small board of directors”, Journal of Financial Economics, 40, 185-211.
33
Figure 1
A. Median PPS
B. Median StockPPS and median OptionPPS
0
5
10
15
20
25
1999 2000 2001 2002 2003 2004 2005
Affected
Control
34
Figure 2
PPS of control group in 2001 PPS of treatment group in 2004
0.0
1.0
2.0
3D
ensi
ty
0 100 200 300 400PPS
0.0
1.0
2.0
3D
ensi
ty
0 100 200 300 400PPS
PPS of treatment group in 2001
0.0
05.0
1.0
15.0
2D
ensi
ty
0 100 200 300 400 500PPS
PPS of treatment group in 2004
0.0
05.0
1.0
15.0
2.0
25D
ensi
ty
0 100 200 300 400 500PPS
35
Table 1
Summary statistics
Table 1 reports the summary statistics of the variables regarding CEO incentives, firm, CEO, and board characteristics. To investigate the effect of the adoption of majority independent rule in 2002, I choose firms that existed in 2000-2001 period and survived at least until 2004. Among this survived sample, I define the treatment groups as the group of firms that had less than 50% of independent directors either in 2000 or 2001 and thus were affected by the adoption of the new rule. The control group consists of firms that are not affected by the adoption of majority independence rule and the rules regulating committee independence.
PPS, the overall CEO pay-performance sensitivity, is defined as the sensitivity of CEO’s stock and option holdings to $1,000 shareholder wealth change. StockPPS stock holdings sensitivity, and OptionPPS is the option holdings sensitivity. Market value represents market value of equity at the fiscal year-end in millions. Sales is the net annual sales in millions. Dollar return variance is defined as the previous 60-month stock return volatility multiplied by the market value at the beginning of the estimation period. 3 year return is 3 year total return to shareholders reinvesting dividends. R&D/Assets is defined as R&D expenditure divided by assets, and missing R&D is a dummy that has value 1 if the data on R&D intensity is not available and 0 otherwise. Indpendent %, the variable of interest, is the percentage of independent directors of a board. Board size is the number of directors in a board. CEO tenure is the number of years the CEO has served as CEO, and CEO age is the age of the CEO. Total Direct Compensation is the total compensation to the CEO including salary, bonus, stock and option grants, long term incentive plan payouts and other compensation both annual and one-time payments.
A. Control group
Variable N Mean Median Stdev
PPS 1,624 23.08 11.12 42.63
StockPPS 1,624 13.33 2.16 40.94
OptionPPS 1,624 9.75 6.38 11.15
Market Value 1,619 9,229.17 2,298.34 22,202.32
Sales 1,624 6,312.80 1,884.97 13,831.03
Dollar return variance 1,536 2,541.67 600.71 7,955.02
3 year return 1,611 8.53 8.21 21.85
R&D/Assets 1,624 0.03 0.00 0.05
missinig R&D 1,624 0.40 0.00 0.49
Independent % 1,624 77.66 80.00 10.77
Board size 1,624 9.47 9.00 2.57
CEO tenure 1,624 6.91 5.00 6.24
CEO age 1,621 55.72 56.00 6.91
Total direct compensation 1,615 5,779.34 3,485.24 6,958.36
36
B. Treatment group
Variable N Mean Median Stdev
PPS 2,003 53.07 19.59 81.47
StockPPS 2,003 43.46 6.96 81.05
OptionPPS 2,003 9.61 6.37 11.84
Market Value 2,000 9,200.14 1,755.27 34,018.00
Sales 2,002 5,254.00 1,309.97 17,998.46
Dollar return variance 1,852 2,570.72 427.58 10,133.29
3 year return 1,988 10.47 9.26 23.32
R&D/Assets 2,003 0.02 0.00 0.04
missinig R&D 2,003 0.52 1.00 0.50
Independent % 2,001 52.01 50.00 15.08
