'a
MIT LIBRARIES
m
3 9080 02237 2673
i
11
iliiiilililillli:
m
!i
^1
iiii
Mliit;
iiiS
:;i|||i
il'I'ii'.fiiiHI'iri.i:.!!/'!'
lffl^^l!f!/Wli;);Mllffl;l'i::
:;:||li|li|||j||ii|f||||||||p^^^
-'iifeiisfv.,
,
llliiiill
iiii:
iiiiiii:
-iiiiii Ipiiili
iiiliiiiiii
\Mi!i(i;i!i;iiSll
:i,;!!iiii|i!l;:|!;!|
IE
LIBRARIES
Digitized by the Internet Archive
in
2011 with funding from
Boston Library Consortium IVIember Libraries
http://www.archive.org/details/doceossettheirowOObert
^^H^y
Massachusetts Institute of Technology
Department of Economics
Working Paper Series
DO CEOs SET THEIR OWN PAY?
THE ONES WITHOUT PRINCIPALS DO
NBER
& NBER
Marianne Bertrand, UChicago &
Sendhil Mullainathan, MIT
Working Paper 00-26
September 2000
Room
E52-251
50 Memorial Drive
Cambridge, MA 02142
This paper can be downloaded without charge from the
Social Science Research
Network Paper Collection at
http:/ /papers. ssrn.com/paper.taf?abstractJd=223736
MASSACHUSETTS INSTITUTE
OF TECHNOLOGY
OCT
3
2000
LIBRARIES
ABSTRACT
We empirically examine two
competing views of CEO pay. In the contracting view, pay is used
to solve an agency problem: the compensation committee optimally chooses pay contracts that
give the CEO incentives to maximize shareholder wealth. In the skimming view, pay is the
result of an agency problem: CEOs have managed to capture the pay process so that they set
their
own pay,
constrained somewhat by the availability of cash or by fear of drawing
shareholders' attention.
To
luck, observable shocks to
luck, such as
changes
responds about as
better
distinguish these views, we first examine how CEO pay responds to
performance beyond the CEO's control. Using several measures of
in oil price for the oil industry,
much
governed firms pay
their
CEOs
less for luck.
are charged for the options they are granted.
might offer an appealing way
in better-governed firms are
that
we
find substantial
to a "lucky" dollar as to a general dollar.
to
Our second
test
pay
for luck.
importantly,
Since options never appear on balance sheets, they
skim. Here again
we
find a crucial role for governance:
and
little
fits
better (pay
charge for options) while in well-governed firms, the contracting view
better (filtering out
of luck and charging for options).
JELNo. G3,J3,J4, L2
CEOs
charged more for the options they are given. These results suggest
both views of CEO pay matter. In poorly governed firms, the skimming view
for luck
Pay
we find that
examines how much CEOs
Most
fits
Introduction
1
There are two predominant ways to think about
CEO
pay.
contracting view, relies on principal-agent models. Because
The
CEOs
they control, shareholders face a classic moral hazard problem:
firm value in making their decisions.
one, which
often
we
own very
CEOs may
CEO
refer to as the
little
of the firm
not always maximize
Under the contracting view, shareholders
the board or the compensation committee) use
for
first
(acting through
pay to reduce moraJ hazard.
Explicit
pay
performance, such as long term contracts or options, and implicit pay for performance, such
as a bonus payment, are all used
by boards to increase CEOs' incentives
to
maximize shareholder
wealth.^
The second
view, which
such as Crystal (1991).
pay.
refer to as the
It also
that this separation allows
board with their
we
begins with the separation of ownership and control, but
CEOs
friends, or
skimming view, has been championed by practitioners
to gain control of the
pay setting process
any other mean of entrenchment, many
They skim what they can from
possible.
Whereas pay
skimming view
We
target.
is
packing the
amoimt
own
of funds in
groups or by a fear of
Within these constraints, however, they pay themselves as much as
in the contracting view is
an attempt
to solve moral hcizard, pay in the
the result of moraJ hazard.
propose two
'Murphy
jictivist
By
argues
de facto set their
shareholders, constrained perhaps by the
the firm, by an unwDlingness to draw the attention of shareholder
becoming a takeover
CEOs
itself.
it
tests to differentiate these
(1985, 1986)
b a
models. In the
first test,
we examine whether CEOs
forerunner of the vast literature that has empirically analyzed
CEO
compensation
summarized in Murphy (1999). The major
econometric work has been to test for vaJue-optimizing incentive scheme by studying the correlation between pay
and performance. Jensen and Murphy (1990) use this framework to argue that political considerations constrain pay
so that incentives are too low. Joskow, Rose £ind Shepard (1993) empirically examine this argument in regulated
industries. Other tests of the agency framework can be found in Gibbons and Murphy (1990), who test for relative
performance evaluation and career concerns, Garen (1994) and Aggarwal and Samwick (1999a), who test for riskreturn treideofe, and Hubbard and Palia (1994) and Bertrand and Mullainathan (1998), who test whether pay
Incentives substitute for other disciplining devices (competition and takeovers respectively).
in the context of the princip>al-agent
model. The resulting literature
is
are rewarded for observable luck.
CEOs'
CEO
By
luck,
we mean changes
in jBrm
performance that are beyond
In simple agency models, pay should not respond to luck since by definition the
control.
cannot influence luck. Tying pay to luck
risk to the contract (Holrostrom, 1979).^
respond to luck since the
CEO
not provide better incentives;
will
Under the skimming view, on
it
will only
add
the other hand, pay will
can divert those "lucky" dollars to pay herself as easily as she can
divert earned dollcu^.
To empirically examine the responsiveness of pay
luck. First,
we perform a
to affect firm
to luck,
we use
case study of the oil industry where large
performance on a regular
basis.
three different measures of
movements
in oil prices
Second, we use changes in industry-specific exchange
rate for firms in the traded goods sector. Third,
we use
year-to-year differences in
performance to proxy for the overall economic fortune of a sector. This
last
mean
CEO
pay
is
CEO
as sensitive to
pay responds
a "lucky
to luck. In fact,
dolleir" as to
we
a "general
indiistry
measure very much
For
resembles the approach followed in the relative performance evaluation literature.
measures, we find that
tend
all
three
find that for all three luck measures
dollar."
This basic finding of pay for luck, however, can be explained by complicating the basic agency
model. For example, suppose boards wanted their
Tying pay to luck in
rewarding the
CEO
open before the
this case
may be
with industry fortunes.
to forecast or respond to luck shocks.
necessary to provide better incentives. In the
after the fact for seeing
fact. Alternatively,
CEOs
an
oil
shock coming encourages him to keep his eyes
suppose the "value" of a CEO's hiiman capital
One would then
oil industry,
find that
rises cuid falls
pay correlates with luck because the CEO's
outside wage moves with luck.
