Do Firms Replenish Executives’ Incentives After Equity
Sales?
Tomislav Ladika∗
University of Amsterdam†
March 4, 2013
Abstract
Boards grant executives equity to align their incentives with those of shareholders. Yet
executive equity sales are common — 61 percent of executives sell firm equity during their
tenure — and can cause holdings in the firm to become suboptimally low. Theory finds
that boards can restore executives’ incentives by shifting the composition of subsequent pay
toward more equity. However, this paper shows empirically that boards do not replenish
incentives lost from sales. Firm-level changes such as reduced growth opportunities may
cause executives to sell equity and simultaneously reduce their need for incentives, potentially biasing empirical estimates of incentive replenishment. I develop a novel identification
strategy to account for such variables: I compare executives who sell equity to other top
executives at the same firm who do not sell. I show that boards grant the same pay to selling
and non-selling executives at the same firm. This result is robust to a variety of reasons
why the board may want selling executives to own less equity. My findings suggest that
boards do not maintain executives’ incentives at an optimal level, as predicted by efficient
contracting theory.
∗
I thank Ivo Welch, Ross Levine, and Ken Chay for advice and guidance, and Brown University’s William
R. Rhodes Center and C.V. Starr Program in Commerce, Organizations, and Entrepreneurship for financial
support. I also thank Kevin Murphy, Geoffrey Tate, Zacharias Sautner, Florian Peters, Pierre Chaigneau, and
seminar participants at Brown University and the University of California-Los Angeles for helpful comments. All
remaining errors are my own.
†
Contact Information: The author can be reached by email at [email protected], and by phone at 31-(0)205255351 or 1-401-3786405. The author’s mailing address is University of Amsterdam Faculty of Economics and
Business, Finance Group, Roetersstraat 11, 1018 WB Amsterdam, The Netherlands.
1
Keywords: Executive compensation, corporate governance, agency theory, incentive compensation, equity compensation, managerial incentives, equity sales, options
2
1
Introduction
Corporate boards commonly state that the purpose of equity compensation is to align top executives’ incentives with those of shareholders. Managerial ownership of firm equity can alleviate
agency conflicts with shareholders, because a manager with a large fraction of wealth tied to
firm performance may be more likely to pursue value-maximizing decisions than private benefits
like shirking or empire-building (Jensen & Meckling (1976)). But executives also sometimes sell
large amounts of firm equity, reducing their stake in the firm — 61 percent of top executives
at S&P 1500 firms sell equity during their tenure, and the median annual sale decreases total
firm holdings by 15 percent.1 What happens to executives’ compensation after such sales? Although sales can cause an executive’s incentives to become suboptimally low, potentially leading
to value-destroying decisions, little is known about how boards act in response.
Economic theory predicts that there is a unique level of incentives which is optimal for an executive to own (Holmstrom & Milgrom (1987)), and that boards should adjust annual compensation
to keep the executive’s incentives at this optimum. Edmans, Gabaix, Sadzik & Sannikov (2012)
show that in a dynamic setting, when incentives fall below the optimal level, boards should
respond by simultaneously increasing equity pay and reducing fixed pay. By shifting the composition of pay instead of granting new equity on top of existing pay, boards avoid rewarding the
executive for the decrease in incentives. In this paper, I empirically test whether boards restore
incentives after equity sales in this manner, using a simple hypothesis:
An executive who sells equity should subsequently receive a higher proportion of pay
in equity than an executive who does not sell.
A key empirical challenge to testing this hypothesis is to differentiate between sales that cause
incentives to become suboptimally low, and sales that coincide with a decrease in the optimal
incentive level. In the latter case, an executive’s post-sale holdings in the firm may remain at
1
These statistics are from Table 2, and are for a sample of top executives at S&P 1500 firms from 1996 to
2007. See also Section 2 for a description of my data sources for equity transactions and executives’ holdings of
firm equity.
3
the optimal level, and the board should leave pay unchanged. It is important to emphasize
that the optimal incentive level is determined by the shareholders’ objective function. Executives
sometimes choose to sell equity for reasons that may not benefit shareholders, such as to diversify
wealth or to purchase a home. The board should restore incentives after such sales, unless doing
so exacerbates agency problems.
Some firm-level changes cause both executives and shareholders to benefit from the executive
owning less firm equity. For example, a young firm may rely on high levels of equity pay to
incentivize executives because the board is not able to evaluate which growth opportunity to
pursue. If the executives choose a successful strategy and establish a profitable business, they
may sell equity to lock in gains from stock price appreciation. At the same time, the board
may be able to substitute direct monitoring for equity incentives (Prendergast (2002)). Such
firm-level changes are often unobservable, and may bias empirical estimates of the relationship
between sales and pay changes toward zero.
I develop a novel identification strategy to account for such changes: I compare executives who sell
equity to other top executives at the same firm who at the same time do not sell. The executives
in my sample include the Chief Executive Officer, Chief Financial Officer, Chief Operating Officer,
and non-executive Chairman of the board. These executives are responsible for decisions that
affect the entire firm, so firm-level changes should have a similar effect on each executive’s optimal
incentive level. For example, when growth opportunities increase, it is unlikely that the best way
to reduce agency problems is to grant more equity to the COO and less to the CFO.
My empirical specification assumes that at the start of each year, boards grant compensation
that sets each executive at their unique optimal incentive level.2 Boards should therefore grant
a higher proportion of pay in equity to executives who sold more during the previous year.
Because executives differ in their level of risk aversion, my specification includes executive-level
2
This assumption is not crucial. In particular, if two executives have incentives below the optimal level, and
one of them sells equity while the other does not, the selling executive’s incentives will fall further below the
optimum. In response, the board should grant the selling executive a higher proportion of pay in equity. However,
a key assumption for my specification is that boards do not systematically allow some executives to accumulate
too much equity.
4
fixed effects and tests whether boards grant a larger increase in equity pay to selling executives.
Throughout this paper, I also take empirically observed vesting conditions as given. Boards could
prevent equity sales by setting long vesting periods, but in practice new equity grants typically
begin to vest within a year (Cadman, Rusticus & Sunder (2012)), and few firms explicitly restrict
executives from selling vested equity (Cook (2009)).3
My empirical results show that boards in fact replenish little to none of the incentives lost due
to an equity sale, and instead grant similar pay to selling and non-selling executives at the same
firm. I measure equity sales in two ways: the annual dollar value of equity sold minus the cost of
exercising stock options, and the decrease in the executive’s total incentives due to the sale. (I
follow Baker & Hall (2005) by defining incentives as the change in dollar value of equity given a
one-percent change in the firm’s stock price.) I find that for each $1 decrease in incentives due to
a sale, an executive receives at most $0.07 in new incentives, relative to a non-selling executive
at the same firm. I also find no change in the composition of pay following equity sales. While
it may not be surprising that boards do not respond to small equity sales, my results show that
boards also do not adjust pay after large sales which decrease an executive’s total firm holdings
by 25 percent or more.
While boards could respond to equity sales in different ways, it is unlikely that the optimal
response is to grant similar pay to executives who sell large amounts of equity as to executives
who do not sell. Boards seem to either allow selling executives’ incentives to decrease below the
optimum, or to provide non-selling executives with too much equity in the firm. In either case,
my results suggest that our understanding of executive compensation is incomplete.
I conduct numerous robustness tests to address limitations to my identification strategy, and my
results do not change. First, boards might evaluate the performance of division-level managers
differently from that of other top executives, so firm-level changes may have a different effect on
3
It may be optimal for boards to allow executives ex ante flexibility to sell equity by allowing it to vest within
a few years. For example, an executive’s wealth may become too closely tied to the firm when the stock price
appreciates substantially. This may cause the executive to undertake suboptimally conservative decisions, unless
he is able to sell some of his holdings in the firm (Edmans et al. (2012)). In this paper, I do not study how boards
set vesting conditions.
5
the pay of these executives (Aggarwal & Samwick (2003)). I therefore repeat my analysis on a
subsample of just “C-Suite” executives — those who have “Chief” in their title and oversee the
entire firm.
Second, several years after hiring an executive the board may learn that it initially overestimated
his optimal incentive level, and reduce subsequent equity pay. This may occur just as the executive
is beginning to sell his holdings in the firm. To overcome this problem, I compare only executives
who joined the firm’s upper ranks at the same time or have worked together for several years.
The board should learn information about these executives at the same time.
Third, my specification does not account for shocks to executives’ liquidity or wealth held outside
the firm. Such shocks may increase an executive’s risk aversion, and the board’s optimal response
may be to allow the executive to sell equity to diversify his remaining wealth. To account for this
possibility, I repeat my tests using only executives who voluntarily set aside a portion of their
annual salary into a deferred compensation account (similar in purpose to a 401-k account).
I also test whether boards do not replenish incentives because they anticipate that executives will
sell equity, and incorporate this expectation ex ante into executives’ annual equity grants. Ofek
& Yermack (2000) find that executives frequently sell stock shortly after receiving new equity
pay, and if boards expect such sales they may adjust the flow of annual equity grants prior to
the sale. However, I show that boards also do not replenish incentives following unanticipated
equity sales, such as sales following early exercise of stock options.
One limitation of my paper is that it does not offer a conclusive answer for why boards do not
adjust pay following equity sales. However, it does raise important questions about the prominent
explanations for why boards grant equity pay. Much of the literature on executive compensation
has traditionally viewed contracts as the product of arms-length negotiations between the board
and executives.4 This “Efficient Contracting” perspective argues that boards attempt to set ex4
Examples of optimal contracting models in the presence of agency problems include Ross (1973), Grossman
& Hart (1983), Holmstrom & Milgrom (1987), Holmstrom & Milgrom (1991), Prendergast (2002), and Edmans
et al. (2012). See also Adams, Hermalin & Weisbach (2008), Murphy (1999), and Core, Guay & Larcker (2003)
for literature reviews on corporate governance and executive pay.
6
ecutives’ equity holdings at the unique level that most effectively mitigates agency problems. My
result that boards grant similar pay to selling and non-selling executives is difficult to reconcile
with the targeting of an optimal incentive level. This result may be consistent with the “Managerial Power” view in which executives use leverage over the board to negotiate favorable pay
(Bebchuk & Fried (2004)). By not responding to equity sales, boards effectively allow executives
to choose how much equity to hold in the firm. However, this does not prove that equity pay is
disguised wealth extraction from the firm.
My paper makes two additional contributions to the executive compensation literature. First,
while much empirical work has examined whether boards structure annual pay grants efficiently
( Jensen & Murphy (1990), Yermack (1995), Hall & Liebman (1998), Morck, Schleifer & Vishny
(1988)), I examine whether boards maintain executives’ overall holdings in the firm. Core &
Guay (1999) examine a similar question, and find that boards grant more equity to CEOs whose
total equity incentives are below the value predicted by a regression model. However, they do
not measure equity sales directly, and their empirical methodology differs from mine.5 Previous
work has also examined selling patterns by executives (Ofek & Yermack (2000), Jaffe (1974),
Seyhun (1986)). However, to the best of my knowledge my paper is the first to examine how
boards adjust pay following equity sales.
Second, my identification strategy of comparing top executives at the same firm may be useful
for empirically identifying important questions in corporate governance. The literature almost
exclusively focuses on the compensation of CEOs, but my paper provides some new findings on
how the boards sets the compensation of non-CEO executives. One is that an executive’s equity
pay increases when he is promoted to a C-Suite position.6
The rest of this paper is organized as follows. Section 2 discusses data and empirical measures.
Section 3 develops the empirical specification and identification strategy. Section 4 presents
5
Core & Guay (1999) test whether the level of new incentives is higher for CEOs whose incentives fall below
a benchmark model. I test whether the change in incentives is higher for executives who recently sold equity, to
account for differences in the level of incentives across executives.
6
This finding may be consistent with either the efficient contracting or managerial power perspectives. This
is because boards which have been captured by executives may disguise large increases in total pay by granting
equity which the executive can easily sell.
7
my main results, and Section 5 presents results from robustness tests. Section 6 concludes.
Information on the construction of all variables is in the Data Appendix.
2
Data and Descriptive Statistics
In this section I describe my sample, data sources, and empirical measures of executive equity
sales and pay changes. I also present descriptive statistics on the frequency and size of executive
equity sales.
2.1
Data and Sample
I use Compustat Execucomp to obtain data on the annual compensation of firms’ top executives,
as well as their total holdings of firm equity. Execucomp covers mostly S&P 1500 firms starting
in 1992. It contains the dollar value of annual total pay, as well as the value of pay components
such as salary, cash-based incentive payments, stock options, and regular stock. It also contains
data on executives’ total holdings of firm equity, including the total number and value of regular
shares and stock options owned at the end of each fiscal year. Firms are required to report
this data for the CEO and four other highest-paid executives in their annual proxy statement to
shareholders. Most Execucomp firms report data for five to seven top executives.7
I obtain data on executives’ transactions of firm equity from Thomson Insiders, which covers all
publicly traded firms with full data available starting in 1996. All top executives, directors, and
large shareholders must to report any transaction of firm equity to the Securities and Exchange
Commission. (Some transactions, such as short sales, are banned.) The most common transactions are stock sales, option exercises, open-market equity purchases, and the disposition of
shares at option exercise to pay tax obligations or exercise cost.
7
In 80% of the firm-year observations in my sample, compensation data is available for five to seven top
executives. In eight percent of firm-years, data is reported for fewer than five executives, even though the Securities
and Exchange Commission requires firms to report data for their five highest-paid executives. In 12 percent of
firm-years, data is reported for eight or more executives.
8
My sample starts with all firms which are in both databases. I drop 42 Execucomp firms which I
am unable to match to firms in Thomson Insiders, 37 firms which do not grant equity to a single
executive in any year, and 142 firms which do not remain in Execucomp past 1995.8 My final
sample contains 30,875 executives at 2,819 firms, and it covers the years 1996 to 2007.
Table 1, Panel A shows the different types of executives who comprise my sample, with observations at the firm-executive-year level. Forty-seven percent of observations are C-Suite executives.
Of these, 19 percent are CEOs, 15 percent are CFOs, eight percent are COOs, and four percent are other chief executives. Additionally, 6 percent of executives are non-executive board
chairmen, 6 percent have only the title “Senior Vice President”, and 9 percent have titles which
indicate a specialized role in the firm.9 The remaining 30 percent of executives do not clearly
fall into any particular category. Panel A also shows that about half of observations are junior
executives who have been at the top of the firm for less than five years.
2.2
Pay and Equity Sales Measures
Throughout the paper, I use three different empirical measures of executive pay changes:
1. Pct. Chg. New Incentives [t] is the annual percentage change in incentives received from
new equity compensation. I define incentives throughout this paper as the dollar change in
the value of equity for a one percent change in the stock price, a standard measure in the
literature (Baker & Hall (2005)).10
8
In order to merge the two databases, I matched firms by CUSIP number and stock ticker. I verified all matches
by comparing firm names, and hand-checked any matches for which the names were not virtually identical. After
matching firms, I matched executives using an algorithm which compares the first and last name of all Execucomp
executives to the first and last names of all Thomson executives at the same firm. I then compared unmatched
executives by last name and first initial, and hand-checked all resulting matches. All programs used to generate
my combined dataset are available upon request.