Board size 2,003 9.13 9.00 2.68
CEO tenure 2,003 9.99 7.00 9.41
CEO age 1,996 56.33 56.00 8.80
Total direct compensation 1,992 6,014.89 2,402.76 18,308.16
37
Table 2
PPS and board independence in cross-section
I analyze the relationship between PPS and board independence in cross-section using the entire sample from 1996 to 2005. The dependent variable of column (1) and (2) is PPS, the sensitivity of CEO’s stock and option holdings to $1,000 shareholder wealth change. Column (3) and (4) investigate StockPPS, stock holdings sensitivity, and column (5) and (6) do OptionPPS. Indpendent %, the variable of interest, is the percentage of independent directors of a board. Rank of dollar return variance is defined as the annual rank percentile of the previous 60-month stock return volatility multiplied by the market value at the beginning of the estimation period. Ln(sales) is the natural log of sales. R&D/Assets is defined as R&D expenditure divided by assets, and missing R&D is a dummy that has value 1 if the data on R&D intensity is not available and 0 otherwise. CEO tenure is the number of years the CEO has served as CEO and board size is the number of the directors in a board. 3 year total returns to shareholders is 3 year total return to shareholders reinvesting dividends. Column (1), (3), and (5) includes CEO-firm pair fixed effects and year dummies while Column (2), (4), and (6) conducts OLS regression controlling for industry (SIC 2-digit) and year dummies. The t-statistics are reported in parentheses. The t-statistics of column (2), (4), and (6) are based on robust standard errors. * and ** denote significance at the 5% and 1% level, respectively.
(1) (2) (3) (4) (5) (6)
PPS PPS StockPPS StockPPS OptionPPS OptionPPS
Independent % ‐0.093 ‐0.478 ‐0.055 ‐0.521 ‐0.038 0.043
(3.03)** (14.20)** (1.88) (15.73)** (4.22)** (5.90)**
Rank of dollar return variance
‐10.737 ‐12.019 ‐9.404 ‐7.988 ‐1.333 ‐4.031
(4.14)** (4.01)** (3.80)** (2.77)** (1.76) (4.25)**
ln(sales) ‐7.971 ‐2.587 ‐5.391 ‐0.915 ‐2.580 ‐1.672
(7.61)** (4.01)** (5.41)** (1.53) (8.46)** (6.56)**
R&D/Assets 6.982 ‐46.074 ‐1.876 ‐56.899 8.857 10.824
(0.50) (5.20)** (0.14) (6.50)** (2.19)* (3.32)**
Missing R&D ‐2.522 6.335 ‐1.948 6.024 ‐0.574 0.311
(0.89) (3.61)** (0.72) (3.58)** (0.70) (0.80)
CEO Tenure 0.604 2.566 0.411 2.470 0.193 0.095
(1.98)* (27.26)** (1.42) (26.01)** (2.17)* (6.09)**
Board size ‐0.699 ‐2.765 ‐0.648 ‐2.366 ‐0.051 ‐0.399
(3.23)** (11.95)** (3.14)** (10.77)** (0.82) (7.27)**
3 year total returns to shareholders
0.036 0.020 0.021 0.007 0.015 0.013
(2.82)** (0.59) (1.73) (0.23) (4.01)** (1.70)
Constant 103.948 135.122 75.426 116.090 28.522 19.032
(13.32)** (8.43)** (10.15)** (7.00)** (12.55)** (9.96)**
Industry FE No Yes No Yes No Yes
Time FE Yes Yes Yes Yes Yes Yes
CEO‐Firm FE Yes No Yes No Yes No
Observations 12214 12214 12214 12214 12214 12214
# of CEO‐Firm pairs 3406 3406 3406
38
Table 3
Year-by-year board independence and CEO incentives
Table 3 reports the time trend of summary statistics of the control group and treatment group on board independence and CEO incentives. Indpendent % is the percentage of independent directors of a board. PPS, the overall CEO pay-performance sensitivity, is defined as the sensitivity of CEO’s stock and option holdings to $1,000 shareholder wealth change. StockPPS stock holdings sensitivity, and OptionPPS is the option holdings sensitivity.