While we wiU argue that these arguments may be incorrect, they motivate us
to search for further
^Note our emphasis on observable luck. In any model, given the r£indomness of the world, CEOs (and almost
everybody else) will end up being rewarded for unobservable luck. Note also our emphasis on the fact that this
prediction holds in simple agency models. As we will discuss shortly, complications to the agency model can in
principle alter this result.
tests.
be
We
therefore
examine another direct implication of the skimming model, that skimming
important in well governed firms.
less
So
control of the pay process.
if
pay
Good
for luck
governance will make
hard for the
it
comes from skimming, we expect
CEO
will
to gain
to see less of
it
in
the better governed firms.
We test
this hypothesis using several
CEO
the board and overall),
proxy
for
measures of governance: presence of large shareholders {on
tenure (interacted with the presence of large shareholders to better
entrenchment), board size and fraction of directors that are insiders.
skimming, we generally find that the better governed firms pay
strongest for the presence of large shcireholders
23 and 33%. Large shareholders are especially important as
CEOs
the idea that unchecked
CEO
These
less for luck.^
on the board who reduce pay
Consistent with
for luck
effects are
by between
tenure increases, consistent with
can entrench themselves over time. The findings on governance cast
doubts on the alternative interpretations which make pay for luck optimal through complications
to the agency model. If
firms use less of
to
be rewairded
at least
it.
pay
were in fact optimal, we would not expect to see well governed
For example, whether or not a large shareholder
for a rise in the vcilue of their
some of the pay
Our second
for luck
around another aspect of
tion in recent yejirs: the greinting of
new
present,
CEOs would
have
humjin capital. These findings instead suggest that
for luck in poorly governed firms is
test revolves
is
CEO
due to skimming by CEOs.
compensation that has received atten-
stock options. Principal-agent theory predicts that
when
options are granted, other components of compensation should be adjusted downwards to leave the
CEO
indifferent. In other words, the
CEO
should be charged for the options she
is
granted,
if
not
their Black-Scholes value then at least this value times a risk correction factor.'* Supporters of the
^Whenever we
for performance.
refer to "less pay for luck" we mean that there is less pay for luck relative to the amount of pay
Thus, these results would not be driven by well governed firms simply giving less overedl pay for
performance.
^The basic idea
is
that options have value. For example, suppose a
price of 50, a horizon of 3 years
because even
if
and the firm
is
CEO
is
granted 10,000 options with a strike
and of themselves
currently trading at 50. These options have value in
the firm under-perfonns relative to the market and earns a cumulative 3-year{nomincil) return of only
skimming view, on the other hand, emphasize that stock options do not appear on balance
sheets.
Because of accounting treatments, firms do not take a financial charge for granting options.
CEOs
can therefore pay themselves through option grants without affecting the company's bottom
Under the assumption that shareholders only
line, the
(or mostly)
line.
pay attention to the accounting bottom
granting of new stock options would represent an easy way to skim more without attracting
No
shareholders' attention.
cut in the other components of pay would be necessary.
Thus while
agency theory predicts a large charge for options, skimming predicts about no charge.
Unfortxmately, a direct measure of
how much CEOs
not possible because of a natural omitted variable bias.
we
findings,
are in fact charged for their options
Taking
on the question of how governance
therefore focus
oiu: lead
from the pay
less for their
new options
grants.
for luck
affects the charge for options.
CEOs
the same measures of governance, we generally find that more poorly governed
For example, for an options grant worth
is
Using
are charged
million dollars (in
1
Black- Scholes terms), an extra large shareholder on the bocird increases the charge for options by
between 30 and 50 thousand
dollars.
As
before,
we
also find that as
a
CEO
tenure increases, the
charge for options diminishes in firms without a large shareholder.
In summary, this paper contains three
main
results. First,
CEO compensation shows on average
a significant amount of pay for luck. Second, well governed firms display
that
it is
possible to filter out luck. Third,
stock options grants.
view.
CEOs
At a
first
CEOs in
seem
10%, he
still
will arise
gains
(.1)
50
the firm's natural
each option.
volatility.
more
for their
skimming
and seem to be able to skim using options. But notice that they
and do manage to charge more
for options.
do manage
We
feel
to filter out
some
that the results suggest
000 = 50, 000 dollars. Even if the option is granted out of the money, the same effect
may come into the money just because of movements in share price that comes from
Agency theory therefore predicts that CEOs should be charged for the gift implicit in
10,
because the option
for luck, suggesting
to provide support for the
also provide support for the contracting view. Well governed firms
luck from performance
pay
well governed firms are charged
glance, these results
are rewarded for luck
less
as a whole that both views of
CEO
is
an
is
more of an
Worse governance means that there
his
own
pay.^
The
rest of this
performance
paper
is
be
pay contracts. Better
active "principal" (or principals) present to actually design
governance means that there
to
skimming or the contracting model depending on the extent to
better characterized by either the
which there
pay are true. In practice, executive compensation seems
is less
active principal and optimal contracting
organized as follows.
We
filtering (section 2).
CEO is
of an active principal and the
first
We
start
by presenting our
more
fits
better.
likely to set
test of the lack of
present a very simple theoretical background (section
2.1),
review the existing evidence (section 2.2) and explain the empirical methodology (section
2.3).
We
then establish the existence of pay-for-luck, both in the specific case of the
(section 3.1)
and
for
more general
sets of industries (section 3.2). In section 3.3,
the role of governance in limiting the extent of pay-for-luck.
same governance
variables also matter in limiting the gift nature of
siunmarize and conclude in Section
Pay
2
Luck
Test:
new
we
4,
industry
we demonstrate
establish that the
stock option grants.
We
5.
Background
Theoretical Background
2.1
Our
for
In section
oil
first test
model
will
focuses
make
on whether
precise
CEOs
are rewarded for observable luck.
what agency theory says about the reward
for luck.
A
simple theoretical
Consider a standard
agency setup where risk-neutral shareholders, perhaps operating through the board, try to induce
a risk-averse top manager to maximize firm performance.
Since the actions of the
CEO
can be
'This confirma the sentiment that emerges from conversations with compensation consultants as they describe
being hired by two types of firms: the ones where they clearly have to respond to the
clearly
have to respond to the shaureholders.
CEO
and the ones where they
hard to observe, shareholders
will
be unable to sign a contract that specifies these actions. Instead,
shareholders will offer a contrax:t to the
CEO
where her compensation
level is
made
depend
to
on the firm's performance. Let p represent firm performance and a the CEO's actions, which by
assumption are unobservable by the shareholders. Firm performance depends on the actions of the
CEO
and on random
factors.