I was unable to directly match 15 Execucomp firms to Thomson Insiders. Additionally, for 24 firms I was unable
to match any of the CEOs listed in Execucomp to individuals listed in Thomson, and for three firms I was unable
to match non-CEO executives listed in Execucomp to individuals listed in Thomson. For firms remaining in my
sample, I was unable to match 964 executives across databases.
9
I label an executive as specialized if his title includes certain terms, such as “group”, “division”, “marketing”,
“research and development”, etc. This definition is a coarser version of the measure used by Aggarwal & Samwick
(2003), who identify specialized executives by hand-checking all executive titles X
in Execucomp.
10
Specifically, incentives from new equity granted in year t are St · Pt /100 +
Ni,t · ∆i · Pt /100, where St is
9
i
2. Chg. Pay Ratio [t] is the annual change in the ratio of equity pay to total pay, multiplied
by 100. Equity pay is the dollar value of new stock grants and the Black-Scholes value of
new option grants. Total pay is the dollar value of all pay received by the executive during
the year.
3. Chg. Total Incentives [t] is the amount by which the executive’s total incentives increase
following a new equity grant. It equals incentives from annual equity pay scaled by total
incentives from the start of the previous year, multiplied by 100.11 This variable is bounded
from below by zero.
Each of these variables tests a different possible response by boards to equity sales. The first
tests whether boards grant more equity incentives to executives following a sale. The second
tests whether boards shift the composition of annual pay toward more equity. The third tests
the extent to which boards replenish executives’ total incentives.12 While my hypothesis is that
boards should respond to sales by changing the pay ratio, I use the other two measures because
it is also important to know whether boards grant a larger amount of equity pay.
I measure incentives using the sensitivity of pay to firm performance based on the Black-Scholes
option delta. This may lead to biased estimates of the level of incentives, because the BlackScholes delta does not vary with the executive’s marginal utility (Ross (2004), Jenter (2002)).13
Because my empirical specification compares changes in incentives instead of the level of incentives, this will not affect my results as long as the change in marginal utility is approximately
the same for executives at the same firm.
I use two different measures of equity sales in this paper:
the number of regular shares granted in year time t, Pt is the grant-date stock price, Ni,t is the number of stock
options in each new grant i made to the executive in year t, and ∆i is the option delta. I follow the literature
by defining the option delta as the derivative of the Black-Scholes value of the stock options with respect to the
stock price.
11
I do not scale by end-of-year total incentives because equity sales during the year decrease this number,
creating a mechanical relationship between sales and subsequent changes in total incentives.
12
Changes in annual equity pay may not accurately capture the increase in an executive’s total holdings. For
an executive who received abnormally low pay in the previous year, the percentage change in annual pay may be
high even if the incentives from new pay are small relative to the amount of equity sold.
13
An executive’s incentives depend on his marginal utility with respect to changes in effort. Jenter (2002)
argues that marginal utility is negatively correlated with changes in pay-for-performance sensitivity. For example,
the option delta rises with a stock price increase, but the executive’s marginal utility may decrease. This causes
commonly used empirical measures of incentives, which are based on pay-for-performance sensitivity, to overstate
the true level of incentives.
10
1. Sale Indicators: I generate three binary indicators for small, medium, and large equity
sales. I calculate the ratio of dollar value of equity sold in a year to dollar value of the
executive’s total holdings of firm equity at the start of the year.14 For stock options, I
subtract the option exercise price from the dollar value of shares sold. Small sales equal
less than 10 percent of total holdings, medium sales are 10 to 25 percent of total holdings,
and large sales are more than 25 percent of total holdings.15 I also create a purchase
indicator equal to one if the executive buys any equity on the open market.
To measure changes in the amount of equity sold, I take the change in each indicator
from year t − 2 to t − 1. For example, “Large Sale [t]” equals one if the executive sold a
large amount of equity in the previous year but not two years ago, and equals zero if the
executive sold large amounts in neither or both years. This accounts for the possibility that
the board increases equity pay in response to a sale in year t − 2, and then leaves equity
pay unchanged following a sale in year t − 1. (My results do not change substantially if I
use regular indicators.)
2. Pct. Drop Incentives [t-1] is the amount by which an equity sale causes an executive’s
total incentives to decrease. I first measure the total incentives that an executive would
have had at the end of the year, had no equity been sold or purchased.16 I then take the
difference between this number and the executive’s actual year-end incentives, divide by
total incentives at the start of the year, and multiply by 100.
The sale indicators are a simple way to test whether selling executives receive a larger increase
in equity pay than non-selling executives, and whether pay changes vary with the size of sales.
However, the indicators do not use all of the variation in equity sales, and they are also correlated
with contemporaneous stock returns — increases in the stock price increase the value of equity
sold during the year.
“Pct. Drop Incentives” is a different and perhaps more precise measure. It measures incentives
instead of dollars sold, and it is not affected by contemporaneous stock returns or changes in
volatility, because both hypothetical and actual equity holdings are valued using the same yearend inputs. Additionally, the relationship between “Pct. Drop Incentives” and “Chg. Total
14
When generating these indicators, I also include the dollar value of equity which the executive relinquishes
to the firm to pay tax obligations or cover option exercise costs (these dispositions are labeled with transaction
code F in Thomson Insiders). The results do not change if I omit these equity dispositions.
15
The 10 percent and 25 percent cutoffs are approximately equal to the 33rd and 66th percentiles of the
distribution of dollars of equity sold divided by total dollar value of equity holdings, conditional on a sale occurring.
16
Calculating incentives in this way is not equivalent to observing the counterfactual in which the executive
does not sell equity. The reason is that the year-end stock price, which I use to calculate the incentives, may be
affected by decisions the executive makes after selling equity. However, such decisions should be more likely to
lead to stock price decreases, because the executive has weaker incentives after selling equity (also, the market
could interpret managerial sales as a negative signal). Therefore, my measure likely understates the true decrease
in incentives due to equity sales.
11
Incentives” is a direct measure of the replenishment rate, because it estimates how much an
executive’s total incentives increase following each 1 percent decrease in total incentives due to
sales. If boards are replenishing an executive’s equity holdings, a new equity grant should increase
total incentives by an amount similar to the decrease caused by the sale (i.e., the replenishment
rate should be about 1).
2.3
Descriptive Statistics
Table 1, Panel B presents descriptive statistics on annual pay and total equity holdings for the
executives in my sample. The median executive receives about $1 million in annual total pay and
$340,000 in annual equity pay; the median equity pay ratio is 36.5 percent of annual pay. The
median executive owns $2.6 million worth of firm equity, but mean ownership is much higher,
about $21 million.
Panel B also presents statistics on equity incentives. The statistics indicate that a one-percent
increase in the stock price causes the value of total firm equity to increase by $61,000 at the
median ($471,000 on average). Incentives from new equity compensation are substantially lower,
$5,400 at the median ($22,400 on average).
Panel C presents statistics for the three measures of equity pay which I use as dependent variables
throughout this paper. The median change in annual pay is zero, for both new equity incentives
and the equity pay ratio. The change in total incentives increases by 11 percent at the median
(this variable, unlike the other two, is bounded from below by zero). Panel D breaks down the
mean and median values for each variable by executive type.
Table 2 presents descriptive statistics on the frequency and size of equity sales. Panel A shows
that significant variation exists across firms in the fraction of top executives selling equity in any
given year. In 27 percent of firm-years no top executives sell equity, and in another 27 percent of
firm-years fewer than half of the top executives sell. In 32 percent of firm-years more than half of
executives sell, and in 14 percent of firm-years all top executive sells. Furthermore, in 54 percent
12
of firm-years at least one executive engages in a medium or large equity sale. The final column
of Panel A shows that substantial variation also exists in selling frequency at firms experiencing
above-median stock returns.
Panel B of Table 2 shows that 61 percent of executives sell equity at least once while in my sample.
While some of these sales are small, 44 percent of executives engage in a medium or large sale at
least once, and 23 percent of executives do so every other year or more often. Additionally, 20
percent of executives buy equity on the open market.
Panel C presents statistics on the size of annual equity transactions (conditional on a transaction
occurring). The median amount of equity sold is $558,000, equal to 15 percent of the dollar value
of total firm holdings. Total incentives decrease by seven percent at the median (10 percent on
average) following equity sales. On the other hand, executives purchase equity in much smaller
amounts. The median purchase is $56,000, equal to four percent of total holdings.
3
Empirical Specification
In this section I develop a regression specification based on the hypothesis that boards use annual
pay to adjust executives’ incentives toward the optimal level. The specification implies that
boards should grant a higher proportion of equity pay to executives who sold more equity in the
previous year. I also describe several variables which may confound my empirical analysis by
causing executives to sell equity and simultaneously reducing optimal incentives, and explain my
identification strategy for accounting for some of these variables.
3.1
Regression Model
My empirical specification is for executive m at firm f at time t. The specification includes the
following variables:
13
• paymf t : a measure of equity pay granted at time t.
∗
• αmf
t : the executive’s optimal incentive level at the beginning of year t.
• αmf t : the executive’s actual incentives in year t, before the board determines annual pay.
• ~xf t : a vector of time-varying firm-level characteristics.
• ~zmt : a vector of time-varying executive-level characteristics.
• λf , µt , δm : firm, year, and executive fixed effects.
The theoretical benchmark for my regression model is based on Holmstrom & Milgrom (1987),
who predict that each executive has a unique level of incentives which most effectively mitigates
agency problems, and that boards grant pay to set executives’ incentives at this optimal level.
My empirical model does require boards to set incentives precisely at this level, but does assume
that annual pay grants adjust incentives back toward the optimum.
I model an executive’s annual compensation as a linear function of the difference between optimal
incentives and the executive’s actual incentives from holdings of firm equity:17
∗
paymf t = β0 + β1 · (αmf
t − αmf t ) + λf + δm + µt + mf t
(1)
∗
The optimal incentive level αmf
t is unobservable empirically, because it depends on variables such
as the curvature of the utility function. However, theory predicts that it varies with firm- and
∗
executive-level characteristics, so I model αmf
t as a linear function of these variables:
∗
αmf
xf t + ηm,2 · ~zmt + emf t
t = ηm,0 + ηm,1 · ~
(2)
17
The literature finds that compensation also depends directly on firm-level characteristics such as tax and
accounting considerations. For example, since 1993 the limit on corporate tax deductions for non-performancebased executive pay has been one million dollars, causing some firms to award pay raises primarily through
increased equity pay (Perry & Zenner (2001), Yermack (1995)). Additionally, prior to 2006 firms did not have to
write down in their accounting statements the cost of granting at-the-money stock options, and some firms granted
executives large amounts of stock options in order to avoid reducing their earnings (Carter & Lynch (2001), Core
& Guay (1999)).
While ~xf t does not directly enter my specification for annual pay, equations (1) and (2) are both linear, so the
final regression specification will not depend on how firm and executive-level characteristics enter the model.
14
Substituting (2) into (1) and rearranging terms yields:
paymf t = βm,1 · ~xf t + βm,2 · ~zmt + β3 · αmf t + λf + µt + δm + ˜mf t
(3)
where βm,1 = β1 · ηm,1 , βm,2 = β1 · ηm,2 , β3 = −β1 , and ˜mf t = β1 · emf t + mf t . Equation (3) states
that annual compensation depends on firm- and executive-level variables which affect optimal
incentives, as well as the executive’s actual incentives from firm holdings at the start of year t.
This is similar to the model developed by Core & Guay (1999), who also hypothesize that new
equity incentives depend on the deviation between an executive’s optimal and actual incentives.
The m subscript on βm,1 allows firm-level characteristics to have a different effect on the optimal
incentives (and hence compensation) of each executive at firm f . I construct the coefficient in
this manner to clarify the key assumption behind my identification strategy below.
Optimal incentives depend partly on executive-level characteristics which are unobservable but
fixed over time, such as utility function curvature. I therefore formulate my specification in terms
of changes in pay by taking the first difference of (3):
∆paymf t = βm,1 · ∆~xf t + βm,2 · ∆~zmt + β3 · ∆αmf t + µt + µt−1 + ∆˜mf t
(4)
where ∆paymf t = paym,f,t − paym,f,t−1 , etc.
I call equation (4) the across-firm model, and in the next section I use this specification to estimate
the relationship between pay changes and equity sales for top executives across firms. This model
states that changes in equity pay depend on two channels: 1) changes to the optimal incentive
level caused by firm- and executive-level characteristics, and 2) changes to the executive’s actual
incentives due to changes in total holdings in the firm. Because equity sales decrease total holdings
and reduce the executive’s actual incentives, the primary coefficient of interest is β3 .
15
3.2
Identification Strategy
The across-firm model predicts that boards should grant larger equity pay increases to executives
whose actual incentives are below the optimal level, but this does not imply that the board should
always grant more equity following a sale. This is because the executive’s optimal incentive level
may also decrease at the time of the sale, and actual incentives may remain at the optimal level.
In such cases, the board’s optimal response should be to leave pay unchanged.
Various elements of ∆~xf t and ∆~zmt are likely unobservable empirically, and are therefore in the
error term of any cross-sectional regression. If these variables are correlated with equity sales
and also cause equity pay to decrease, my empirical estimate of β3 will be biased downward. If
the effect of the variables is large relative to the size of the equity sale, β3 may be statistically
indistinguishable from zero.
One example of such a firm-level change is a reduction in investment opportunities or stock price
volatility, which may increase the board’s monitoring ability and simultaneously cause executives
to sell equity. In such cases, the board may not need to rely as much on equity pay to convey
incentives (Prendergast (2002)). Alternatively, executives may sell equity in anticipation of poor
future firm performance, and the board may reduce executives’ total and equity pay upon learning
about the poor performance. (This can occur before the market learns about firm performance
and updates stock prices.) Executive-level variables can also bias my empirical estimate of β3 .
For example, an executive may begin selling equity several years after being hired by the firm,
but at this time the board may also learn information about the executive that helps improve its
monitoring.
Such confounding variables show that an empirical estimate of no relationship between equity
sales and firm performance might be consistent with theory. To account for some of these variables, I compare executives who sell equity to executives at the same firm who in the same year
do not sell equity. I derive this specification by replacing βm,1 · ∆~xf t in the across-firm model
16
with a firm-year fixed effect µf t :
∆paymf t = µf t + βm,2 · ∆~zmt + β3 · ∆αmf t + ∆˜mf t
(5)
I call (5) the within-firm model. In this model, β3 measures the effect of changes in total incentives
on the change in executive m’s equity pay relative to other executives at the same firm. Top
executives who do not sell equity serve as a control group for executives at the same firm who
do.
This specification is invalid if the pay changes of non-selling executives would differ from the
pay changes of selling executives, in the counterfactual case in which they do not sell. In other
words, any variable that causes an executive to sell equity should not affect that executive’s pay
differently than the pay of non-selling executives at the same firm. My key identifying assumption
is that βm,1 is the same for each top executive m at firm f . Otherwise, µf t 6= βm,1 · ∆~xf t , and
some elements of ∆~xf t will be in the error term of a regression based on (5). If these elements
are correlated with ∆αmf t , they will bias my empirical estimates.
I argue that this identifying assumption is likely valid for the top executives in my sample. Most
of these executives likely make decisions that affect the entire firm, so it is reasonable to assume
that firm-level changes affect their pay in a similar manner. Boards also set the annual pay for
each of these top executives.