A. Control group
Independent% PPS StockPPS OptionPPS year N mean median stdev mean median stdev mean median stdev mean median stdev 2000 248 77.68 80.00 10.24 22.38 10.02 41.15 13.61 2.04 39.96 8.77 5.34 9.71 2001 281 78.05 80.00 10.31 24.45 11.48 48.37 15.20 2.33 47.34 9.25 5.80 9.97 2002 279 77.19 80.00 11.28 25.27 11.50 47.50 14.83 2.33 45.33 10.44 7.39 11.39 2003 279 77.31 80.00 11.11 22.65 11.79 38.68 11.88 2.07 36.18 10.77 7.27 12.95 2004 277 77.51 78.57 10.92 21.32 11.32 36.59 11.56 2.00 35.18 9.76 6.33 10.67 2005 260 78.25 80.00 10.72 22.22 10.42 42.14 12.84 2.13 40.22 9.38 5.50 11.76
B. Affected firms
Independent% PPS StockPPS OptionPPS year N mean median stdev mean median stdev mean median stdev mean median stdev 2000 323 42.48 44.44 11.67 64.10 23.15 89.99 54.55 10.18 88.72 9.54 5.33 14.10 2001 344 45.21 44.44 12.85 58.27 22.62 84.14 49.14 8.96 84.56 9.14 5.19 11.12 2002 337 48.72 50.00 13.71 55.40 20.94 86.67 45.44 7.40 86.60 9.96 7.13 11.20 2003 339 54.12 54.55 13.87 50.59 19.01 78.03 40.73 6.90 77.34 9.86 6.71 11.31 2004 341 59.34 58.33 13.35 48.39 17.58 77.06 38.17 6.12 76.71 10.23 6.95 11.91 2005 317 62.44 62.50 14.02 41.33 15.08 69.64 32.42 4.98 68.99 8.91 5.61 11.23
39
Table 4
PPS and board independence: diff-in-diff approach
I investigate the effect of the adoption of majority independent rule in 2002 on the dependent variable PPS, the sensitivity of CEO’s stock and option holdings to $1,000 shareholder wealth change. (Year04-05)*(<50% independent directors before 02) is dummy that has value 1 if year is 2004 or 2005 and the firm had less than 50% of independent directors on its board before 2002, and 0 otherwise. (<50% independent directors before 02) is a dummy that has value 1 if the firm had less than 50% of independent directors on its board before 2002, and 0 otherwise.
Rank of dollar return variance is defined as the annual rank percentile of the previous 60-month stock return volatility multiplied by the market value at the beginning of the estimation period. Ln(sales) is the natural log of sales. R&D/Assets is defined as R&D expenditure divided by assets, and missing R&D is a dummy that has value 1 if the data on R&D intensity is not available and 0 otherwise. CEO tenure is the number of years the CEO has served as CEO and board size is the number of the members of a board. 3 year total returns to shareholders is 3 year total return to shareholders reinvesting dividends.
Column (1) includes CEO-firm pair fixed effects and year dummies. Column (2) conducts OLS regression controlling for industry (SIC 2-digit) and year dummies. Column (3)-(5) run robust regressions. (3) is median regression, (4) is 25 percentile, and (5) is 75 percentile. The t-statistics are reported in parentheses. The t-statistics of column (2) are based on robust standard errors. * and ** denote significance at the 5% and 1% level, respectively.