We
split
the
random
feictors into
observed by shareholders and those that cannot. For an
an observable random
term,
factor.
Under some technical conditions, Holmstrom
two
p and
o,
oil
CEOs
s
for
u be the unobservable
cuid
Milgrom (1987) calculate the optimal incentive
Since shareholders can only observe
= Q + /3(p - Jo) = a + ;0(a + u)
filters
(1)
the observable luck firom performance. This
because leaving o in the incentive scheme provides no added benefit to the principal
definition, the agent has
no control over
Beyond providing no
effects.
o.
"Incentivizing" her
How
more
benefit, it actually costs the principal
we need
filtering of luck to
to assume that the
general result can be found in
^In Section 3.2.4,
by
because not
filtering
out luck
mean pay
risk averse agent.
might we expect
'Essentially,
A much
as,
on o has therefore no incentive
increases the variance of the incentive scheme, thereby forcing the principal to increase
compensate the
noise
performance net of the observable factor.
In other words, the optimal incentive scheme
to
would be
the incentive scheme could at most depend on these two variables. In fact,
shareholders wiU only reward
is
crude
as:
for this model.^ Let s denote this incentive scheme.
variables,
firm, the price of
Letting o be the observable factor and
we assiune that performance can be written
scheme
oil
two components: those that can be
we
will consider
CEO
occur in practice? Explicit incentive contracts, such as
has
CARA
Holmstrom
utility
and that outcomes
follow
(1979).
changes to this basic model that might alter these
results.
a Brownian process.
options, rarely
filter.
For example, options axe rarely
if
ever indexed against market performance.
This need not be inconsistent with a lack of filtering, however. One might expect that the best way
to filter luck
through the use of subjective performance evaluation. In practice, we would expect
is
the board to be the primary mechanism for doing this.
components of
pay, such as bonus, salary or
new options
The board should
use the discretionary
grants, to respond to luck shocks.
Existing Evidence
2.2
Previous work already hints at a relationship between pay and luck.
and
de-Silanes
Shleifer (1992) find that windfall gains
from court rulings
Second, as we just mentioned, the stock options that are granted to
While
to the market.
interesting, these
luck, but rather the result of the
CEOs
two facts are only suggestive.
CEO's work. And,
First,
A
Blanchard, Lopez-
pay
raise the
of
CEOs.
are very rarely indexed
court ruling
as noted, while options
may
may
not be
not be indexed
(perhaps for simplicity or tax reasons), the board can always reissue them or adjust other parts of
pay.
For many, the apparent lack of relative performance evaluation (RPE)
existing piece of evidence of
First,
may
mean
pay
for luck.*
Even with
this evidence though,
industry movements aie a special form of luck.
fact.
themselves note that relative performance evaluation can distort
actions that affect the average output of the reference group."^
a
specific
model
in
probably the best
some problems
arise.
Filtering this specific kind of luck
not be optimal from an agency theoretic point of view. In
also develop
is
Gibbons and Murphy (1990)
CEO
incentives
if
they can "take
Aggarwal and Samwick (1999b)
which relative performance evaluation schemes may not be optimcil
*While Gibbons and Murphy (1990) find some evidence for RPE, some of their results are puzzling. For example,
there appears to be more filtering of general stock market shocks than of shocks specific to the relevant industry.
Most
of the other
evidence of
®They
work on the
topic, such as
Janakiraman
et al. (1992)
and Aggarwal and Sajnwick
(1999a), find
no
RPE.
list
externalities.
four different kinds of such actions:
sabotcige, collusion, choice of reference
group and production
for shareholders aa they try to provide
We
criticisms.
movements.
will
CEOs
with the best incentives. ^"^ Our test addresses these
examine a variety of shocks, including but not
We will verify that other shocks to performance that are
managerial influence, such as shocks to the price of crude
be
oil
restricted to
mean
industry
even more objectively beyond
or exchange rates shocks, also
fail to
filtered.
Empirical Methodology
2.3
Within the agency framework, moat of the empirical
Vit
where
denotes total
yit
fixed effects,
Xn
CEO
=
oii
+ at + ax
*
^it
compensation in firm
are firm (and
resents a performance measure.
CEO)
The
i
literature estimates
+ P * per fit +
at time
t, a,-
an equation
eu
(2)
are firm fixed
efi'ects,
specific variables such as tenure or firm size,
coefficient
^
of the form:
at are time
and per fit rep-
captures the strength of the pay for performance
relationship.
Performance
and we
will xise
literature
is
typically
measured either as changes
both measures.^*
and our paper
is
in accounting profits or stock
In measuring compensation,
no exception.
Ideally, the
y^t,
market returns
one problem permeates the
compensation in a year would include the
change in value of unexercised options granted in previous years. Such a calculation requires data
on the accumulated stock of options held by the
information on
include this
new options granted each
component of the change
CEO each year,
whereas existing data contains only
year. Consequently, our
As we
in wealth.
compensation measure does not
discuss in Section 3.2.1, this data problem
'''The basic idea is that the niean performance in the rest of the industry may actually provide some information
about the CEO's effort to attenuate or strengthen the level of competition in that industry. For example, if product
market competition in an industry needs to be softened, shcireholders might in an efficient contracting environment
reward the CEO when the rest of the industry is doing well as this might indicate that the CEO acted successfully
in
attenuating the competitive pressures in that industry.
"These
profits as
are flow measures. In practice, given the firm fixed effects,
measures of perfn.
10
we
will
use market value and level of accounting
should,
if
anything, bias our results towards overstating the case towards filtering and understating
the extent of pay for luck.
To estimate the general
sensitivity of
pay to performance, we
will
foUow the
estimate equation (2) using a standard Ordinary Least Squares (OLS) model.
pay to
sensitivity of
stage,
we
luck,
will predict
we need
more complicated two stage
to use a
two stage procedure
where the luck variable
is
Letting o be luck, the
procedirre.
estimate the
In the
first
is
how
sensitive
pay
is
to these predictable changes
essentially aJi Instrumental Variables (IV) estimation
the instrument for performance.^^
first
equation we estimate
is:
~ ai+at + ax * X^ + b * oa +
perfit
where ou represents the luck measure
(oil
price for example).
predict a firm's performance using only information about luck.
this luck
To
and
performance using luck. This will isolate changes in performance that are
caused by luck. In the second stage, we will see
in performance. This
literatiure
en
(4)
Again, this equation allows us to
We
then ask how pay responds to
induced performance, perf^^:
yu
The estimated
=
ai
+ at + ax * Xu + PlucK
coefficient pLj^ck indicates
firom luck. Since
such changes should be
how
pay
sensitive
filtered, basic
*
per fit
is
+ en
to changes in
(5)
performance that come
agency theory predicts
/3[,y,ck
should equal
0.