However, some executives in my sample manage divisions or subsidiaries of the firm, or are
responsible for specialized tasks such as directing product development. Boards may base the
pay of these executives primarily on the performance of the unit of the firm which they oversee
(Aggarwal & Samwick (2003)). To account for this possibility, I re-estimate my results using a
sample of just “C-Suite” executives whose decisions affect the entire firm. All specifications also
include an indicator for executives whose titles indicate a specialized role in the firm. Additionally,
I repeat the estimation of each specification using a sample of all top executives except the CEO.
This accounts for the possibility that some unobserved firm-level changes affect only the CEO’s
17
pay. For example, because the CEO is often the public face of the company, the board may
decrease just the CEO’s pay following a failed acquisition. I do not report these results, but they
do not differ substantially from my main results in the next section.
My identification strategy does not account for unobservable elements of ∆~zmt , some of which
likely affect ∆αmf t . I conduct numerous robustness checks in Section 5 using subsamples of
executives for whom individual characteristics are unlikely to bias regression results, and conclude
that the most pertinent elements of ∆~zmt do not appear to bias my results.
4
Main Results
In this section, I present my main results on the relationship between equity sales and subsequent
changes in equity pay. I first present results in tables 3 and 4 based on the across-firm model (4),
which compare pay changes for all executives who sell equity to all executives who do not. After
examining these cross-sectional results, I then implement my identification strategy of compare
pay changes for selling and non-selling executives at the same firm.
4.1
Results for Across-firm Model
In Table 3, the specifications measure equity sales using the sale indicators, and in Table 4 using
the decrease in total incentives. In each table, the dependent variable in the first two regressions
is the percentage change in incentives from new equity pay. In the next two regressions, it is the
change in pay ratio, and in the last two regressions it is the change in total incentives. For each
dependent variable, I compute results first using all top executives, and second using just CEOs.
Each regression includes various firm and executive-level control variables, and fixed effects for
year, industry, and the executive’s role. The null hypothesis in both tables (and throughout this
paper) is that no relationship exists between equity sales and subsequent pay.
If boards are responding optimally to equity sales, the coefficients on all of the sale indicators
18
should be positive and increasing in sale size. Positive coefficients in the first two regressions of
Table 3 would indicate that selling executives receive larger increases in equity incentives than
the excluded group of executives who do not sell equity. Positive coefficients in the next two
regressions would indicate that boards shift the composition of selling executives’ pay toward
more equity, and in the last two regressions that boards grant larger increases in total incentives
to selling executives.
Instead, in each regression in Table 3 the coefficients on each sale indicator are either statistically
indistinguishable from zero or negative. In addition to having the wrong sign, the coefficient on
large sales is smaller than the coefficients on small and medium sales in five of the six models.
For example, the first regression indicates that executives who sell medium amounts of equity on
average receive a 1.5 percent smaller increase in new incentives than executives who do not sell,
while executives who sell large amounts of equity on average receive a 3.7 percent smaller increase
in new incentives (statistically significant at the five percent level). The results show that no
relationship exists between equity sales and changes in pay composition, and that executives who
sell large amounts of equity receive smaller increases in total incentives than executives who sell
little or none. The estimates from regressions using just CEOs are similar to those for the entire
sample.
In Table 4, the measure of equity is “Pct. Drop Incentives”, the percentage change in total
incentives due to an equity transaction. For this variable, larger values indicate larger decreases
in total incentives due to sales, so the coefficient should also be positive if boards are replenishing
incentives. Unlike in the previous table, the relationship between sales and subsequent pay
in Table 4 is positive and statistically significant in some regressions. The second regression
indicates that for each 1 percent drop in total incentives due to an equity sale, a CEO receives a
0.13 percent larger increase in new incentives. This estimate is statistically significant at the five
percent level (for all top executives, the coefficient is smaller and statistically insignificant). Also,
the last two columns indicate a replenishment rate of 8 percent of total incentives (15 percent for
CEOs), which is statistically significant at the 1 percent level. However, the results again show
no relationship between sales and changes in pay composition.
19
Although the results in Table 4 provide some evidence that selling executives receive larger
pay increases than non-selling executives, the magnitude of replenishment is small at best. An
executive’s total incentives are substantially larger than new equity incentives from annual pay, so
the 0.13 percent increase in new incentives that selling CEOs receive offsets only a small portion
of total incentives lost from a sale. The second regression of Table 4 implies that the median CEO
receives just a $0.01 increase in total incentives for each $1 decrease in total incentives from equity
sales, relative to CEOs who do not sell.18 The replenishment rate of 0.08 in the fifth regression
of Table 4 is also substantially below 1, the value that would indicate full replenishment.
Furthermore, all specifications in tables 3 and 4 strongly reject alternative null hypotheses that
boards restore a substantial amount of incentives. The specifications in Table 3 reject with 99
percent confidence the alternative hypothesis that the coefficient on any of the sales indicators
is greater than or equal to 5 (which would indicate that selling executives receive a 5 percent
larger increase in new equity pay, or about a 0.4 percent larger increase in total incentives for the
median executive). My results in Table 4 also reject with 99 percent confidence the alternative
hypothesis that the coefficient on “Pct. Drop Incentives” is 1 in any specification.
Some firm- and executive-level variables do explain changes in equity pay. In both tables, larger
firms and those with higher stock returns grant larger increases in equity pay. Firms with higher
stock volatility or larger increases in Tobin’s Q receive larger increases in total incentives. Interestingly, junior executives on average receive larger increases in annual and total incentives
than more senior executives, which is not consistent with the hypothesis that senior executives are
more likely to become entrenched, and hence need stronger incentives. The results also show that
executives receive large pay increases upon being promoted — executives receive an 18 percent
larger equity pay increase when promoted to the C-Suite and 44 percent larger increase when
promoted to CEO — and substantially lower pay after stepping down from CEO.
18
The median CEO has total incentives of $246,000 and receives new annual incentives of $15,700. Therefore,
the CEO receives $20.4 in new incentives for each $2,460 decrease in total incentives, or $0.008 for each $1 decrease.
20
4.2
Results for Within-firm Model
I now present results based on the empirical specification (5), which implements my identification
strategy of comparing executives at the same firm. These results, unlike those in tables 3 and
4, are unlikely to be biased by unobservable firm-level variables. All regressions in this section,
and throughout the rest of the paper, include firm-year fixed effects. I first test whether selling
executives receive larger increases in annual equity pay, and then whether boards replenish selling
executives’ total incentives.
Table 5 tests whether selling executives receive larger increases in new equity pay than nonselling executives using the three indicators for small, medium, and large sales. The dependent
variable in the first two regressions is the percentage change in new incentives, and in the next
two regressions it is the change in pay ratio. For both dependent variables, I first estimate
specifications using all top executives, and then repeat the analysis using a subsample of just
C-Suite executives.19
Positive coefficients in Table 5 would indicate that selling executives receive larger increases
in equity pay than non-selling executives at the same firm. Instead, the coefficients on all of
the sale indicators are negative. For example, the first regression shows that executives who
sell large amounts of equity subsequently receive 0.7 percent smaller increases in new incentives
than executives at the same firm who do not sell. All of these coefficients are statistically
indistinguishable from zero. The sign of the coefficient on equity purchases in the first regression
is consistent with efficient contracting — boards may optimally grant less equity to executives
who purchase stock on their own — but the relationship is also statistically insignificant. The
coefficients on the sales indicators are similar when I restrict the sample to just C-Suite executives,
indicating that my results are not affected substantially by boards setting pay differently for
specialized executives.
The structure of Table 6 is similar, but equity transactions are measured using “Pct. Drop
19
In all specifications with C-Suite executives, I restrict my sample to firm-years in which at least three such
executives are listed in Execucomp.
21
Incentives”. The coefficient in each specification is small in magnitude and statistically indistinguishable from zero. For example, the fourth regression indicates that a C-Suite executive
who sells equity receives a 0.02 percent larger increase in pay ratio than a non-selling C-Suite
executive at the same firm. This result implies that an executive who sells all of his equity would
receive a two-percent increase in pay ratio, from 36.5 to 38.5 percent of total pay for the median
executive.
The specifications in tables 5 and 6 cannot reject the null hypothesis that selling executives
receive the same pay changes as non-selling executives at the same firm, but all specifications
reject alternative hypotheses of boards granting more equity pay to selling executives. In Table 5,
all specifications reject with 99 percent confidence the alternative hypothesis that the coefficient
on the sale indicators is greater than or equal to five, and in Table 6 the specifications reject the
alternative null hypothesis that the coefficient on “Pct. Drop Incentives” is equal to 1.
The control variables in the two tables show that junior executives receive larger increases in
equity incentives, and a greater shift in pay composition toward more equity, during their first
four years at the firm (the latter result is not statistically significant for C-Suite executives).
Executives also receive substantial pay increases following promotion to the C-Suite or CEO, and
when becoming board chairman. For example, an executive receives a 15 larger increase in equity
incentives in the year of promotion to the C-Suite than other top executives at the same firm.
Next, Table 7 estimates the degree to which boards replenish an executive’s total incentives from
holdings in the firm after an equity sale. The dependent variable in this table is the increase in
total incentives due to new equity pay. The first two specifications measure equity sales using
sale indicators, and the second two using the decrease in total incentives from equity sales. I
estimate both sets of regressions first for all top executives, and then just the C-Suite.
The first two columns show that executives who sell large amounts of equity receive about a 1.6
percent smaller increase in total incentives than executives who do not sell (2.3 percent smaller
increase for C-Suite executives). This relationship is statistically significant at the one percent
level. The next two columns show that “Pct. Drop Incentives” is positively associated with
22
subsequent increases in total incentives. The coefficient in the third column indicates that for
each 1 percent drop in total incentives due to equity sales, the executive subsequently receives
a 0.04 percent larger increase in total incentives than an executive who does not sell. This
replenishment rate rises to 0.07 for C-Suite executives. Both results are statistically significant
at the 1 percent level, but far below the coefficient of 1 which would imply full replenishment.
Furthermore, the specifications strongly reject the alternative null hypothesis that the coefficient
on “Pct. Drop Incentives” is even as high as .15.20
4.3
Discussion of results
The main results presented in this section are as follows:
1. Executives who sell equity do not subsequently receive a higher proportion of annual pay
in equity, or a larger increase in annual equity incentives, than executives at the same firm
who do not sell.
2. Boards replenish at most 7 percent of the decrease in incentives caused by an equity sale.
The data strongly reject full replenishment by the board.
3. The results are the same for C-Suite executives, indicating that my estimates are likely not
affected by differential pay changes across executives.
Given these results, it seems unlikely that the true replenishment rate of incentives after equity
sales is 1. Any confounding variable which remains in the error term of the within-firm regressions
would have to cause a substantial decrease in optimal incentives whenever an executive sells
equity. Such a variable would have to reduce the optimal replenishment rate of close to 1 to the
20
The discrepancy in Table 7 between the positive coefficient on “Pct. Drop Incentives” and negative coefficients
on “Large Sale” is largely driven by executives with small holdings in the firm. When executives with the smallest
quintile of total incentives at the start of year t − 1 (about $17,500 in total incentives) are excluded from the
second and fourth regressions, the coefficient on “Pct. Drop Incentives” falls to about zero and becomes statistically
insignificant.
23
empirically observed estimate of 0.07. I conduct additional tests in the next section to account
for some possible confounding variables, and my results do not change substantially.
The economic implication of these results is that the incentives of executives who sell equity
decrease relative to those of executives at the same firm who do not sell. It is important to
emphasize the relative interpretation of these results — the board either allows the incentives
of selling executives to decrease below the optimal level, or allows the incentives of non-selling
executives to increase above the optimum. My results imply that one group of executives has
suboptimal incentives, but they do not indicate which it is.
In either case, shareholders seemingly would be better off according to agency theory if the
board adjusted incentives toward the optimal level. If boards allow some executives’ incentives
to become too low after equity sales, these executives might pursue private benefits instead of
value-maximizing decisions. On the other hand, if some executives do not sell equity for years and
accumulate a larger-than-necessary stake in the firm, then the board could reduce compensation
cost by granting these executives more fixed pay and less equity. Both shareholders and executives
would benefit from this shift in pay composition — underdiversified executives should prefer to
receive cash to equity pay, and because the executive is exposed to less risk the board could
reduce total compensation. On the other hand, it is difficult to explain how granting similar pay
to selling and non-selling executives maximizes shareholder value.
5
Robustness Checks
In this section I test whether the reason that boards do not restore executives’ incentives following
equity sales is because it is in shareholders’ best interests not to do so. While my empirical
specification accounts for firm-level changes which affect the optimal incentives of all of a firm’s
top executives, some unobservable variables may cause equity sales and simultaneously reduce the
optimal incentives of only selling executives. Such variables could bias downward my empirical
estimates from the previous section, leading me to falsely conclude that no relationship exists
24
between equity sales and subsequent pay. This section identifies several such variables, and
presents results from robustness tests which use subsamples of executives who are likely unaffected
by the variables.
I also address another potential concern with my empirical specification: the assumption that
boards adjust executives’ compensation after observing whether sales occurred in the previous
year. Boards may instead grant some executives a large amount of equity because they anticipate that sales will occur in the future (Ofek & Yermack (2000)), and then leave compensation
unchanged after the sales occur. To test this possibility, I conduct robustness tests examining
whether boards adjust compensation after sales that they are less likely to anticipate. My main
results from the previous section are robust to both sets of tests.
5.1
Executive Tenure and Board Learning
When a board hires an executive, it may not know with certainty his propensity to consume
private benefits. A board that is contracting efficiently may form an expectation about the executive’s type, and grant compensation that conveys optimal incentives based on this expectation.
The board may then may revise its belief about the executive after observing his first few years
at the firm. If the board learns that it initially overestimated the executive’s willingness to shirk
or underestimated his risk aversion, it may reduce his optimal incentives. This may occur just
as the executive is beginning to sell his holdings in the firm.
To account for the possibility that many equity sales occur as the board is learning new information about selling executives, I restrict my sample to cohorts of executives — groups of three
or more executives who joined a firm’s top management in the same year, or who have served
together in top management for at least four years.21 Executives in the same cohort have similar
21
Specifically, I identify all firm-years in which at least three executives are listed for the first time as top
executives at that firm. (These executives did not necessarily join the firm at the same time; some may have
previously worked in lower-level management positions.) I define this group of executives as a cohort in all
subsequent years in which at least three of the executives stay at the firm.
I cannot observe the tenure of the top executives listed in the firm’s first year in the Execucomp database, if the
firm is listed in Execucomp in 1992, the first year of database coverage. However, starting in 1996, I am certain
25
tenure, so if the board learns valuable information about one executive shortly after he joins the
firm’s top management, it is likely to also learn information about other executives who joined
at the same time. Forty-three percent of executives are part of a cohort at some point during my
sample, and 45 percent of firm-years in my sample have a cohort of executives.
Table 8 presents results using just this subsample of executives. The table presents results for
two sets of regressions, the first using indicator variables to measure equity sales, and the second
using the decrease in total incentives. Each set of regressions also includes results for all three
measures of pay changes.
The results are similar to my results for all top executives. The coefficients on the sale indicators
are all either statistically insignificant or, in the case of the relationship between medium or
large sales and changes in total incentives, have the wrong sign. The coefficients on “Pct. Drop
Incentives” are all close to zero and statistically insignificant, including the replenishment rate
of 0.02.