(1) (2) (3) (4) (5)
PPS PPS PPS PPS PPS
(Year04‐05)*(<50% independent directors before 02) ‐4.736 ‐8.115 ‐1.797 ‐1.352 ‐8.476
(3.12)** (2.14)* (2.14)* (2.80)** (4.03)**
<50% independent directors before 02 17.455 3.043 1.529 13.351
(6.63)** (5.72)** (5.06)** (10.22)**
Rank of dollar return variance ‐3.450 ‐14.540 ‐8.131 ‐5.842 ‐8.431
(0.90) (2.11)* (6.68)** (8.31)** (2.70)**
ln(sales) ‐7.955 ‐3.641 ‐1.615 ‐1.267 ‐3.095
(4.21)** (2.52)* (6.41)** (9.01)** (4.62)**
R&D/Assets 1.317 ‐40.387 ‐12.295 ‐0.852 ‐40.484
(0.06) (1.56) (1.90) (0.23) (2.44)*
Missing R&D 0.582 9.433 0.532 0.189 ‐0.112
(0.15) (2.99)** (0.82) (0.49) (0.07)
CEO Tenure 0.991 2.896 1.711 0.902 3.962
(2.66)** (17.35)** (66.70)** (59.05)** (60.09)**
Board size ‐0.369 ‐3.527 ‐0.953 ‐0.642 ‐1.625
(1.02) (6.82)** (9.10)** (10.76)** (6.02)**
3 year total returns to shareholders ‐0.017 ‐0.090 0.003 0.001 ‐0.001
(0.83) (1.69) (0.32) (0.18) (0.04)
Constant 98.705 102.071 37.506 26.520 69.520
(6.98)** (7.15)** (11.07)** (13.63)** (7.62)**
Industry FE No Yes Yes Yes Yes
Time FE Yes Yes Yes Yes Yes
CEO‐Firm FE Yes No No No No
Observations 3388 3388 3388 3388 3388
# of CEO‐Firm pairs 942
40
Table 5
StockPPS and board independence: diff-in-diff approach
I investigate the effect of the adoption of majority independent rule in 2002 on the dependent variable StockPPS, the sensitivity of CEO’s stock holdings to $1,000 shareholder wealth change. (Year04-05)*(<50% independent directors before 02) is dummy that has value 1 if year is 2004 or 2005 and the firm had less than 50% of independent directors on its board before 2002, and 0 otherwise. (<50% independent directors before 02) is a dummy that has value 1 if the firm had less than 50% of independent directors on its board before 2002, and 0 otherwise.
Rank of dollar return variance is defined as the annual rank percentile of the previous 60-month stock return volatility multiplied by the market value at the beginning of the estimation period. Ln(sales) is the natural log of sales. R&D/Assets is defined as R&D expenditure divided by assets, and missing R&D is a dummy that has value 1 if the data on R&D intensity is not available and 0 otherwise. CEO tenure is the number of years the CEO has served as CEO and board size is the number of the members of a board. 3 year total returns to shareholders is 3 year total return to shareholders reinvesting dividends.
Column (1) includes CEO-firm pair fixed effects and year dummies. Column (2) conducts OLS regression controlling for industry (SIC 2-digit) and year dummies. Column (3)-(5) run robust regressions. (3) is median regression, (4) is 25 percentile, and (5) is 75 percentile. The t-statistics are reported in parentheses. The t-statistics of column (2) are based on robust standard errors. * and ** denote significance at the 5% and 1% level, respectively.