'^One might wonder why we should use this procedure rather than simply include o directly into the pay
performance equation (2) and run OLS to estimate:
Vit
=
at
+ at + ax
*
Xit
+
*
perfu
+7
o
+ e<e
for
(3)
hard to interpret, however. Even if there is no pay for luck, the coefficient 7 will not equal —p but
see from equation I. Since we we do not estimate S, the estimated coefficient 7 can be small
either because there is pay for luck or simply because S is small. The first equation in the IV procedure, on the other
hand, scales the effect of luck on performance, circumventing this problem.
This equation
rather —05, as
is
we can
11
Testing for Pay for Luck
3
Study
Oil Industry
3.1
We now tiim to
the
the price of crude
oil
oil
industry to test whether in fact there
profits.
would have found that a measurable variable
Over
US CEO.
this period,
(oil prices)
Moreover, these large fluctuations in crude
control of a single
oil
its
OPEC.
price increase between 1979
Oil price
As Figure
1
shows,
and 1981
is
of one of these firms
greatly influenced the performance of
prices are likely to have
oil
been beyond the
CEOs
US
of
end of 1985
price at the
petroleum policy and to increase
an action hardly attributable (and never attributed) to the
oil
CEO
a
For example, the sharp decline in crude
was caused by Saudi Arabia's decision to reform
large
zero pay for luck.
has fluctuated dramatically over the last 25 years. These large fluctuations
have caused large movements in industry
his firm.
is
oil firms.
it
production,
Similarly, the
usually attributed to an internal policy change by
movements therefore provide an
ideal place to test for
pay
for luck:
they affect
performance, are measurable and axe plausibly beyond the control of the CEOs.
We use a data set on the pay and
performance for the 51 largest
and 1994 to implement the methodology of the previous
analysis,
it is
In Figure 2,
industry.
useful to look directly at
we have graphed changes
Two
how pay
section. ^^
same
pay changes
sign:
both
au^e
oil
companies between 1977
Before moving to regression
compared to the movements in Figm-e
in oil prices for each year
striking facts emerge. First,
In 12 of the 17 years, they are of the
fluctuates
US
1.
and changes in mean log pay in the
cind oil price
up or both
changes correlate quite weU.
are
down. This
is
suggestive of
a large amount of pay for luck. The second fact comes firom the remaining 5 years where they are
'^We are extremely grateful to Michael Haid for making the data set used in this Bection available to us. See
Haid (1997) for further details about the construction of the data set. Tiible Al in the appendix provides mean
and standard deviations for the main variables of interest. While the original data set covers 51 companies over the
period 1977 to 1994, information on CEO pay is available for only 50 of these original 51 companies. Moreover, CEO
pay is available from 1977 on for only 34 of these 50 companies. The final data set we use covers 827 company /year
observations.
12
of opposite sign:
all
these 5 years
CEOs
an asymmetry: while
for
bad
luck.
This
to the
While we
figure,
pay
will
cire
years in which
oil
are always rewarded for
price drops but pay does not. This hints at
good
luck, they
not formally pursue the asymmetry,
it is
may
not always be punished
worth keeping in mind.
how
however, does not allow us to quantify the size of pay for luck:
for general
performance?
might be changing over time. Table
It also
1
does
it
compare
does not control for other firm specific variables that
follows the empirical
methodology presented
in Section 2.3,
which allows a more systematic analysis. The regressions use log(total compensation) as dependent
variable
and include firm
dependent
variables.*'*
fixed effects, age
We
also allow for a year quadratic to allow for the fact that
has been trending up during this period.
change
in
and tenure quadratics and a performance measure as
accounting performance.
The
Column
(1) estimates
in pay.
coefficient of .82 suggests that
if
an
oil
Note that the sign and magnitude of all the other covariates
firm increases
* .01
one percentage point increase in accounting returns leads to a
Pay increases with age and
.8
=
.0082
«
1
its
log
percent increase
in the regression
seem
Both the age and tenure
to a lesser extent with teniu-e.
pay
the sensitivity of pay to a general
accounting return by one percentage point, total compensation rises by .82
points. Roughly, a
CEO
sensible.
profile are
concave (the negative coefficient on the quadratic term).
Colimin
with log
due to
(2) estimates the sensitivity of pay to luck.
oil prices.'^
oil price
As
described,
we instrument
for
performance
This aUows us to narrow down on movements in accounting returns that are
changes.
The
coefficient in
column
percentage point rise in accoimting returns due
standcird errors, one cannot reject that the
coefficient are the Scune.
One can however
'"Tbtal compensation in this table
and
all
pay
(2)
now
rises to 2.15.
to luck raises
This suggest that a one
pay by 2.15 percent. Given the large
for luck coefficient
and pay
for general
strictly reject the hypothesis of
complete
performance
filtering:
oil
others includes salary, bonus, other incentive payments and value of
options granted in that year.
"This table does not report the first stage regressions of performance measure on
expect from Figure 1, these regressions show very significant coefiicients on oil price (p
13
oil price.
<
.001).
But as one would
CEOs
comes from
are paid for luck that
Columna
(3)
and
(4)
The
shareholder wealth.
oil
price movements.
perform the Scune exercise but for a market measure of performance:
coeflBcient of .38
on column
(3) suggests that
shareholder wealth leads to roughly a .38 percent increase in
CEO
a one percent increase in
pay. In
column
(4),
we
a one percent increase in shareholder wealth due to luck leads to .35 percent increase in
Again, pay
for luck
matches pay
More General
3.2
The
oil
for general
CEO
pay
performance.
much
to
a lucky dollar than
how
foreign competition, exchange rate
Movements
influence
in
how
CEO
mean
It is
only one industry,
We
we
focus on two measures of luck:
By
affecting the extent of
movements can strongly
affect
will
examine
movements
in
import penetration
a firm's profitabihty.
industry performance also proxy for luck to the extent that a
CEO
does not
the rest of her industry performs.^^
Compensation Data
To implement these
1991 period.
is
generalizable these results are. In this section,
exchange rates and mean industry performance.
and hence
performance dollar. This
to a general
luck shocks that affect a broader set of firms.
Shlerfer.
pay.
Tests
however, and one might ask
3.2.1
CEO
industry case study has been instructive about the magnitude of pay for luck.
responds as
find that
we use compensation data on 792
tests,
The data
set
was graciously made available to us by David Yermack and Andrei
extensively described in
the corporations'
SEC
Yermack
Proxy, 10-K, cmd 8-K
(1995).
filings.