These results indicate that differences in the amount of information that the board possesses
on executives are not affecting my estimates. The board should update its information on all
executives in a cohort at approximately the same time, so it is unlikely that only the optimal
incentive level of selling executives is revised downward.
One possibility not addressed in Table 8 is that equity sales reveal an executive’s preferences to the
board. A large equity sale could signal to the board that it has exposed the executive to too much
risk, and the board may choose to grant the executive lower incentives after the sale. I argue that
this is unlikely to be the reason why I observe no relationship between equity sales and subsequent
pay changes. First, it implies that most executives are initially granted too much equity, and
that almost all sales are in the best interests of shareholders. Second, in unreported results I
repeat my analysis on a subsample of executives who sell equity repeatedly, and my results do
not change. The executive’s first equity sale may signal that the board mistakenly granted him
that any of the initial executives who are still listed at the firm have worked there for at least four years. I define
these executives as a cohort if at least three of them are still with the firm after four years. Finally, if a firm has
just begun to trade publicly, I define its initial top executives as a cohort.
26
too much equity, but this does not explain why boards do not respond to subsequent sales.
5.2
Wealth effects
If an executive’s utility is concave in wealth (e.g. a constant relative risk aversion utility function),
the executive will become more risk averse following a decrease in wealth held outside the firm.
If the executive does not sell equity, a higher fraction of his remaining wealth will be held in the
firm, and this may cause the executive to begin undertaking decisions which shareholders consider
to be suboptimally conservative.22 Allowing the executive to reduce his holdings in the firm may
therefore be in shareholders’ best interests. If a large fraction of executives in my sample sell
equity after experiencing wealth shocks, this may explain the observed lack of replenishment.
I cannot directly control for wealth changes, because executives do not need to publicly report
any information on personal wealth held outside the firm. Instead, I identify a subsample of
executives who likely did not suffer major shocks to outside wealth, using data on contributions
to non-qualified deferred compensation accounts. NQDC accounts are similar in purpose to
401(k) accounts.23 Executives allocate a portion of their annual pay (usually salary or a cash
bonus) into the account, and the firm invests the savings on behalf of the executive. Income
taxes are only paid when money is withdrawn from the account.
About half of the firms in my sample offered NQDC plans in 2006, the first year in which the SEC
required firms to provide detailed data on deferred compensation.24 The median executive who
22
A wealth shock may also increase an executive’s incentives to work hard, even after he sells some of his
equity, which could allow the board to reduce equity pay. The executive chooses optimal effort by weighing the
marginal benefit of working harder against the marginal cost. The executive’s marginal utility with respect to pay
may increase following a wealth shock, and if the marginal cost of effort does not change, the executive will work
harder. The precise effect of a wealth shock on the executive’s effort level depends on several factors, including
the degree to which a change in effort affects the executive’s pay (Ladika (2011)).
23
One of the purposes of these plans is to allow executives to bypass the $22,000 cap on annual contributions
to a 401(k) account. However, unlike the savings in a 401(k) account, NQDC accounts are not protected from
creditors in the event of bankruptcy. Reda, Reifler & Thatcher (2007) provide a detailed description of the rules
regarding these accounts.
24
Execucomp does not include a variable indicating which firms have active NQDC plans, but there are separate
variables for the amount of money contributed by the executive and firm to the account. I define a firm as having
an active plan if in 2006 at least one executive contributed money to an NQDC account, or the firm made a
contribution on behalf of at least one executive. By this definition, 49% of firms in my sample offered plans in
27
contributes to an NQDC account defers 17 percent of annual salary and has savings of $564,000
in the account (equal to about 13 percent of the value of firm equity holdings). Therefore,
executives who contribute to NQDC accounts voluntarily delay receipt of a significant fraction
of their annual income. I argue that such executives are less likely to have recently experienced
major wealth or liquidity shocks
I conduct analysis on two subsamples: 1) executives who contributed money to an NQDC account
in the same year they sold equity, and 2) executives whose NQDC accounts are equal to at least
10 percent of the value of their firm equity, and who did not withdraw money from the account,
in the year after a sale. Because data is not available before 2006, I am only able to compute
results using the last two years of my sample. I restrict analysis to firms in which at least three
executives made NQDC contributions or had substantial savings in the accounts, leaving about
300 firms for the first subsample and 400 firms for the second subsample.
The results are in Table 9. Panel A shows results for the first subsample of executives who
contribute to NQDC accounts, and Panel B shows results for the second subsample of executives
with substantial savings in their accounts. Each panel presents two sets of regressions using the
two different measures of equity sales, and I show results for each of the three measures of pay
changes.
In both panels, the coefficients on the sale indicators are either statistically insignificant or negative. In the first subsample, large equity sales are associated with a 1.6 percent increase in pay
ratio, but the relationship is not statistically significant. There is also no relationship between
“Pct. Drop Incentives” and any measure of pay changes. In particular, the replenishment rate is
also negative in this sample (but statistically insignificant). These results indicate that at least
in the final two years of my sample, wealth shocks are not the primary reason that boards do not
replenish equity holdings.
One limitation of using NQDC contributions as an indicator for the absence of a wealth shock
is that young executives who experience shocks may continue to contribute to the accounts, to
2006.
28
reduce their tax payments. In unreported tests, I re-estimate my main results using alternate
measures of wealth shocks. I exclude firms which experienced below-median stock returns in the
previous year, firms from industries which experienced poor performance (measured by annual
or two-year value-weighted industry stock returns), and firms headquartered in states which
experienced decreases in real estate prices. My results do not change substantially in these tests.
5.3
Anticipation of equity sales
If the board is forward-looking, it may know that executives will at some point sell the equity that
they receive each year. In particular, the board may anticipate that executives will sell equity
soon after it completely vests. Instead of replenishing executives’ incentives after an equity sale,
the board instead may set the flow of annual equity grants ex ante such that incentives remain
at the optimal level. In this case, the board grants new equity each year to replace the equity
that is about to vest, which the executive may soon sell.
If boards are maintaining executives’ incentives in this manner, they will not adjust pay in
response to anticipated sales. However, they should still respond to equity sales which occur
earlier than expected. I identify unanticipated equity sales as those which occur following the
early exercise of stock options, defined as the exercise of an option that has at least six years left
until expiration. Because options usually have a 10-year lifespan, most of these exercises occur
within four years of the option’s grant date, before the option has fully vested.25 Boards may be
less likely to anticipate executives’ exercises and sales of options that have not fully vested.
In Table 10, I define “Early Exercise” as an indicator equal to one if at least half of the dollar
value of equity sold by the executive in year t − 1 is from shares acquired after the early exercise
of a stock option.26 The interaction of the sale indicators and “Early Exercise” captures the
25
Options typically finish vesting three to five years after the grant date (Cadman et al. (2012)), so these
exercises mainly occur before the option has fully vested. I also try to use a seven-year threshold, and obtain
similar results.
26
Specifically, I identify the date on which an executive exercises a stock option with at least six years left until
expiration, and the number of shares acquired from the option exercise. I then measure the number of acquired
shares that the executive proceeds to sell in the same fiscal year, and the dollar value obtained from selling the
29
difference in pay changes between executives who sell equity following early option exercise and
those who engage in other types of sales. Positive coefficients on the interaction terms would
indicate that boards adjust pay more in response to unanticipated equity sales.
The results in Table 10 indicate that boards do not respond differently to equity sales following
early stock option exercise. For example, the coefficient of -2.4 on “Large Sale [t-1] * Early
Exercise” in the first column indicates that boards grant 2.4 percent smaller increases in new
equity incentives to executives who sold large amounts following early option exercise, than to
executives who engaged in other large sales. This coefficient, as all the other interaction term
coefficients, is statistically indistinguishable from zero or negative.
Another potential concern is that when boards set annual equity pay for the upcoming fiscal
year, they anticipate sales which will occur early in the year, and replenish incentives ex ante.
One reason that boards may be able to anticipate such sales is that executives sometimes draft
pre-arranged written trading plans that specify when and how many shares the executive will
sell. The SEC provides a legal defense against insider trading allegations for executives who trade
within the guidelines of such plans (Jagolinzer (February 2009)).27
Boards typically determine executives’ annual pay for year t toward the end of year t − 1. If
the board knows that the executive is planning to sell equity at the beginning of year t, it may
increase equity pay in the upcoming year instead of in the year after the sale. If many sales in
my sample are pre-arranged, I may estimate no relationship between sales in year t and pay in
year t + 1, when boards do in fact maintain optimal incentives. However, the board is less likely
to anticipate sales which occur toward the end of year t, so it should respond to such sales by
replenishing incentives in the following year.
shares. I divide this value by the total value of equity sold in the fiscal year.
27
In October 2000, the SEC established Rule 10b5-1, which states that an individual can be found to have
engaged in insider trading if he transacted equity while in possession of material information. The rule also provides
an affirmative defense against litigation for insiders who transact equity within the guidelines of a pre-arranged
trading plan. In order to qualify for the defense, the insider must develop a written plan that specifies equity
selling dates and the amounts of equity to be sold on each date, specifies a formula or algorithm for transacting
equity, or delegates the selling decision to a third party which does not possess inside information. Rule 10b5-1
grants insiders flexibility in choosing the length of the plan or the number of shares sold on each trading date
(Jagolinzer (February 2009)).
30
I account for this possibility in Table 11 by distinguishing between equity sales which occur in
the first and second halves of each fiscal year. In the table, “Small Sale Early [t]” is the change
in an indicator equal to 1 if the executive sold a small amount of equity in the first half of the
fiscal year, while “Small Sale Late [t]” is the change in an indicator for small sales in the second
half of the fiscal year (other sale indicators are similarly defined). If boards do not anticipate
sales which occur later in the fiscal year, and restore incentives after such sales, the coefficients
on the late sales should be positive and statistically significant. There may on the other hand be
no relationship between early sales and subsequent pay.
The results in Table 11 show that boards do not restore incentives after equity sales which occur
early or late in the fiscal year. For example, the coefficients on “Medium Sale Early [t]” and
“Medium Sale Late [t]” in the first column are similar in size (and both statistically insignificant),
indicating that boards grant similar decreases in equity incentives to executives who sell equity
early in the fiscal year as to executives who sell equity later in the year. The results are similar
for other types of sales and measures of pay changes. I obtain similar results when I generate
indicators for sales which occur in the first three months and last nine months of each fiscal year,
or the first nine months and last three months of each year (I do not report these results).
6
Conclusion
Theory predicts that corporate boards should set an executive’s incentives at the unique level
which most effectively mitigates agency problems, and Edmans et al. (2012) show that boards
should replenish incentives that fall below this level by granting a higher fraction of pay in equity.
An executive’s incentives may become suboptimally low when they sell a large fraction of firm
equity, but this paper shows that subsequent pay granted by the board does not replenish the
executive’s lost incentives.
The board may leave pay unchanged if the executive’s optimal incentive level falls concurrently
with the equity sale. Unobservable firm-level variables could simultaneously decrease optimal
31
incentives and cause executives to sell. To account for this possibility, I compare top executives
who sell equity to other top executives at the same firm who do not sell. I show that boards grant
similar pay to selling and non-selling executives at the same firm, and restore at most 10 percent
of a selling executive’s incentives. Boards grant similar pay not only after small equity sales,
but also following large decreases equal to 25 percent or more of an executive’s total holdings in
the firm. These results are robust to a variety of alternative explanations for why boards may
optimally allow an executive’s incentives to decrease.
If the purpose of equity pay is to incentivize executives, it is difficult to explain why boards
let some executives reduce their holdings in the firm, while at the same time maintaining the
holdings of other top executives. If shareholders are better off when an executive decreases their
holdings in the firm, why did the board grant the executive substantial equity prior to the sale?
Why does the board grant the same amount of equity to non-selling executives, whose incentives
did not decrease? Although the optimal contract can take on a variety of different forms, it is
unlikely that granting similar pay to selling and non-selling executives is optimal. This finding
suggests that our understanding of executive compensation is incomplete.
I suggest three alternative explanations for this finding, which can be tested in future research.
First, boards commonly benchmark their executives’ compensation to that of their peers (Faulkender & Yang (2010), Bizjak, Lemmon & Naveen (2008), Bizjak, Lemmon & Nguyen (2011)).
Perhaps boards place higher priority on benchmarking annual pay grants than on properly adjusting total holdings of firm equity. For example, the board may choose not to grant selling
executives a high proportion of pay in equity when rival firms grant their executives salary increases. One way to test this is to examine whether boards respond to increases in equity pay at
peer firms by increasing their own executives’ equity pay, even following large equity sales.
Second, the Managerial Power perspective predicts that boards disguise wealth extraction from
the firm by granting executives large amounts of equity pay, and allowing them to later unwind
their equity incentives. My results are consistent with boards allowing executives to choose their
level of equity holdings in the firm, but they do not prove that boards grant equity to facilitate
32
wealth extraction. This can be tested by examining whether executives who receive large increases
in equity pay, such as following a promotion or merger, subsequently sell equity and prevent their
holdings in the firm from increasing.
Third, boards may choose to grant similar pay to all top executives to promote teamwork and
collegial relations, and to avoid potentially negative repercussions of granting executives widely
varying pay. While potentially important, such internal constraints have not been incorporated
into most models of optimal executive compensation. Future research could examine the degree
to which contracting features other than pay composition vary across top executives at the same
firm.
33
References
Adams, R., Hermalin, B. & Weisbach, M. (2008), The role of boards of directors in corporate
governance: A conceptual framework and survey. NBER Working Paper 14486.
Aggarwal, R. & Samwick, A. (2003), ‘Performance incentives within firms: The effect of managerial responsibility’, Journal of Finance 58, 1613–1649.
Baker, G. & Hall, B. (2005), ‘Ceo incentives and firm size’, Journal of Labor Economics
22(4), 767–798.
Bebchuk, L. & Fried, J. (2004), Pay without Performance: The Unfulfilled Promise of Executive
Compensation, Cambridge, MA.
Bizjak, J., Lemmon, M. & Naveen, L. (2008), ‘Does the use of peer groups contribute to higher
pay and less efficient compensation?’, Journal of Financial Economics 90, 152–168.
Bizjak, J., Lemmon, M. & 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.
Cadman, B., Rusticus, T. & Sunder, J. (2012), Stock option grant vesting terms: Economic and
financial reporting determinants. Working Paper.
Carter, M. & Lynch, L. (2001), ‘An examination of executive stock option repricing’, Journal of
Financial Economics 61, 207–225.
Cook, F. W. (2009), Stock ownership guidelines — prevelance and design of executive and director
ownership guidelines among the top 250 companies, Technical report, Frederic W. Cook and
Co., Inc.
Core, J. & Guay, W. (1999), ‘The use of equity grants to manage optimal equity incentive levels’,
Journal of Accounting and Economics 28, 151–184.
Core, J. & Guay, W. (2002), ‘Estimating the value of employee stock option portfolios and their
sensitivities to price and volatility’, Journal of Accounting Research 40(3), 613–630.
34
Core, J., Guay, W. & Larcker, D. (2003), ‘Executive equity compensation and incentives: A
survey’, Federal Reserve Bank of New York Economic Policy Review pp. 27–50.
Edmans, A., Gabaix, X., Sadzik, T. & Sannikov, Y. (2012), ‘Dynamic ceo compensation’, Journal
of Finance 67(5), 1603–1647.