(1) (2) (3) (4) (5)
StockPPS StockPPS StockPPS StockPPS StockPPS
(Year04‐05)*(<50% independent directors before 02) ‐4.217 ‐8.531 ‐1.249 ‐0.363 ‐7.558
(2.92)** (2.22)* (1.86) (1.93) (4.57)**
<50% independent directors before 02 19.532 2.513 0.784 12.504
(7.39)** (5.91)** (6.41)** (12.35)**
Rank of dollar return variance ‐2.984 ‐10.742 ‐1.812 ‐1.314 ‐6.465
(0.82) (1.54) (1.86) (4.89)** (2.66)**
ln(sales) ‐6.222 ‐2.488 ‐0.495 ‐0.081 ‐0.950
(3.46)** (1.70) (2.46)* (1.50) (1.82)
R&D/Assets 2.290 ‐53.276 ‐21.029 ‐5.414 ‐43.281
(0.11) (1.96)* (4.07)** (3.96)** (3.29)**
Missing R&D 0.413 8.565 0.284 0.418 0.867
(0.11) (2.66)** (0.54) (2.88)** (0.67)
CEO Tenure 0.914 2.805 1.240 0.394 3.474
(2.58)** (16.82)** (60.33)** (71.34)** (64.29)**
Board size ‐0.282 ‐3.196 ‐0.488 ‐0.187 ‐1.154
(0.82) (6.20)** (5.86)** (8.17)** (5.54)**
3 year total returns to shareholders ‐0.018 ‐0.100 ‐0.011 ‐0.003 ‐0.042
(0.92) (1.89) (1.43) (1.26) (2.12)*
Constant 76.131 82.457 11.492 3.675 52.219
(5.66)** (5.53)** (4.24)** (5.33)** (7.28)**
Industry FE No Yes Yes Yes Yes
Time FE Yes Yes Yes Yes Yes
CEO‐Firm FE Yes No No No No
Observations 3388 3388 3388 3388 3388
# of CEO‐Firm pairs 942
41
Table 6
Board meeting frequency, firm size and board independence
I investigate the effect of the adoption of majority independent rule in 2002 on the dependent variables the number of board meetings (# meetings) for a given year in column (1)-(2) and PPS, the sensitivity of CEO’s stock and option holdings to $1,000 shareholder wealth change, in column (3)-(6). I compare the change in PPS of large treatment group firms (upper 30 percentile) to large control group firms in column (4)-(5) and that of small treatment group firms (bottom 30 percentile) to small control group firms in column (5)-(6). (Year04-05)*(<50% independent directors before 02) is dummy that has value 1 if year is 2004 or 2005 and the firm had less than 50% of independent directors on its board before 2002, and 0 otherwise. (<50% independent directors before 02) is a dummy that has value 1 if the firm had less than 50% of independent directors on its board before 2002, and 0 otherwise.
The controls used are as follows. Rank of dollar return variance is defined as the annual rank percentile of the previous 60-month stock return volatility multiplied by the market value at the beginning of the estimation period. Ln(sales) is the natural log of sales. R&D/Assets is defined as R&D expenditure divided by assets, and missing R&D is a dummy that has value 1 if the data on R&D intensity is not available and 0 otherwise. CEO tenure is the number of years the CEO has served as CEO and board size is the number of the members of a board. 3 year total returns to shareholders is 3 year total return to shareholders reinvesting dividends. For the analysis on the number of board meetings, I include a dummy for the CEO’s last year of service, and a dummy for the CEO’s first year of service.
Column (1), (3), and (5) include CEO-firm pair fixed effects and year dummies. Column (2), (4), and (6) conduct OLS regression controlling for industry (SIC 2-digit) and year dummies. The t-statistics are reported in parentheses. The t-statistics of column (2), (4), and (6) are based on robust standard errors. * and ** denote significance at the 5% and 1% level, respectively.
(1) (2) (3) (4) (5) (6)
Large firms Small firms
# meetings # meetings PPS PPS PPS PPS
(Year04‐05)*(<50% independent directors before 02) 0.404 0.582 ‐6.259 ‐10.960 ‐1.086 ‐3.435
(2.35)* (2.54)* (3.32)** (2.75)** (0.34) (0.38)
<50% independent directors before 02 ‐0.525 17.797 19.234
(4.50)** (5.91)** (2.89)**
Industry FE No Yes No Yes No Yes
Time FE Yes Yes Yes Yes Yes Yes
CEO‐Firm FE Yes No Yes No Yes No
Observations 3390 3390 972 972 1064 1064
# of CEO‐Firm pairs 950 315 373
top related