Compensation data was
As we mention
before, this \ast assumption
like
is
more
SEC
questionable.
the exx^hange rate movements (and like
14
collected
from
Other data was transcribed from the Forbes
magazine annual survey of CEO compensation as well as from
movements operate exactly
large corporations over the 1984-
oil
Registration statements, firms'
In practice,
we
find that
price movements).
mean
industry
direct correspondence with firms, press reports of
Annual Reports,
stock prices pubhshed by Standard
&
CEO
hires
and departures, and
Firms were selected into the sample on the basis
Poor's.
of their Forbes rankings. Forbes magazine publishes annual rcinkings of the top 500 firma
on four
dimensions: sales, profits, assets and market value. To qualify for the sample a corporation must
appear in one of these Forbes 500 rankings at
least four times
between 1984 and 1991. In addition,
the corporation must have been publicly traded for four consecutive years between 1984 and 1991.
Yermack data
is
attractive in that
it
provides both governance variables and information
options granted, not just information on options exercised.
in the value of options held.
Our compensation measure,
It
on
does not, however, include changes
therefore,
wiU not include the change in
value of pre-existing options. If anything, this biases us towards finding less pay for luck than there
Since options are not indexed, changes in the value of options held wUl covary perfectly with
is.
luck. Including
them would only increase the measured pay
for luck. This data limitation therefore
pushes us towards favoring the filtering hypothesis, making our statistical rejection of complete
filtering that
much
stronger.
Tkble 2 presents summary
statistics for the
main
All nominal variables are expressed in 1991 dollars.
in salary
and bonus. His
thousand
dollars.
The average
compensation
is
CEO
earns 900 thousand dollars
amount
CEO's pay
CEO
data.^'^
at one million
and 600
that are due to options
As
governance goes, the average firm in our sample has 1.12 large shareholders, of which
less
them are
'''in
CEO
nearly twice that
difference indicates the large fraction of a
is
roughly 57 years old and has been
than a fourth are sitting on the board. There
of
The average
Yermack
yccirs.
grants.
far as
The
total
variables of interest in the full
cire
of the firm for 9
on average 13 directors on a board. 42 percent
insiders.^®
practice,
depending on the required regressors, the various
various subsamples of the original data.
None
of these
main
tests in the following sections will
be performed on
variables of interest significantly differ in any of these
subsamples.
'^Ttechnically,
we
define "insiders" to
be both inside and grey
15
directors.
An
inside director
is
defined as a director
Exchange Rate Movements as Measures of Luck
3.2.2
Our
first
general measure of luck focuses on exchange rate movements.
exchange rates between the
US
dollar
may be more
same
yeax.
affected
may be more affected by
by Bolivia, these two industries
by
different countries'
exchange
rates.
We
For
Japanese imports while the lumber industry
may
experience very different shocks in the
This allows us to construct industry-specific exchange rate movements which we are
arguably beyond
The exchange
for
exploit the fact that
and other country currencies fluctuate greatly over time.
also exploit the fact that different industries are affected
example, since the toy industry
We
CEOs'
control since they are primarily determined by
rate shock measure
is
macroeconomic
variables.
based on the weighted average of the log real exchange rates
importing countries by industry. The weights are the share of each foreign country's import in
total industry
imports in a base year (1981-1982). Real exchange rates are nominal exchange rates
(expressed in foreign currency per dollar) multiplied by U.S.
CPI and divided by
the foreign country
CPI. Nominal exchange rates and foreign CPI's are from the International FinanciaJ Statistics of
the International Monetary Fund.
Panel
A
of Table 3 examines the effect of this measure of luck. Note that since the exchange
rate measure can only be constructed for industries where
is
much
smaller here than for our
full
as well as for quadratics in tenure
in millions
and pay
levels
is
and
Column
age.^^
and do not include
(1) uses as
dependent variable the
level
Thus, relative to our standard specification, we
Vcilue of
options granted. Since profits are reported
reported in thousands, the coefficient of .17 in colmnn (1) suggests that
is a current or former ofRcer of the company. A grey director
has substantial business relationships with the company.
that
'*We do not report the
data, the sample size
sample. All regressions control for firm and year fixed effects
(not the log) of cash (not total) compensation.
run this regression in
we have imports
coefficients
is
a
relative of
a corporate
officer, or
on age Jind tenure to save space, but they resemble the
and age.
positive but diminishing effects of tenure
16
someone who
oil indxjstry findings:
SI, 000 increase in profits leads to
now
exercise but
As
pay
for
in the oil case,
for general
Columns
pay
we
for luck:
pay
find a
performance
(3)
through
a 17 cent increase in performance. Column
we instrument
for
performance using the exchange rate
for luck coefficient that is of the
we
(6)
run the more standard regression where we use the logarithm of pay
When
(either total or cash) rises
movements
columns
in
(5)
and
(6)
total assets). In
we use
columns
total compensation.
be about the same as the
find the sensitivity of pay to luck to
to a general shock.
as the
coefficient.
we use only cash compensation, while
In both cases,
shocks.^'^
same order of magnitude
and an accounting measure of performance (operating income divided by
(3) BJid (4)
performs the same
(2)
sensitivity of
pay
accounting performance rises by one percentage point compensation
by about 2 percent, whether that
rise
was due to luck—exchange rate
—or not.
Colmnns
(7)
through
(9) replicate these foiir
colimans but for market measures of performance.
Again, we find pay for luck that matches the pay sensitivity to a general shock.
wealth of one percent leads to a
rise in
pay (again either total or cash) of about
A rise in shareholder
.3
percent irrespective
of whether this rise was caused by luck.
Two important points should be taken away
average firm rewards
its
There seems to be very
CEO
as
little if
much
any
from this panel.
for luck as
filtering.
it
Since
First, to
a first approximation, the
does for a generaJ movement in performance.
we
use a totally different shock, these findings
address theoretical concerns about the use of mean industry shocks (such as those raised in Gibbons
and Murphy (1990) and Aggaxwal and Samwick (1999b)) and shows that the lack of
observed in
RPE
Second, there
^''First
filtering
findings generalizes to other sources of luck.
is
as
much pay
for luck
on
totally discretionary
stage regressions are reported in 'Rible A2. In practice
we use
components of pay
as instruments a set of
indicate any significant exchange rate appreciations or depreciations in the current
are jointly highly significant.
17
and past
dummy
year.