Faulkender, M. & Yang, J. (2010), ‘Inside the black box: The role and composition of compensation peer groups’, Journal of Financial Economics 96, 257–270.
Grossman, S. & Hart, O. (1983), ‘An analysis of the principal-agent problem’, Econometrica
51(1), 7–45.
Hall, B. & Liebman, J. (1998), ‘Are ceos really paid like bureaucrats?’, Quarterly Journal of
Economics 113, 653–691.
Holmstrom, B. & Milgrom, P. (1987), ‘Aggregation and linearity in the provision of intertemporal
incentives’, Econometrica 55(2), 303–328.
Holmstrom, B. & Milgrom, P. (1991), ‘Multitask principal-agent analyses: Incentive contracts,
asset ownership and job design’, Journal of Law, Economics and Organization 7, 24–52.
Jaffe, J. (1974), ‘Special information and insider trading’, Journal of Business 47, 410–428.
Jagolinzer, A. (February 2009), ‘Sec rule 10b5-1 and insiders’ strategic trade’, Management Science .
Jensen, M. & Meckling, W. (1976), ‘Theory of the firm: Managerial behavior, agency costs and
ownership structure’, Journal of Financial Economics 3(4), 305–360.
Jensen, M. & Murphy, K. (1990), ‘Performance pay and top-management incentives’, Journal of
Political Economy 98, 225–264.
Jenter, D. (2002), Executive compensation, incentives, and risk. Working Paper.
Ladika, T. (2011), Amplification of shocks and managerial risk aversion. Working Paper.
35
Morck, R., Schleifer, A. & Vishny, R. (1988), ‘Management ownership and market valuation’,
Journal of Financial Economics 20, 293–315.
Murphy, K. (1999), ‘Executive compensation’, Handbook of Labor Economics, eds. O. Ashenfelter
and D. Card. 3.
Murphy, K. (2003), ‘Stock based pay in new economy firms’, Journal of Accounting and Economics 34, 129–147.
Ofek, E. & Yermack, D. (2000), ‘Taking stock: Equity-based compensation and the evolution of
managerial ownership’, Journal of Finance 55, 1367–84.
Perry, T. & Zenner, M. (2001), ‘Pay for performance? government regulation and the structure
of compensation contracts’, Journal of Financial Economics 62, 453–488.
Prendergast, C. (2002), ‘The tenuous trade-off between risk and incentives’, jpe 110(5), 1071–
1102.
Reda, J., Reifler, S. & Thatcher, L. (2007), The Compensation Committee Handbook, 3rd Edition,
Hoboken, NJ.
Ross, S. (1973), ‘The economic theory of agency: The principal’s problem’, aer 63(2), 134–139.
Ross, S. (2004), ‘Compensation, incentives, and the duality of risk aversion and riskiness’, Journal
of Finance 59(1).
Seyhun, H. (1986), ‘Insiders profits, costs of trading, and market efficiency.’, Journal of Financial
Economics 16, 189–212.
Yermack, D. (1995), ‘Do corporations award ceo stock options effectively?’, Journal of Financial
Economics 39, 237–269.
36
7
Data Appendix
This section provides detailed definitions of the variables I use in my empirical tests.
Firm-level Variables
Variable
Definition
Assets [t]
AT[t]
Compustat
Firm Q [t]
(PRCC[t]*CSHO[t] + LT[t]) / AT[t-1]
Compustat
Book-Market [t]
SEQ[t] / (PRCC[t]*CSHO[t] + LT[t])
Compustat
Change Assets [t]
(AT[t] − AT[t-1])/ AT[t-1] ∗ 100
Compustat1
Change Firm Q [t]
(Firm Q[t] − Firm Q[t-1])/ Firm Q[t-1] ∗ 100
Compustat1
Log Stock Return [t]
Stock Volatility [t]
Sales Growth [t]
Net Income Growth [t]
Tech Firm
h
ln 1 +
(PRCC[t]+DVPSX[t]−PRCC[t-1])
PRCC[t-1]
Data Sources and
Notes
i
Compustat1,2
St. deviation of past 24 months’ returns
Compustat3
(SALE[t] − SALE[t-1]) / SALE[t-1]
Compustat1
(NI[t] − NI[t-1]) / SALE[t-1]
Compustat1
Indicator equal to 1 for tech firm
1
Murphy (2003)
This variable is winsorized at the 5th and 95th percentiles.
All stock prices are adjusted for stock splits. Specifically, PRCC[t] in the definition is actually PRCC[t] /
ADJEX[t].
3
Monthly returns are defined as ln[ ( PRCCM[t]/ADJEXM[t] + DVPSXM[t]/ADJEXM[t] +
CHEQVM[t]/ADJEXM[t] ) / ( PRCCM[t-1]/ADJEXM[t-1] )*100]. I set Stock Volatility [t] equal to missing
if monthly stock returns are not available for at least 12 of the 24 months preceding month t.
2
37
Executive-level Variables
Variable
Definition
Notes
Junior Executive
Indicator equal to 1 if executive is
in first four years of tenure
For executives working at firm
when it enters Execucomp, indicator missing until executive’s fifth
year in sample
Promoted CEO[t]
Indicator equal to 1 if executive
became CEO in year t
Stepped Down CEO[t]
Indicator equal to 1 if executive
left CEO post in year t
Promoted Chair[t]
Indicator equal to 1 if exec. became chairman in year t
Only consider chairmen who are
not currently CEO, CFO, or COO
Promoted C-Suite[t]
Indicator equal to 1 if executive
promoted to C-Suite in year t
C-Suite executive is any with
“chief”
in
TITLEANN
in
Execucomp
CEO
Indicator equal to 1 for CEO
CEOANN equals
Execucomp
CFO
Indicator equal to 1 for CFO
TITLEANN contains “CFO” or
“chief financ”.
COO
Indicator equal to 1 for COO
TITLEANN contains “COO” or
“chief operating”.
Chair
Indicator equal to 1 for nonexecutive Chairman of board
TITLEANN contains “chairman”
or “chmn”, and not currently
CEO, CFO, or COO
Senior V.P.
Indicator equal to 1 for Vice President
TITLEANN contains “exec. v-p”,
“executive vp”, “senior vp”, “sr.
v-p”, or “v-p”.
Specialized Exec.
Indicator equal to 1 if executive
has specialized role at firm
4
4
“CEO”
in
I set this indicator equal to 1 if TITLEANN contains one of the following terms: “marketing”, “sales”, “manufacturing”, “research & development”, “product development”, “engineering”, “merchandis”, “international”,
“regional”, “division”, “group”, “subs”.
38
Measures of pay changes
Variable
Pct. Chg.
tives[t]
Definition
New Incen-
Chg. Pay Ratio[t]
Chg. Total Incentives[t]
5
6
Data Sources and Notes
New Incentives[t]−New Incentives[t-1]
New Incentives[t-1]
Equity Pay[t]
Annual Pay[t]
−
Equity Pay[t-1]
Annual Pay[t-1]
New Incentives[t]
Total Incentives[t-1]
∗ 100
Execucomp, Compustat5,6
∗ 100
Execucomp5,7
Execucomp, Compustat5,8
∗ 100
This variable is winsorized at the 5th and 95th percentiles.
X
Incentives from new equity granted in year t are St · Pt /100 +
Ni,t · ∆i · Pt /100, where St is the number
i
of regular shares granted in year t, Pt is the grant-date stock price, Ni,t is the number of stock options in each
new grant i made to the executive in year t, and ∆i is the Black-Scholes option delta. For new incentives, the
formula inputs are: 1) the grant-date stock price; 2) the number of days between the end of the previous fiscal
year and the option expiration date; 3) stock return volatility, which is defined in the same way as “Volatility[t]”
(see definition in the “Firm-Level Variables” table); 4) the yield on the 10-year U.S. treasury bond in fiscal year
t − 1; and 5) the value of dividends granted in fiscal year t − 1 divided by the end-of-year stock price. Option
grant data is from Execucomp, while stock price and dividend data is from Compustat. For any executive, new
incentives are set to missing if the delta of any of his new option grants is undefined.
7
“Equity Pay[t]” is the dollar value of new equity pay granted in year t, and is defined using Execucomp
data as RSTKGRNT + OPTION AWARDS BLK VALUE (OPTION AWARDS FV + STOCK AWARDS FV
starting in 2006). “Annual Pay[t]” is the dollar value of new total pay granted in year t, and is defined as TDC1
in Execucomp.
8
For “Total Incentives[t-1]”, I first compute incentives from shares by multiplying the total number of beneficially owned shares at the end of fiscal year t − 2 (SHROWN EXCL OPTS[t-2] in Execucomp) by the stock price
at the end of fiscal year t − 2 (PRCC[t-2] in Compustat). I add incentives from stock options by multiplying the
number of exercisable and unexercisable options at the end of fiscal year t − 2 (OPT UNEX EXER NUM[t-2] and
OPT UNEX UNEXER NUM[t-2]) by their respective Black-Scholes option deltas. I use the same inputs as for
“New Incentives”. However, prior to 2006 I need to estimate the option exercise prices and time to expiration.
To do this, I follow the procedure developed by Core & Guay (2002) with slight modifications. Finally, I add
incentives from new equity granted at the start of fiscal year t − 1.
39
Measures of equity sales
Variable
Definition
Data Sources and Notes
Small Sale [t-1]
Change from year t-2 to t-1 in
an indicator equal to one if dollar value of equity sold is less
than 10 percent of total equity
holdings.9
Thomson Insiders for equity
sales, Execucomp for equity
holdings.
Medium Sale [t-1]
Same, except sale size between
10 and 25 percent.
Thomson
Execucomp.
Insiders
and
Large Sale [t-1]
Same, except sale size greater
than 25 percent.
Thomson
Execucomp.
Insiders
and
Purchase [t-1]
Change from year t-2 to t-1 in an
indicator equal to one if the executive purchased equity on the
open market.10
Thomson
Execucomp.
Insiders
and
Pct. Drop Incentives [t-1]
(Total
incentives[t-1]
had
exec.
not sold equity —
Total
Incentives[t-1])/(Total
Incentives[t-2])*100.
Thomson
Insiders
Execucomp.11
and
9
More specifically: In each year, I generate an indicator “Sale [t]” equal to one if (dollar value of equity sold
in year t) / (dollar value of total equity holdings in year t) is less than .1. I set “Sale [t]” equal to zero if the
executive is in my sample but reports no transactions in Thomson Insiders for year t. I then define “Small Sale
[t-1]” as “Sale [t-1]” - “Sale [t-2]”. Note that this variable is missing for the executive’s first two years in the
sample.
I define the dollar value of equity sold as follows: For stock transactions (transaction code “S” or “F” in
Thomson Insiders), it is equal to the transaction date stock price multiplied by number of shares sold (“tprice”
multiplied by “shares” in Thomson Insiders). For stock sales following option exercise, it is equal to “tprice”
multiplied by “shares”, minus the cost of exercising the options (“tprice” multiplied by “shares” when transaction
code equals “M”). I count the number of shares sold and options exercised during the year, and stop deducting
the exercise cost once the number of options exercised exceeds the number of shares sold. I skip all transactions
with missing values of “tprice” or “shares”, or transactions of indirectly held equity (“ownership” equal to “I”).
I define the dollar value of total equity holdings as follows: using Execucomp data, I define the
value of year-end holdings as SHROWN EXCL OPTS[t]*PRCC[t] + OPT UNEX EXER EST VAL[t] +
OPT UNEX UNEXER EST VAL[t]. The dollar value of total equity holdings in year t is this number, plus
the dollar value of equity sold during year t, minus the dollar value of equity purchased in year t.
10
The value of open market purchases is “tprice” multiplied by “shares” when transaction code equals “P” in
Thomson Insiders. I skip all transactions with missing values of “tprice” or “shares”, or transactions of indirectly
held equity (“ownership” equal to “I”).
11
“Total Incentives[t-1]” is the executive’s total incentives from equity owned at the end of year t-1, and is
measured in the same way as for the variable “Chg. Total Incentives[t]” (see the definition in the “Measures of
pay changes” table in this appendix). To estimate “Total incentives[t-1] had exec. not sold equity”, I first collect
data on all of the executive’s stock and option holdings at the start of year t-1. These are the same holdings used
to calculate incentives from the end of the previous year, “Total Incentives[t-2]”. I then add to the executive’s
portfolio equity granted during year t-1, and remove any stock options that expire during the year. I then calculate
40
the total incentives from this hypothetical portfolio of equity holdings using stock price data from the end of year
t-1. The portfolio is hypothetical in the sense that it ignores any equity that the executive purchases or sells
during year t-1, instead estimating what incentives would have been had the executive not engaged in any equity
transaction.
41
Table 1. Descriptive Statistics: Executive Roles and Compensation
This table presents descriptive statistics for executives in my sample. My sample is all firms in the Execucomp and Thomson Insiders databases, and
covers the year 1996 to 2007. All statistics are at the firm-year level. In Panel A, “Other C-Suite" is an executive with "chief" in his title who is not a CEO,
CFO, or COO. "Non-Exec. Chairman" is a chairman of the board who is also not CEO, CFO, or COO. "Specialized Exec." is an executive whose title
includes terms like "division", "subsidiary", "marketing", or "research and development". Junior Exec. ia one who has been in the firm's top management
for less than five years. In Panel B, "Equity Pay" is the sum of annual stock and stock option compensation. "Pay Ratio" is the ratio of annual equity to
annual total pay. "Equity Holdings" is the dollar value of firm stock, valued at fiscal-year-end price, plus the Black-Scholes value of all stock options.
"Total Incentives" is the year-end incentives from total equity holdings in the firm. Incentives are defined as the change in dollar value of equity for a onepercent change in stock price, as in Baker & Hall (2005). "New Incentives" is the incentives conveyed by annual equity compensation. In Panel C and D,
"Pct. Chg. New Incentives" is the percentage change in new incentives from year t-1 to t. "Chg. Pay Ratio" is the change in the pay ratio from year t-1 to t,
multiplied by 100. "Chg. Total Incentives" is new incentives in year t divided by total incentives from the start of year t-1, multiplied by 100. See the Data
Appendix for complete details on variable construction.
Panel A. Executive Roles
CEO
CFO
COO
Other C-Suite
N
20,327
16,362
8,649
4,428
Pct.
19
15
8
4
Non-Exec. Chairman
Senior Vice President
Specialized Exec.
Junior Exec.
N
5,919
6,870
9,918
48,355
Pct.
6
6
9
49
Total Pay ($ Thous.)
Salary ($ Thous.)
Equity Pay ($ Thous.)
Pay Ratio
Equity Holdings ($ Thous.)
Total Incentives ($ Thous.)
New Incentives ($ Thous.)
Panel B. Executive Pay and Incentives
N
Mean
St. Dev.
25th Per.
115,917
2,381
6,210
529
115,917
389
272
221
115,917
1,347
5,464
46
115,917
36.6
27.9
10.0
104,855
21,354
87,314
665
79,197
470.7
6,241.8
18.6
111,382
22.4
86.2
0.6
Median
1,051
315
343
36.5
2,645
60.8
5.4
75th Per.
2,280
475
1,108
58.3
9,789
196.9
18.3
Pct. Chg. New Incentives
Chg. Pay Ratio
Chg. Total Incentives
Panel C. Changes in Equity Pay
N
Mean
St. Dev.
25th Per.
87,934
25.6
95.7
-38.0
89,413
-0.9
29.9
-12.0
58,506
19.0
23.8
1.8
Median
0
0
10.9
75th Per.
93.7
11.4
24.6
Panel D. Pay Changes by Executive Type
Type of Executive
CEO
CFO or COO
Senior V.P.
Other
Pct. Chg. New Incentives
Mean
Median
26.1
0
26.8
2.8
29.0
7.2
25.2
0
Chg. Pay Ratio
Mean
Median
-1.3
0
-1.5
0
-0.2
0
-0.6
0
Chg. Total Incentives
Mean
Median
15.0
7.4
21.6
13.1
19.2
12.3
20.0
11.8
Table 2. Descriptive Statistics: Executive Equity Transactions
This table presents descriptive statistics on the frequency and size of executives' equity transactions. My sample is all firms in the Execucomp and Thomson
Insiders databases, and covers the year 1996 to 2007. Throughout the table, small, medium, and large sales refer to annual sales with dollar value equal to less than
10 percent of equity holdings, 10 to 25 percent of holdings, and more than 25 percent of holdings, respectively. In Panel A, "few executives" means less than half
of the firm's executives, and "most executives" means half or more, but not all, executives. The "All sales when stock ret. high" column is for firm-years in which
the firm's stock return is above the median return for all firms in that year. In Panel B, the CEO column shows statistics for executives who served as CEO at some
point in their tenure; other columns are similarly defined. In Panel C and D, all statistics are conditional on a sale or purchase occurring. "Equity Sold" is the
annual dollar value of all equity sold, minus the exercise cost of stock options which are sold. "Equity Holdings" is the dollar value of all firm equity owned by the
executive at the start of year t-1. "Pct. Drop Incentives" is the total incentives from equity holdings that the executive would have had at year end if he did not sell
equity minus actual total incentives at year end, divided by total incentives at the start of the year, multiplied by 100. "Equity Purchased" is the annual dollar value
of all open-market stock purchases. In Panel D, the CEO column shows statistics for executives serving as CEO at the time; other columns are similarly defined.
See the Data Appendix for complete details on variable construction.
Panel A. Variation in Annual Sales Among Firm's Top Executives
All sales
Medium and Large Sales All sales when stock ret. high
Pct. of firm-years in which…
27
46
19
No executives sell
27
32
24
Few executives sell
32
17
35
Most executives sell
14
4
18
All executives sell
Panel B. Frequency of Equity Transactions
Pct. of executives who…
All
CEO
CFO / COO
Sell equity at least once
61.2
70.7
64.3
Sell equity at least every other year
45.3
44.4
44.9
Sell medium / large amount at least once
44.1
48.8
47.2
Sell medium / large amount every other year
22.9
15.0
20.9
Purchase equity on open market
19.5
36.8
25.8
Panel C. Size of Annual Equity Transactions
N
Mean
St. Dev.
25th Per.
Equity Sold ($ Thous)
49,061
9,193
584,196
139
Equity Sold / Equity Holdings
37,388
0.31
0.46
0.05
Pct. Drop. Incentives
34,891
10.0
21.6
0.5
Pct. Drop. Incentives - small sale
15,923
3.3
15.7
-0.1
Pct. Drop. Incentives - medium sale
9,910
12.0
21.0
3.1
Pct. Drop. Incentives - large sale
8,333
20.4
26.8
4.2
Equity Purchased ($ Thous.)
8,979
557
7,685
18
Equity Purchased / Equity Holdings
5,294
0.09
0.11
0.01
Type of Executive
CEO
CFO or COO
Senior V.P.
Other
Senior V.P.
67.1
49.1
49.9
23.3
17.2
Other
57.4
45.5
41.5
25.5
12.5
Median
558
0.15
6.9
2.8
12.1
22.2
56
0.04
75th Per.
2,004
0.36
20.8
8.2
23.1
44.9
188
0.14
Panel D. Size of Annual Equity Sales by Executive Type
Equity Sold / Equity Holdings
Pct. Drop. Incentives
25th Per.
Median
75th Per.
25th Per.
Median
75th Per.
0.03
0.09
0.23
0.4
5.1
15.7
0.05
0.16
0.39
0.3
7.0
21.5
0.05
0.14
0.35
0.8
7.4
20.3
0.06
0.17
0.42
0.7
7.9
23.0
Table 3. Across-firm pay changes after equity sales (measured by dollar value)
This table examines whether executives who sell equity subsequently receive larger increases in equity pay than executives who do not sell, across firms in my sample. My sample is all
firms in the Execucomp and Thomson Insiders databases, and covers the year 1996 to 2007. Standard errors are reported in parentheses, and *, **, *** denote statistical significance at the
1, 5, and 10 percent levels. In all regressions, standard errors are adjusted for heteroskedasticity and clustering at the firm level. For the dependent variables, "Pct. Chg. New Incentives" is
the percentage change in new incentives from year t-1 to t. "Chg. Pay Ratio" is the change in the ratio of equity to total pay from year t-1 to t, multiplied by 100. "Chg. Total Incentives" is
new incentives in year t divided by total incentives from the start of year t-1, multiplied by 100. All dependent variables are winsorized at the 5-95 level. Incentives are defined as the
change in dollar value of equity for a one-percent change in stock price, as in Baker & Hall (2005). "Small sale [t-1]" is the change from year t-2 to t-1 in an indicator variable equal to 1 if
the dollar value of annual equity sales is less than 10 percent of total equity holdings. For "Medium sale [t-1]" and "Large sale [t-1]", the indicator equals 1 if the executive sells equity
worth 10 to 25 percent of total holdings, or more than 25 percent of total holdings, in year t-1. For "Purchase [t-1]", the indicator equals 1 if the executive purchases any equity on the open
market in year t-1. "Junior Executive" is an indicator equal to 1 if the executive has been in the firm's top management for less than five years. "Firm Q" is defined as firm market value
plus liabilities, divided by assets. "Log Stock Return [t-1]" is the log of 1 plus the stock return from year t-1 to t. "Stock Volatility [t-1]" is the standard deviation of monthly logarithmic
stock returns in years t-1 and t-2. "Sales Growth [t-1]" is the fractional change in sales from year t-2 to t-1, and "Net Income Growth [t-1]" is the change in net income from year t-2 to t-1,
divided by sales from year t-2. "Tech Firm" is an indicator equal to 1 if a firm is a "New Economy" firm as defined by Murphy (2003). "Promoted to CEO", "Stepped Down CEO",
"Promoted to Chair" and "Promoted to C-Suite" are indicators equal to 1 if the given event occurred in year t. "Promoted Chair", and executive role indicators. Industry fixed effects are
defined according to the Fama-French 12 industry classification. See the Data Appendix for complete details on variable construction.
The null hypothesis is that the coefficient on the sale indicators is zero. The alternative hypothesis that selling executives receive ≥ 5%
more equity pay than non-selling executives is rejected in all regressions with 99 percent confidence.
Dependent Variable
Type of Executive
Pct. Chg. New Incentives [t]
[st. dev. = 95.7]
Chg. Pay Ratio [t]
[st. dev. = 29.8]
Chg. Total Incentives [t]
[st. dev. = 23.8]
All
CEO
All
CEO
All
CEO
Small Sale [t-1]
-1.54
(1.23)
-0.19
(2.10)
-0.49
(0.37)
-0.10
(0.65)
-0.43*
(0.23)
-0.41
(0.37)
Medium Sale [t-1]
-1.52
(1.39)
1.22
(2.52)
0.02
(0.44)
0.77
(0.87)
-0.77***
(0.26)
-0.66
(0.48)
-3.74**
(1.68)
-5.20*
(3.08)
-0.05
(0.54)
-0.27
(1.12)
-2.38***
(0.32)
-2.08***
(0.60)
0.04
(1.78)
-1.49
(2.61)
0.08
(0.56)
-0.45
(0.84)
0.30
(0.40)
-0.18
(0.56)
Junior Executive
7.20***
(1.14)
9.34***
(2.93)
0.28
(0.31)
0.63
(0.88)
11.27***
(0.35)
17.54***
(0.93)
Log Assets [t-1]
2.21***
(0.44)
1.53***
(0.54)
0.25**
(0.11)
0.31**
(0.14)
1.56***
(0.16)
1.74***
(0.20)
Firm Q [t-1]
-0.98**
(0.42)
-1.48***
(0.47)
-0.05
(0.13)
-0.22
(0.17)
-0.50***
(0.14)
-0.54***
(0.17)
-2.87
(4.56)
-10.42**
(5.17)
-0.17
(1.32)
-1.83
(1.68)
3.95**
(1.58)
1.85
(1.91)
Large Sale [t-1]
Purchase [t-1]
Book-Market [t-1]
Change Assets [t-1]
-4.33
(3.95)
-6.66
(4.68)
-2.55**
(1.29)
-3.31**
(1.56)
-0.98
(0.90)
-1.17
(1.12)
Change Firm Q [t-1]
3.96
(6.25)
5.49
(7.35)
-1.65
(2.01)
-1.11
(2.43)
6.76***
(1.56)
7.70***
(1.83)
Log Stock Return [t-1]
37.25***
(3.89)
34.72***
(4.60)
7.74***
(1.32)
6.68***
(1.60)
9.56***
(1.04)
7.96***
(1.20)
Log Stock Return [t-2]
-5.14**
(2.33)
-6.66**
(2.84)
1.75**
(0.72)
1.30
(0.90)
-7.34***
(0.56)
-6.45***
(0.68)
Stock Volatility [t-1]
-0.14
(0.14)
-0.02
(0.17)
-0.10**
(0.04)
-0.05
(0.05)
0.24***
(0.05)
0.22***
(0.06)
Sales Growth [t-1]
1.79
(4.07)
2.84
(4.92)
0.92
(1.24)
1.12
(1.59)
0.97
(0.98)
-0.46
(1.21)
Net Income Growth [t-1]
-2.36
(5.79)
-4.57
(7.19)
2.86
(2.00)
2.96
(2.54)
1.56
(1.31)
3.03*
(1.69)
Tech Firm
3.55
(2.43)
3.32
(3.38)
0.78
(0.66)
1.06
(0.94)
2.02**
(0.83)
0.85
(1.09)
Promoted CEO [t-1]
44.19***
(4.16)
51.46***
(5.13)
6.17***
(1.07)
6.67***
(1.27)
12.30***
(0.99)
14.48***
(1.26)
Stepped Down CEO [t-1]
-46.06***
(4.21)
-14.44***
(1.59)
-12.15***
(0.96)
Promoted Chair [t-1]
17.47***
(4.50)
3.85***
(1.42)
7.90***
(1.10)
Promoted C-Suite [t-1]
17.71***
(2.87)
2.30***
(0.81)
5.19***
(0.72)
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
43,565
0.05
12,381
0.05
43,868
0.02
12,470
0.02
43,446
0.15
12,350
0.16
Executive Type Fixed Effects
Year Fixed Effects
Industry Fixed Effects
Firm-Year Fixed Effects
Observations
R-squared
Takeaway:
Across firms in my sample, executives who sell larger dollar amounts of equity do not receive larger
increases in equity pay than executives who do not sell.
Table 4. Across-firm pay changes after equity sales (measured by drop in incentives)
This table examines whether executives who sell equity subsequently receive larger increases in equity pay than executives who do not sell, across firms in my sample. My sample
is all firms in the Execucomp and Thomson Insiders databases, and covers the year 1996 to 2007. Standard errors are reported in parentheses, and *, **, *** denote statistical
significance at the 1, 5, and 10 percent levels. In all regressions, standard errors are adjusted for heteroskedasticity and clustering at the firm level. For the dependent variables,
"Pct. Chg. New Incentives" is the percentage change in new incentives from year t-1 to t. "Chg. Pay Ratio" is the change in the ratio of equity to total pay from year t-1 to t,
multiplied by 100. "Chg. Total Incentives" is new incentives in year t divided by total incentives from the start of year t-1, multiplied by 100. All dependent variables are
winsorized at the 5-95 level. Incentives are defined as the change in dollar value of equity for a one-percent change in stock price, as in Baker & Hall (2005). "Pct. Drop
Incentives" is the total incentives from equity holdings that the executive would have had at year end if he did not sell equity minus actual total incentives at year end, divided by
total incentives at the start of the year, multiplied by 100. For "Purchase [t-1]", the indicator equals 1 if the executive purchases equity on the open market. "Junior Executive" is an
indicator equal to 1 if the executive has been in the firm's top management for less than five years. "Firm Q" is defined as firm market value plus liabilities, divided by assets. "Log
Stock Return [t-1]" is the log of 1 plus the fractional change in stock price from year t-1 to t. "Stock Volatility [t-1]" is the standard deviation of monthly stock returns in years t-1
and t-2. "Sales Growth [t-1]" is the fractional change in sales from year t-2 to t-1, and "Net Income Growth [t-1]" is the change in net income from year t-2 to t-1, divided by sales
from year t-2. "Tech Firm" is an indicator equal to 1 if a firm is a "New Economy" firm as defined by Murphy (2003). "Promoted to CEO", "Stepped Down CEO", "Promoted to
Chair" and "Promoted to C-Suite" are indicators equal to 1 if the given event occurred in year t. "Promoted Chair", and executive role indicators. Industry fixed effects are defined
according to the Fama-French 12 industry classification. See the Data Appendix for complete details on variable construction.
The null hypothesis is a zero coefficient for "Pct. Drop Incentives [t-1]". The alternative hypothesis that the coefficient is 1 is
rejected in all regressions with 99 percent confidence.