(salary
jmd
variables that
The instruments
bonus) as there
pay
for luck
is
on other components such as options granted. This
arise because firms
might
commit
plans where the number of options grants
is
(implicitly or explicitly) to multi-year stock option
fixed ahead of time.
options rise and so does the value of options grants (Hall, 1999).
subjective
component of pay
for
performance
quite suggestive. Boards are rewarding
3.2.3
Mean
In Panel
mance
B
CEOs
rules out the notion that
sensitivity.
for luck
To
find
As
firm vaJues rise, the prices of
More
pay
bonus
generally,
for luck
even when they could
on
this
is
the most
component
is
filter it.
Industry Movements as Measures of Luck
of Table
we
3,
replicate Panel
of the industry, which
More
firms in the industry.
in a given year,
meant
except that our measure of luck becomes
specifically, as
firm in a given year
to.
to,
an instrument
for firm-level rate of
Similarly, as
by
perfor-
all
the
accounting return
rate of accounting return in that year in the 2-digit
excluding the firm itself from the calculation.^^
the share of
is
mean
to capture external shocks that are experienced
we use the weighted average
industry that firm belongs
the firm belongs
is
A
its total assets in
an instrument
The weight
of a given
the "total" total assets of the 2-digit industry
for firm-level logarithm of shareholder wealth in
a given year, we use the weighted average of the log values of shareholder wealth in the
2-digit
industry in that year, again excluding the firm itself from the calculation and using total assets to
weight each individual
Like in Panel A,
for a quadratic in
firra.^^
all
regressions include firm fixed
CEO
age and a quadratic in
twice the data points of Painel
'*
We
A
CEO
tenure.
because we can now use
also investigated the use of 1-digit
and
3-diglt industry
and year
eflFects
means
The
fixed efiects.
We
also control
regressions include
all firms,
more than
not only those in the traded
as instruments
and found
qualitcitively similar
results.
'^These mean industry performance measures are constructed from
COMPUSTAT. To
the performance measures between Yermaclc's firms and the rest of industry,
income to assets
present in
ratios
from
COMPUSTAT
in
COMPUSTAT
every year,
we
for the firms in
lose
Yermack's. Because not
about 800 firm-year observations.
18
maximize consistency of
we also compute shareholder wealth and
all
firms in Yermack's data aie
B
goods sector. Panel
all
shows a pattern quite similar
specifications again roughly
matches the pay
to
Panel A.
for general
The pay
for luck relationship in
performance. Besides reinforcing the
findings of Panel A, these latest findings suggest that previous
RPE
results arose probably not
because of miameasurement of the reference industry or of the industry shock but because of true
pay
for luck.
Could Pay
3.2.4
for
Luck Be Optimal?
Since the results so far clearly establish pay for luck or a lack of performance filtering,
to ask whether pay for luck could in fact be optimal.
that would generate pay for luck in
we have assumed that o
First,
is
important
for
most of our
an optimal contract?
We
could think of at least three.
contractible. If o cannot be contracted
findings. First,
natural
Are there complications to the agency model
upon or measured
the empirical testa developed in section 2.3 are no longer valid. In practice,
is
it is
it is difficult
we do not
well,
believe this
to believe that non-contracting issues
can
explain our results in the oU industry case study: the price of crude oil can easily be measured ajid
written into a contract. Second, even in the presence of non-contractibility, subjective performance
evaluation should efl'ectively
filter (see
Baker, Gibbons and Murphy, 1994).
subjective performance evaluation relates to
problem
for the
commitment
issues,
for
boards of directors of the large firms we study here.
example to
forecast,
most readily evaluated
oil
respond
to,
may want
to provide to the
or hedge against luck.^^ This kind of argument can be
in our oil industry application.
Suppose a particularly talented
CEO
in the
industry understood the political subtleties of the Arab countries and forecasted the coming of
^'
the
only problem with
but these are unlikely to be a
Second, we have neglected the possibility of indirect incentives one
CEO,
The
An argument similar
CEO
to hedging has been
made by Diamond
(1998).
Tying pay to luck may generate incentives for
would seem to be in the interest
to change his correlation with the luck variable. In practice, diversification
of management
and not
sheixeholders. Tiifano (1996), for
as share or option ownership, are quite predictive of risk
example, demonstrates that managerial characteristics, such
management
19
in
a sample of gold
firms.
the positive
increasing inventories, or intensifying search for
when
here
is
that those
CEOs who were
what we
test for.
as every other oil
may want
to reward
We
he could have increased his firm's profits
is
to reward
them
CEO who
CEOs
responds to the shock exactly the same way
for exceptional responsiveness to shocks, but there
oil
CEO
could be affected by the
human capital
CEOs more
of oil
is little
Again,
reason
random
factor o.
When
CEOs may simply become more
the
valuable.
simply to match their increased outside options. Thus, pay
Some objections can be raised against
CEO human capital should become more valuable as
bad times that having the right
this
view as well. First,
industry fortunes
rise.
it is
unclear
For example,
it
why
may be
CEO is most valuable. A priori, either relationship seems
Second, we have found some evidence of asjonmetries in the pay-for-luck relationship
that are hard to reconcile with that view (and easy to reconcile with a
luck" seems to matter
industry always goes
more than "bad
up when the
price of crude oil goes
it
well.
optimal here not as an incentive device, but merely becatise the optimal level of pay
increases with luck.
plausible.
shock.
for just average responsiveness.
Firms then pay their
exactly in
oil
rise (or fall, for the negative shocks) in pay.
rewarded. This cannot be a reward for having forecasted
industry enjoys good fortune, the
for luck is
CEO? The important point
use none of the firm to firm variation in response to the
results suggest that a
CEO
oil wells,
exceptional in having forecasted should indeed be rewarded. But
Third, the reservation utility of the
oil
wells,
merely test whether the average firm experiences a
Put another way, our
one
new
increasing output from existing
the shock did come. Shouldn't shareholders reward this farsighted
this is not
We
By
shock at the beginning of the 1980's.
oil
luck". For example, average
price of crude oil goes
skimming view). "Good
CEO
compensation
in the oil
up but does not alwaj^ go down when the
down. Also, when we separate positive and negative mean industry shocks,
appears that pay responds more to positive indiistry shocks than to negative industry shocks
20
(while there are no asymmetries in the sensitivity of pay to individual firm performance).^^
Finally,
one might argue that pay
for luck is
a form of risk sharing between
firms are effectively risk averse, perhaps because of liquidity constraints, they
some of the
oil
risks they face
industry study.
awash
in cash
would need
It is
with their
CEO. This argument
hard to imagine why the large
oil
is
CEOs
ajid firms. If
may want
to share
best evaluated in the context of our
producers we study, which are usually
and have excellent bond ratings (perhaps because large holdings of pledgeable
to share risk
take futures positions in
take on the risk than the
with the
oil
CEO. Even
if
these firms needed to lay off risk, they can readily
which would allow them to shaxe the risk with a market better able to
CEO.
Similarly, exchange rate futures
markets are active. Of course, one
could argue that for
mean industry movements
futures markets
But
would then only apply to
this third luck
this explanation
assets),
may
not be as readily available.
measure, whereas our results hold
similarly for all the measures.
While we have provided arguments against these various extensions of the simple agency model,
in the
end we
stiU believe that they merit serious consideration.
for luck finding does not
corporate governance affects
3.3
The
Effect of
We
per se rule out agency models.
CEO
They suggest
to us that the pay
deal with this issue by examining
how
pay.
Governance
The skimming view makes a
further prediction.
Since
it
emphasizes the CEOs' ability to gain
control of the pay process, corporate governance should play
exactly in the poorly governed firms where
process. This suggests that
To examine how pay
we expect CEOs
we should expect more pay
for luck differs
an important
to
most
role in
skimming.
easily gain control of the
for luck in the poorly
between well and poorly governed
It is
pay
governed firms.
firms,
we estimate two
'*These results, not directly reported here, aie available upon request from the authors. They hold quite strongly
accounting measures of firm performance but are not present for market measures of firm performance.
for
21
equations. First, in order to provide a baseline,
differs
(2)
between well and poorly governed firms.
we allow the pay
except that
=
Vit
where Govu
is
+
ai
at
for
we ask how pay
We
performance
for general
estimate an
coefficient to
OLS
performance (not luck)
equation similar to equation
depend on governance:
+ ax * Xu + ao * Govu + 0* perfu + 7 * {Govu
*
perfu}
+
eu
(6)
a measure of governance. To xmderstand this equation, differentiate both sides with
respect to perfu:
dperfit
The
for
7 would imply
Equation
To
CEO
sensitivity of
get at
procediire.'^
that better governed firms
(6) of
pay
course
for luck,
We
pay to performance depends on the governance
tells
we
greater
pay
for
A
positive value
performance.
us nothing about pay for luck, merely about pay for performance.
re-estimate this equation using our two stage instrumental Vcuriables
then compute an estimate of the
mance, 7 and an estimate of the
effect of
Oiu- test then consists in comparing
governed firms
show
variable.
when poorly governed
effect
of governance on pay for general perfor-
governance on pay for luck, ^luck-
7 and
jLuck-
firms display
We
will
speak of more pay for luck in poorly
more pay
for luck relative to
pay
for general
performance. If poorly governed firms simply gave more pay for performance and pay for luck rose
as a consequence,
is
pay
we would not
for luck that
refer to this as
more pay
for luck. In practice,
we
will see that
it
changes with governance, while pay for performance hardly changes.
^"'An extremely important caveat here:
a different responsiveness of performance
to instrument, appears both directly and
we perform two first stages, one for the
procedure is cruciaJ because it allows the
our approcich allows for the possibility that better governed firms may have
to luck. Technically, performance per fa, the endogenous variable we need
indirectly (the term Govit * perfu) in this equation. ^Vhen we instrument,
direct effect
perfu and one for the interaction term GotHt
on performance to depend on governance.
effect of luck
22
*
ptrfu- This
Large Shareholders
3.3.1
In Table
4,
we implement
this
framework
We
for the case of large shareholders.
presence of large shareholders affects pay for luck. Shleifer and Vishny (1986),
that large shareholders improve governance in a firm.
A
single investor
who
ask whether the
among others, argue
holds a large block of
shares in a firm will have greater incentives to watch over the firm than a dispersed group of small
shareholders.'^ In our context, the idea of large shareholders
our
intiiition of
most naturally as
this
matches
"having a principal around." Yermack data contains a variable which counts the
who own
nimiber of individuals
further
fits
know whether
blocks of at least 5 percent of the firm's
these large shareholders are on the board or not.
that large shareholders
A
common
priori,
on the board have the greatest impact. They can exert
shares.^^
We
one might expect
their control not just
through implicit pressure or voting, but also with a direct voice on the bocird. Since the information
available,
is
The
AU
we
first
will consider the effect of
both
four columns of T^ble 4 use
regressions include the usual controls.
performance depends on governance
for
all
large shareholders
and of only those on the board.
aU large shareholders as our measure of governance.
Column
(1) estimates
how
the sensitivity of pay to
accounting measures of performance.
The
first
row
tells
us
that a firm with no large shareholders shows a sensitivity of log compensation to accounting return
of 2.18.
An
increase in accounting return of one percentage point leads to
about two percent. The second row
tells
in
pay of
us that adding a large shareholder only weakly decreases
the sensitivity of pay to general performance, and this effect
example, a one percentage point increase in accounting retm^i
in
an increase
is
not statisticadly significant.
now
For
leads to a 2.09 percent increase
pay when the fixm has one large shareholder and not a 2.18 percent increase in pay. Colmnn
They
also point out
a possible opposite
rents for thenaselves. This effect
"Whenever CEOs happen
to
is likely
effect:
very large shareholders
may have a
greater ability to expropriate
to be greatest in other countries where investor protection
own such a
block,
we exclude them from the count.
23
is
weakest.
(2) estimates
there
how
pay
significant
is
pay
large shareholders affect
for luck.
for luck.^
The second row here
As
row
before, the first
tells
tells \is,
however, that this pay for luck
A
one percentage point increase in
diminishes significantly in the presence of a large shareholder.
accounting returns due to luck leads to roughly a 4.6 percent increase in pay when there
shareholder but only a 4.2 percent increase in pay
when
additional large shareholder decreases this effect by
.4
us that
there
is
is
no large
one more large shareholder. Each
percent. This
is
a 10% drop in the pay for
luck coefficient for each additional large shareholder.
Columns
,/;
(3)
In this case, the
estimate the same regressions using market measures of performance.
and
(4)
pay
for general
performance does not depend at
shareholder (a coefficient of .001 with a standard error of .007).
diminishes with the presence of a large shareholder.
for luck
10%
the
for
level,
the economic magnitude
is
The pay
larger.
aJl
We
WhUe
on the existence
of a large
again find, however, that pay
the result
for luck coefficient
is
only significant at
now drops
^ « 17%
each large shareholder.
In columns (5) through {8)
We now
focus only
on
we repeat the above
exercise but altering the governance measure.
Comparing columns
on the board.
large shareholders
(6)
and
(2),
we
see that the governance effect strengthens significantly with respect to the filtering of accounting
performance.