Dependent Variable
Type of Executive
Pct. Drop Incentives [t-1]
Pct. Chg. New Incentives [t]
[st. dev. = 95.7]
All
CEO
Chg. Pay Ratio [t]
[st. dev. = 29.8]
All
CEO
Chg. Total Incentives [t]
[st. dev. = 23.8]
All
CEO
0.05
(0.04)
0.13**
(0.06)
-0.00
(0.01)
0.01
(0.02)
0.08***
(0.01)
0.15***
(0.02)
Junior Executive
6.33***
(1.15)
9.61***
(2.96)
0.14
(0.31)
0.42
(0.90)
11.81***
(0.37)
18.73***
(0.92)
Log Assets [t-1]
2.21***
(0.45)
1.86***
(0.54)
0.28**
(0.11)
0.35**
(0.15)
1.45***
(0.17)
1.76***
(0.20)
Firm Q [t-1]
-0.91**
(0.41)
-1.41***
(0.46)
-0.02
(0.13)
-0.19
(0.17)
-0.52***
(0.14)
-0.61***
(0.17)
Book-Market [t-1]
-3.24
(4.52)
-7.15
(5.14)
0.17
(1.33)
-0.72
(1.71)
3.70**
(1.57)
2.18
(1.84)
Change Assets [t-1]
-3.79
(3.97)
-5.77
(4.69)
-2.46*
(1.28)
-3.07**
(1.57)
-0.86
(0.91)
-1.06
(1.11)
Change Firm Q [t-1]
6.00
(6.31)
6.22
(7.36)
-0.65
(2.02)
-0.47
(2.46)
6.90***
(1.58)
7.75***
(1.83)
Log Stock Return [t-1]
34.29***
(3.96)
33.55***
(4.59)
6.93***
(1.34)
6.50***
(1.62)
8.45***
(1.08)
7.00***
(1.19)
Log Stock Return [t-2]
-4.60*
(2.37)
-5.57*
(2.84)
2.03***
(0.72)
1.65*
(0.91)
-7.73***
(0.57)
-6.11***
(0.67)
Stock Volatility [t-1]
-0.21
(0.14)
-0.11
(0.17)
-0.12***
(0.04)
-0.05
(0.05)
0.21***
(0.05)
0.18***
(0.06)
Sales Growth [t-1]
3.10
(4.05)
2.41
(4.95)
1.35
(1.22)
1.19
(1.58)
1.08
(0.99)
-0.47
(1.20)
Net Income Growth [t-1]
-3.60
(5.80)
-5.67
(6.83)
2.38
(1.92)
2.35
(2.47)
0.39
(1.36)
1.98
(1.68)
Tech Firm
3.59
(2.38)
4.37
(3.29)
0.70
(0.65)
1.46
(0.94)
1.52*
(0.82)
0.65
(1.08)
Promoted CEO [t-1]
43.99***
(4.23)
51.91***
(5.20)
5.86***
(1.08)
6.32***
(1.28)
12.22***
(1.00)
14.55***
(1.26)
Stepped Down CEO [t-1]
-46.41***
(4.31)
-14.72***
(1.60)
-12.41***
(0.97)
Promoted Chair [t-1]
14.90***
(4.60)
3.50**
(1.44)
7.82***
(1.12)
Promoted C-Suite [t-1]
18.03***
(2.91)
2.28***
(0.81)
5.03***
(0.72)
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
43,033
0.05
12,220
0.05
43,171
0.02
12,204
0.01
43,033
0.16
12,220
0.18
Executive Type Fixed Effects
Year Fixed Effects
Industry Fixed Effects
Firm-Year Fixed Effects
Observations
R-squared
Takeaway:
There is some cross-sectional evidence that executives whose incentives decrease more due to
equity sales receive larger increases in annual equity pay and total incentives than non-selling
executives. However, the magnitude of the relationship is small.
Table 5. Within-firm pay changes after equity sales (measured by dollar value)
This table examines whether executives who sell larger amounts of equity subsequently receive larger increases in equity pay than executives at the same
firm who do not sell. All regressions include firm-year fixed effects. My sample is all firms in the Execucomp and Thomson Insiders databases, and
covers the year 1996 to 2007. Standard errors are reported in parentheses, and *, **, *** denote statistical significance at the 1, 5, and 10 percent levels.
In all regressions, standard errors are adjusted for heteroskedasticity and clustering at the firm level. The second and fourth regressions include just "CSuite" executives with "Chief" in their title. The dependent variable "Pct. Chg. New Incentives" is the percentage change in new incentives from year t-1
to t, and "Chg. Pay Ratio" is the change in the ratio of equity to total pay from year t-1 to t, multiplied by 100. Incentives are defined as the change in
dollar value of equity for a one-percent change in stock price, as in Baker & Hall (2005). "Small sale [t-1]" is the change from year t-2 to t-1 in an
indicator variable equal to 1 if the dollar value of annual equity sales is less than 10 percent of total equity holdings. For "Medium sale [t-1]" and "Large
sale [t-1]", the indicator equals 1 if the executive sells equity worth 10 to 25 percent of total holdings, or more than 25 percent of total holdings. For
"Purchase [t-1]", the indicator equals 1 if the executive buys equity on the open market. "Junior Executive" is an indicator equal to 1 if the executive has
been in the firm's top management for less than five years. "Promoted to CEO", "Stepped Down CEO", "Promoted to Chair", and "Promoted to C-Suite"
are indicators equal to 1 if the given event occurred in year t. See the Data Appendix for complete details on variable construction.
The null hypothesis is that the coefficient on the sale indicators is zero. An alternative hypothesis that selling
executives receive ≥ 5% more equity pay than non-selling executives is rejected in all regressions with 99
percent confidence.
Dependent Variable
Type of Executive
Pct. Chg. New Incentives [t]
[st. dev. = 95.7]
Chg. Pay Ratio [t]
[st. dev. = 29.8]
All
C-Suite Only
All
C-Suite Only
Small Sale [t-1]
-1.23
(0.79)
-1.78
(1.74)
-0.29
(0.25)
-0.87
(0.54)
Medium Sale [t-1]
-0.55
(0.86)
-0.16
(2.09)
-0.35
(0.28)
-0.86
(0.65)
Large Sale [t-1]
-0.67
(1.06)
-1.24
(2.39)
-0.29
(0.35)
-0.85
(0.85)
Purchase [t-1]
-0.20
(1.16)
1.87
(2.54)
0.10
(0.35)
0.46
(0.77)
Junior Executive
5.05***
(0.78)
3.78*
(1.98)
0.58**
(0.23)
0.36
(0.55)
Promoted to CEO
48.65***
(3.26)
53.79***
(9.34)
8.14***
(0.87)
11.79***
(2.39)
Stepped down CEO
-44.45***
(4.04)
-13.77***
(1.43)
Promoted to Chair
13.34***
(3.57)
3.58***
(1.17)
Promoted to C-Suite
14.79***
(2.28)
11.37***
(3.15)
2.02***
(0.63)
1.67*
(0.92)
+
+
+
+
+
52,113
0.04
+
9,225
0.01
+
52,826
0.02
+
9,367
0.01
Executive Type Fixed Effects
Year Fixed Effects
Industry Fixed Effects
Firm-Year Fixed Effects
Observations
R-squared
Takeaway:
Executives who sell larger dollar amounts of equity do not subsequently receive
larger increases in equity pay than executives at the same firm who do not sell.
Table 6. Within-firm pay changes after equity sales (measured by drop in incentives)
This table examines whether executives whose total incentives decrease more due to equity sales subsequently receive larger increases in equity
pay than executives at the same firm who do not sell. All regressions include firm-year fixed effects. My sample is all firms in the Execucomp and
Thomson Insiders databases, and covers the year 1996 to 2007. Standard errors are reported in parentheses, and *, **, *** denote statistical
significance at the 1, 5, and 10 percent levels. In all regressions, standard errors are adjusted for heteroskedasticity and clustering at the firm level.
The second and fourth regressions include just "C-Suite" executives with "Chief" in their title. The dependent variable "Pct. Chg. New Incentives"
is the percentage change in new incentives from year t-1 to t, and "Chg. Pay Ratio" is the change in the ratio of equity to total pay from year t-1 to
t, multiplied by 100. "Pct. Drop Incentives" is the total incentives from equity holdings that the executive would have had at year end if he did not
sell equity minus actual total incentives at year end, divided by total incentives at the start of the year, multiplied by 100. Incentives are defined as
the change in dollar value of equity for a one-percent change in stock price, as in Baker & Hall (2005). "Junior Executive" is an indicator equal to
1 if the executive has been in the firm's top management for less than five years. "Promoted to CEO", "Stepped Down CEO", "Promoted to
Chair", and "Promoted to C-Suite" are indicators equal to 1 if the given event occurred in year t. See the Data Appendix for complete details on
variable construction.
The null hypothesis is that the coefficient on "Pct. Drop Incentives [t-1]" is zero. An alternative
hypothesis that the coefficient is 1 is rejected in all regressions with 99 percent confidence.
Dependent Variable
Type of Executive
Pct. Drop Incentives [t-1]
Pct. Chg. New Incentives [t]
[st. dev. = 95.7]
All
C-Suite Only
Chg. Pay Ratio [t]
[st. dev. = 29.8]
All
C-Suite Only
-0.00
(0.02)
0.05
(0.06)
-0.01
(0.01)
0.02
(0.02)
Junior Executive
4.96***
(0.79)
3.57*
(1.98)
0.52**
(0.23)
0.21
(0.54)
Promoted to CEO
48.94***
(3.34)
48.36***
(9.50)
8.48***
(0.88)
11.07***
(2.45)
Stepped down CEO
-42.87***
(4.09)
-12.81***
(1.45)
Promoted to Chair
11.29***
(3.65)
2.39**
(1.19)
Promoted to C-Suite
14.97***
(2.30)
12.69***
(3.22)
1.88***
(0.64)
2.01**
(0.95)
+
+
+
+
+
51,309
0.04
+
9,019
0.01
+
51,481
0.02
+
9,069
0.01
Executive Type Fixed Effects
Year Fixed Effects
Industry Fixed Effects
Firm-Year Fixed Effects
Observations
R-squared
Takeaway:
Executives whose incentives decrease more due to equity sales do not
receive more equity pay than executives at the same firm who do not sell.
Table 7. Within-firm changes in total incentives after equity sales (both measures)
This table examines whether boards replenish executives' total equity holdings following equity sales. All regressions include firm-year fixed effects.
My sample is all firms in the Execucomp and Thomson Insiders databases, and covers the year 1996 to 2007. Standard errors are reported in
parentheses, and *, **, *** denote statistical significance at the 1, 5, and 10 percent levels. In all regressions, standard errors are adjusted for
heteroskedasticity and clustering at the firm level. The second and fourth regressions include just "C-Suite" executives with "Chief" in their title.
"Chg. Total Incentives" is new incentives in year t divided by total incentives from the start of year t-1, multiplied by 100. It is winsorized at the 5-95
level. Incentives are defined as the change in dollar value of equity for a one-percent change in stock price, as in Baker & Hall (2005). "Small sale [t1]" is the change from year t-2 to t-1 in an indicator variable equal to 1 if the dollar value of annual equity sales equals less than 10 percent of total
equity holdings. For "Medium sale [t-1]" and "Large sale [t-1]", the indicator equals 1 if the executive sells equity worth 10 to 25 percent of total
holdings, or more than 25 percent of total holdings. For "Purchase [t-1]", the indicator equals 1 if the executive purchases equity on the open market.
"Pct. Drop Incentives" is the total incentives from equity holdings that the executive would have had at year end if he did not sell equity minus actual
total incentives at year end, divided by total incentives at the start of the year, multiplied by 100. "Junior Executive" is an indicator equal to 1 if the
executive has been in the firm's top management for less than five years. "Promoted to CEO", "Stepped Down CEO", "Promoted to Chair", and
"Promoted to C-Suite" are indicators equal to 1 if the given event occurred in year t. See the Data Appendix for complete details on variable
construction.
The null hypothesis is that the coefficient on all sale variables is zero. The first two regressions reject the
alternative hypotheses that selling executives receive ≥ 5% more equity pay than non-selling executives.
The next two regressions reject the alternative that the coefficient on "Pct. Drop Incentives" is 1.
Dependent Variable
Type of Executive
Chg. Total Incentives [t] [st. dev = 23.8]
All
C-Suite Only
Pct. Drop Incentives [t-1]
All
C-Suite Only
0.04***
(0.01)
0.07***
(0.02)
Small Sale [t-1]
-0.05
(0.17)
-0.27
(0.44)
Medium Sale [t-1]
-0.26
(0.20)
-0.43
(0.49)
-1.56***
(0.24)
-2.25***
(0.63)
0.25
(0.30)
1.03
(0.65)
Junior Executive
9.21***
(0.26)
9.40***
(0.68)
9.75***
(0.27)
9.78***
(0.68)
Promoted to CEO
13.08***
(0.81)
9.13***
(2.12)
13.00***
(0.82)
7.69***
(2.17)
Stepped down CEO
-12.97***
(0.98)
-12.66***
(1.00)
Promoted to Chair
6.88***
(0.95)
6.79***
(0.97)
Promoted to C-Suite
5.08***
(0.58)
4.33***
(0.88)
5.34***
(0.60)
5.04***
(0.92)
+
+
+
+
+
51,873
0.12
+
9,158
0.12
+
51,309
0.13
+
9,019
0.12
Large Sale [t-1]
Purchase [t-1]
Executive Type Fixed Effects
Year Fixed Effects
Industry Fixed Effects
Firm-Year Fixed Effects
Observations
R-squared
Takeaway:
Boards replenish at most 7 percent of incentives lost from equity sales. There
is also no relationship between dollar amounts of equity sold and changes in
total incentives.
Table 8. Robustness check: does board learning over time explain lack of replenishment?
This table tests whether my main results change when I restrict analysis to groups of three or more executives who joined the firm's top management in the same year, or who have worked
together for at least four years. These results should not be affected by differences in executive tenure or board information. All regressions include firm-year fixed effects. My sample is all
firms in the Execucomp and Thomson Insiders databases, and covers the year 1996 to 2007. Standard errors are reported in parentheses, and *, **, *** denote statistical significance at the 1, 5,
and 10 percent levels. In all regressions, standard errors are adjusted for heteroskedasticity and clustering at the firm level. For the dependent variables, "Pct. Chg. New Incentives" is the
percentage change in new incentives from year t-1 to t. "Chg. Pay Ratio" is the change in the ratio of equity to total pay from year t-1 to t, multiplied by 100. "Chg. Total Incentives" is new
incentives in year t divided by total incentives from the start of year t-1, multiplied by 100. All dependent variables are winsorized at the 5-95 level. "Small sale [t-1]" is the change from year t-2
to t-1 in an indicator variable equal to 1 if the dollar value of annual equity sales is less than 10 percent of total equity holdings. For "Medium sale [t-1]" and "Large sale [t-1]", the indicator
equals 1 if the executive sells equity worth 10 to 25 percent of total holdings, or more than 25 percent of total holdings. For "Purchase [t-1]", the indicator equals 1 if the executive purchases
equity on the open market. "Pct. Drop Incentives" is the total incentives from equity holdings that the executive would have had at year end if he did not sell equity minus actual total incentives
at year end, divided by total incentives at the start of the year, multiplied by 100. Incentives are defined as the change in dollar value of equity for a one-percent change in stock price, as in
Baker & Hall (2005). "Junior Executive" is an indicator equal to 1 if the executive has been in the firm's top management for less than five years. "Promoted to CEO", "Stepped Down CEO",
"Promoted to Chair", and "Promoted to C-Suite" are indicators equal to 1 if the given event occurred in year t. See the Data Appendix for complete details on variable construction.
Pct. Chg. New
Chg. Pay Ratio [t]
Incentives [t]
[st. dev. = 95.7] [st. dev. = 29.8]
Dependent Variable:
Type of Executive:
Chg. Total
Incentives [t]
[st. dev. = 23.8]
Pct. Chg. New
Chg. Pay Ratio [t]
Incentives [t]
[st. dev. = 95.7] [st. dev. = 29.8]
Chg. Total
Incentives [t]
[st. dev. = 23.8]
Executives belonging to same cohort
Pct. Drop Incentives [t-1]
0.01
(0.05)
-0.00
(0.01)
0.02
(0.02)
Small Sale [t-1]
-1.65
(1.34)
-0.52
(0.43)
-0.25
(0.28)
Medium Sale [t-1]
-1.86
(1.41)
-0.69
(0.48)
-0.52*
(0.29)
Large Sale [t-1]
-1.69
(1.70)
-0.19
(0.56)
-2.31***
(0.39)
Purchase [t-1]
-3.40*
(1.98)
-0.28
(0.59)
0.04
(0.53)
Junior Executive
-1.00
(4.76)
0.31
(1.25)
6.28***
(1.42)
-1.20
(5.10)
-0.54
(1.25)
6.22***
(1.47)
Promoted to CEO
47.36***
(5.33)
6.71***
(1.42)
11.82***
(1.38)
48.10***
(5.57)
6.72***
(1.49)
11.77***
(1.41)
Stepped down CEO
-46.23***
(7.34)
-11.98***
(2.03)
-12.15***
(1.75)
-44.80***
(7.37)
-11.62***
(2.11)
-12.27***
(1.80)
Promoted to Chair
15.74***
(5.62)
3.96**
(1.63)
6.59***
(1.56)
14.77***
(5.62)
4.08**
(1.71)
6.92***
(1.61)
Promoted to C-Suite
19.94***
(3.88)
3.86***
(0.99)
5.35***
(0.97)
21.24***
(3.95)
4.43***
(1.02)
5.76***
(0.98)
+
+
+
+
+
+
+
16,947
0.03
+
17,332
0.01
+
16,837
0.07
+
16,242
0.04
+
16,310
0.01
+
16,242
0.07
Executive Type Fixed Effects
Year Fixed Effects
Industry Fixed Effects
Firm-Year Fixed Effects
Observations
R-squared
Takeaway:
My main results do not change when I restrict the sample to executives in the same cohort, for whom boards
should learn information at the same time.
Table 9. Robustness check: do shocks to outside wealth explain lack of replenishment?
This table tests whether my main results change when I restrict analysis to executives who likely did not experience wealth shocks. Panel A tests shows pay changes from 2006 to 2007 for
executives who deferred a portion of compensation in 2006. Panel B shows pay changes from 2005 to 2006 and 2006 to 2007, for executives whose deferred compensation accounts in 2006
or 2007 are equal to at least 10 percent of the value of firm equity holdings. All regressions include firm-year fixed effects. My sample is all firms in the Execucomp and Thomson Insiders
databases, and covers the year 1996 to 2007. Standard errors are reported in parentheses, and *, **, *** denote statistical significance at the 1, 5, and 10 percent levels. In all regressions,
standard errors are adjusted for heteroskedasticity and clustering at the firm level. For the dependent variables, "Pct. Chg. New Incentives" is the percentage change in new incentives from
year t-1 to t. "Chg. Pay Ratio" is the change in the ratio of equity to total pay from year t-1 to t, multiplied by 100. "Chg. Total Incentives" is new incentives in year t divided by total
incentives from the start of year t-1, multiplied by 100. All dependent variables are winsorized at the 5-95 level. Incentives are defined as the change in dollar value of equity for a onepercent change in stock price, as in Baker & Hall (2005). "Small sale [t-1]" is the change from year t-2 to t-1 in an indicator variable equal to 1 if the dollar value of annual equity sales is less
than 10 percent of total equity holdings. For "Medium sale [t-1]" and "Large sale [t-1]", the indicator equals 1 if the executive sells equity worth 10 to 25 percent of total holdings, or more
than 25 percent of total holdings. "Pct. Drop Incentives" is the total incentives from equity holdings that the executive would have had at year end if he did not sell equity minus actual total
incentives at year end, divided by total incentives at the start of the year, multiplied by 100. Other variables included in the regressions, but not shown, are "Purchase [t-1]", "Junior
Executive", "Promoted to CEO", "Stepped Down CEO", "Promoted to Chair", "Promoted to C-Suite". See the Data Appendix for complete details on variable construction.
Dependent Variable:
Panel A. Executives who contribute to deferred compensation account in 2006
Pct. Chg. New
Chg. Total
Pct. Chg. New
Chg. Total
Chg. Pay Ratio [t]
Chg. Pay Ratio [t]
Incentives [t]
Incentives [t]
Incentives [t]
Incentives [t]
Pct. Drop Incentives [t-1]
-0.03
(0.14)
-0.00
(0.04)
-0.02
(0.04)
Small Sale [t-1]
2.31
(3.30)
1.53
(1.30)
1.40
(1.10)
Medium Sale [t-1]
-0.75
(4.23)
1.20
(1.45)
-0.36
(1.25)
Large Sale [t-1]
-0.54
(4.48)
1.63
(1.81)
-2.13
(1.73)
Purchase [t-1]
-6.02
(12.87)
2.92
(3.70)
4.14
(3.53)
Junior Executive
10.94**
(4.44)
3.54**
(1.43)
12.97***
(1.66)
12.33***
(4.70)
4.69***
(1.37)
12.67***
(1.85)
Promoted to CEO
16.54
(25.75)
3.26
(5.65)
14.12**
(5.92)
21.77
(28.04)
3.33
(6.06)
12.32*
(6.32)
Stepped down CEO
-82.16
(56.26)
-36.45**
(15.95)
-25.73***
(7.27)
-51.54
(48.08)
-29.88*
(16.09)
-26.67***
(6.81)
Promoted to Chair
33.49
(53.86)
19.51
(12.70)
15.30**
(6.27)
5.69
(45.61)
12.09
(12.90)
17.30***
(5.78)
Promoted to C-Suite
10.84
(11.91)
-0.18
(3.66)
3.41
(2.59)
14.62
(13.79)
0.02
(4.20)
4.34*
(2.58)
+
+
+
+
+
+
+
1,155
0.06
+
1,183
0.07
+
1,143
0.18
+
1,008
0.07
+
1,020
0.07
+
1,008
0.18
Executive Type Fixed Effects
Year Fixed Effects
Industry Fixed Effects
Firm-Year Fixed Effects
Observations
R-squared
Pct. Chg. New
Incentives [t]
Dependent Variable:
Panel B. Executives with substantial savings in account in 2006 or 2007
Chg. Total
Pct. Chg. New
Chg. Pay Ratio [t]
Chg. Pay Ratio [t]
Incentives [t]
Incentives [t]
Pct. Drop Incentives [t-1]
Chg. Total
Incentives [t]
-0.04
(0.11)
-0.01
(0.02)
-0.07
(0.05)
Small Sale [t-1]
-0.54
(3.79)
0.24
(1.15)
0.29
(1.03)
Medium Sale [t-1]
-2.66
(4.57)
0.68
(1.30)
-1.22
(1.10)
Large Sale [t-1]
-6.08
(4.73)
-0.18
(1.17)
-2.79*
(1.52)
Purchase [t-1]
8.02
(7.24)
2.24
(1.76)
0.36
(1.82)
Junior Executive
10.64**
(4.17)
1.47
(1.35)
10.38***
(1.56)
11.60***
(4.27)
2.22*
(1.14)
9.92***
(1.60)
Promoted to CEO
52.19***
(19.66)
0.82
(5.16)
13.26**
(5.30)
46.52**
(22.54)
1.20
(4.04)
12.05**
(6.09)
Stepped down CEO
-30.02
(31.93)
-19.67*
(10.32)
-12.79**
(5.29)
-24.65
(23.31)
-7.96
(7.65)
-13.30**
(5.27)
Promoted to Chair
-35.02
(32.57)
-9.08
(10.19)
-2.80
(5.17)
-58.65***
(20.54)
-16.03**
(7.87)
-4.47
(5.24)
Promoted to C-Suite
17.81
(12.11)
2.45
(2.68)
7.94**
(3.52)
22.54*
(11.84)
4.24*
(2.55)
9.48***
(3.42)
+
+
+
+
+
+
+
1,343
0.06
+
1,371
0.06
+
1,340
0.11
+
1,348
0.07
+
2,448
0.06
+
1,348
0.12
Executive Type Fixed Effects
Year Fixed Effects
Industry Fixed Effects
Firm-Year Fixed Effects
Observations
R-squared
Takeaway:
There is no evidence of replenishment on a subsample of executives who are less likely to have recently
experienced large shocks to outside wealth or liquidity.
Table 10. Robustness check: do boards only respond to unexpected sales?
This table tests whether boards are more likely to replenish the incentives of executives who sold equity after exercising stock options early. "Early Exercise" is an
indicator equal to one if at least half of the dollar value of equity sold by the executive in year t-1 is from shares acquired when the executive exercises a stock option
with more than six years left until expiration. All regressions include firm-year fixed effects. My sample is all firms in the Execucomp and Thomson Insiders databases,
and covers the year 1996 to 2007. Standard errors are reported in parentheses, and *, **, *** denote statistical significance at the 1, 5, and 10 percent levels. In all
regressions, standard errors are adjusted for heteroskedasticity and clustering at the firm level. For the dependent variables, "Pct. Chg. New Incentives" is the percentage
change in new incentives from year t-1 to t. "Chg. Pay Ratio" is the change in the ratio of equity to total pay from year t-1 to t, multiplied by 100. "Chg. Total
Incentives" is new incentives in year t divided by total incentives from the start of year t-1, multiplied by 100. All dependent variables are winsorized at the 5-95 level.
Incentives are defined as the change in dollar value of equity for a one-percent change in stock price, as in Baker & Hall (2005). "Small sale [t-1]" is the change from
year t-2 to t-1 in an indicator variable equal to 1 if the dollar value of annual equity sales is less than 10 percent of total equity holdings. For "Medium sale [t-1]" and
"Large sale [t-1]", the indicator equals 1 if the executive sells equity worth 10 to 25 percent of total holdings, or more than 25 percent of total holdings. For "Purchase [t1]", the indicator equals 1 if the executive purchases equity on the open market. "Junior Executive" is an indicator equal to 1 if the executive has been in the firm's top
management for less than five years. "Promoted to CEO", "Stepped Down CEO", "Promoted to Chair", and "Promoted to C-Suite" are indicators equal to 1 if the given
event occurred in year t. See the Data Appendix for complete details on variable construction.
Pct. Chg. New Incentives [t]
Chg. Pay Ratio [t]
Chg. Total Incentives [t]
[st. dev. = 95.7]
[st. dev. = 29.8]
[st. dev. = 23.8]
Small Sale [t-1]
0.72
(1.14)
-0.21
(0.34)
0.22
(0.28)
Medium Sale [t-1]
0.48
(1.19)
-0.80**
(0.38)
-0.14
(0.31)
Large Sale [t-1]
0.39
(1.38)
-0.69
(0.46)
-1.05***
(0.35)
Small Sale [t-1] * Early Exercise
-4.17
(3.74)
-2.06*
(1.20)
-1.11
(0.87)
Medium Sale [t-1] * Early Exercise
-0.76
(3.76)
-0.02
(1.24)
-0.29
(0.88)
Large Sale [t-1] * Early Exercise
-2.41
(3.74)
-0.12
(1.25)
-1.46
(0.96)
Early Exercise
-2.96
(2.08)
-1.20*
(0.71)
2.25***
(0.53)
Purchase [t-1]
-0.73
(2.03)
-0.04
(0.62)
-0.22
(0.57)
Junior Executive
3.49***
(1.15)
0.34
(0.35)
8.13***
(0.36)
Promoted to CEO
49.88***
(4.70)
7.71***
(1.17)
13.07***
(1.20)
Stepped down CEO
-47.22***
(5.49)
-15.70***
(2.09)
-12.55***
(1.39)
Promoted to Chair
15.01***
(4.99)
4.50***
(1.61)
6.76***
(1.27)
Promoted to C-Suite
11.70***
(2.97)
2.28***
(0.80)
4.28***
(0.80)
+
+
+
+
24,952
0.04
+
25,348
0.02
+
24,846
0.12
Dependent Variable:
Other Controls
Year Fixed Effects
Industry Fixed Effects
Firm-Year Fixed Effects
Observations
R-squared
Takeaway:
Boards do not respond to sales following an early exercise of stock options, which
are likely to be unanticipated by the board ex ante.
Table 11. Robustness check: does board's anticipation of sales explain lack of replenishment?
This table tests whether boards anticipate equity sales which occur early in the fiscal year, and hence only replenish incentives ex post for sales which occur later
in the fiscal year. Early sales are those which occur in the first half of the fiscal year, and late sales are those which occur in the second half. All regressions
include firm-year fixed effects. My sample is all firms in the Execucomp and Thomson Insiders databases, and covers the year 1996 to 2007. Standard errors are
reported in parentheses, and *, **, *** denote statistical significance at the 1, 5, and 10 percent levels. In all regressions, standard errors are adjusted for
heteroskedasticity and clustering at the firm level. For the dependent variables, "Pct. Chg. New Incentives" is the percentage change in new incentives from year t1 to t. "Chg. Pay Ratio" is the change in the ratio of equity to total pay from year t-1 to t, multiplied by 100. "Chg. Total Incentives" is new incentives in year t
divided by total incentives from the start of year t-1, multiplied by 100. All dependent variables are winsorized at the 5-95 level. Incentives are defined as the
change in dollar value of equity for a one-percent change in stock price, as in Baker & Hall (2005). "Small Sale Early [t-1]" is the change from year t-2 to t-1 in an
indicator variable equal to 1 if the dollar value of annual equity sales in the first half of the fiscal year is less than 10 percent of total equity holdings. For "Medium
Sale Early [t-1]" and "Large Sale Early [t-1]", the indicator equals 1 if the executive sells equity worth 10 to 25 percent of total holdings, or more than 25 percent
of total holdings. For "Purchase Early [t-1]", the indicator equals 1 if the executive purchases equity on the open market in the first half of the fiscal year. "Small
Sale Late [t-1]", "Medium Sale Late [t-1]", etc. are similarly defined, but for sales which occur in the second half of the fiscal year. "Junior Executive" is an
indicator equal to 1 if the executive has been in the firm's top management for less than five years. "Promoted to CEO", "Stepped Down CEO", "Promoted to
Chair", and "Promoted to C-Suite" are indicators equal to 1 if the given event occurred in year t. See the Data Appendix for complete details on variable
construction.
Pct. Chg. New Incentives [t]
Chg. Pay Ratio [t]
Chg. Total Incentives [t]
[st. dev. = 95.7]
[st. dev. = 29.8]
[st. dev. = 23.8]
Small Sale Early [t-1]
-1.00
(0.80)
-0.65**
(0.25)
0.15
(0.16)
Medium Sale Early [t-1]
-1.04
(1.03)
-0.08
(0.34)
-0.34
(0.22)
Large Sale Early [t-1]
-0.97
(1.41)
-0.51
(0.49)
-1.64***
(0.33)
Small Sale Late [t-1]
-1.33
(0.82)
-0.08
(0.25)
-0.21
(0.17)
Medium Sale Late [t-1]
-1.11
(1.05)
-0.60*
(0.32)
-0.56**
(0.22)
Large Sale Late [t-1]
0.11
(1.49)
0.06
(0.45)
-1.55***
(0.33)
Purchase Early [t-1]
0.08
(1.52)
0.01
(0.43)
1.08***
(0.39)
Purchase Late [t-1]
2.45
(1.58)
0.62
(0.47)
-0.25
(0.39)
Junior Executive
4.52***
(0.81)
0.50**
(0.23)
9.12***
(0.27)
Promoted to CEO
49.42***
(3.42)
7.95***
(0.88)
13.18***
(0.84)
Stepped down CEO
-44.25***
(4.23)
-13.88***
(1.51)
-12.76***
(1.03)
Promoted to Chair
11.87***
(3.61)
3.29***
(1.17)
6.42***
(0.98)
Promoted to C-Suite
14.85***
(2.32)
2.44***
(0.62)
5.09***
(0.59)
Dependent Variable:
Other Controls
Year Fixed Effects
Industry Fixed Effects
Firm-Year Fixed Effects
Observations
R-squared
Takeaway:
+
+
+
46,709
0.04
+
+
47,374
0.02
+
46,501
0.13
Boards do not replenish incentives after sales which occur later in the fiscal year, and are
less likely to be anticipated by the board when setting ex ante compensation.