The
We see that the pay for luck drops by 33 percent for each additional large shareholder.
results are very statistically significant.
also rises
but
less dramatically.
In column
On
(8),
market performance measures, we find the
effect
the pay for luck drops 23 percent with each large
shareholder on the board. Moreover, this last result
is
insignificant. In
summary, our findings
in
Table 4 highlight how large shareholders (especially those on the board) affect the extent of pay
for luck.
*In
all
powerful
Firms with more large shareholders show
that follows,
first
we
stages in the
far less
use mean industry performance as
IV framework.
will
24
pay
otir
for luck.
measure of luck since this produces the most
Entrenchment and Large Shareholders
3.3.2
The
results in
CEO
the effects of
CEOs who
Table 4 simply compare firms with large shareholders to firms without. This ignores
pay
governance
belief
is
that
have been with the firm longer have had a chance to become entrenched, perhaps by
appointing friends on the board.
greatest
A common
tenure, another important determinant of governance.
for luck.
we
woxild expect high tenure
Moreover, we would expect this
weak and there
is
In this case,
no large shareholder present
is
Hence, in the absence of large shareholders, we expect
be strongest
effect to
fairly
CEOs
to
show the
in those firms
where
to limit the increased entrenchment.
CEO's
strong governance early in a
tenure but this governance should weaken over time as he entrenches himself. In the presence of
large shareholders,
we not only expect stronger governcince but
shoidd last throughout the
there
is
CEO's
tenure. It
is
CEO
harder for a
a large shareholder aroimd. Thus we expect a
aJso that this stronger governance
rise in
pay
to begin stacking the
for luck
board when
with tenure in the absence
of a large shareholder, but less of a rise (or even no rise) in the presence of a large shareholder.
T^ble 5 tests this idea.
We
first
sort firms into two groups based
on whether they have a
large shareholder present on the board.^^ This produces 740 or so data points for firms with large
shareholders and 3880 or so data points for firms without large shareholders. For each
separately estimate regression 6 for these
Columns
(1)
shareholder.
^5^
w
with columns
^We
(2) focus
The second row
greatly increases
about
and
pay
35%
(3)
for luck. In fact, a
(4)
of performance in firms without a large
us that while tenure does not affect pay for performance,
CEO
with (roughly) the median
greater pay for luck than one
and
we now
two groups with tenure as our governance measure.
on accounting measures
tells
set,
who just began
which estimates the same
effect for firms
teniure of 9 years
shows
at the firm. Let us contrast this
with a large shareholder present.
focus only on large shajeholdeis on the bojird because these provided the strongest results in Table 4.
have used
aJl
large shareholders
and found
similar,
though
statistically
25
it
weaker, results.
We
Here we find that tenure does not
for
performance
slightly. '^°
pay
affect
Thus pay
for luck at all, while, if
for luck increases with tenure
anything,
it
seems to
pay
raise
without large sliareholders but
does not change with tenure with them.
Columns
(5)
through
repeat the tests of columns (1) through (4) but for market measures
(8)
of performance. Here the results are less stark but
and
(5),
we
see that both pay for performance
rises three times as fast (.003 versus .009).
significant at the
10%
level.
The economic
both diminish with
3.3.3
(7),
we
and pay
The
very suggestive. Comparing columns
for luck rise
coefficient
with tenm-e, but pay
on the pay
see that
=^^^ « 31% but a
if
rise in
a
pay
(6)
for luck
for luck, however,
significance however stays large as
of 9 yeaxs shows an increase in pay for luck of
only 10%. In columns (8) and
still
only
is
CEO with a tenure
for
performance of
anything pay for luck and pay for performance
teniu-e.
Other Governance Measures and Robustness Checks
While large shareholders correspond most
closely to the idea of
a principal, other governance
measures could also be used. Our data contain two variables that have shown to be important
governance measures in the past: the size of the board and the fraction of board members that are
insiders in the firm.^'^ Small boards are thought to
(1996), for example,
be more
shows that smaller boards correlate with larger q values
four columns in Table 6 estimate the effect of board size
that for accounting measures, the direction of the effect
the coefficient
'^Gibbons
effective at governing firms.
is statisticcilly
axid
Murphy
also
examined
insignificant.
on pay
is
CEOs
Columns
the opposite of
Note that the actual
(1992) present evidence that
for luck.
for firms.
(1)
and
Yermack
The
(2)
first
show
what we postulated but
size of the coefficient is tiny:
even a
with longer tenure have greater pay for performance
generally.
^*We have
CEO
ownership and whether the founder is present. We do not report these for space
effects. Founders and CEOs with high insider ownership both show
reasons but both produce genercJIy significant
greater pay for luck.
26
huge increase
and
(3)
(4),
in
board
size of 10
board members leads only to a
however, show that there
and these are of the expected
cire
significant effects for
The
6.
a drop of
'^'^^^j^^-
» 30%
in the
pay
other measure we examine
The
is
But,
coefficient of .228.
(6)
it
The
shows a pay
directors
for
10%
is
for luck coefficient of .177.
is
This represents
a measure of insider presence on the board. This variable
level).
negative and small.
directors.
Columns
insider presence dramatically increases the
pay
(5)
for luck
In a board with ten directors, turning one of the outside
from an outsider to an insider increases pay
performance
show a pay
smaller board continues to
for luck coefiicient.
show that on accounting measures,
coefficient (significant at the
Columns
market measures of performance
measured as fraction of board members that are firm insiders or grey
and
for luck.
big board firm shows a pay for performance
coeSicient of .240 and a pay for luck coefficient of .229.
performance
drop in pay
Consider the difference between two boards, one of which
sign.
has 10 board members and one of which has
for
2%
Columns
(7)
by
for luck
and
(8)
^l^*'^
^
20%. The
effect
on pay
show that on market performance
measures, insider board presence again increases pay for luck, but while the coeflacient continues
to be economically large
We
turn to our
last
index that aggregates
all
it is
statistically quite insignificant.
where we construct an
governance measure in columns
(9)
through
the governance measures \ised so
far:
number of large shareholders, number
(12),
of large shareholders on the board, board size and insider presence on board.
we demean each
take the
sum
of the four govemEince variables, divide
it
of these standardized variables. For bocird size,
procedure. For fraction of insiders on the board,
by
its
we use negative of board
we use one minus
*'This particular
way
index,
standard deviation, and then
size in this
that fraction. This guarantees
that the resulting governance index has larger values whenever the firm
and once as generaJ
To form the
is
better governed.^'^
of proceeding will tend to count large shareholders on the board twice, once as on the board
is a crude way of incorporating our prior belief (supported by the findings
l^ble 4) that large shareholders on the board matter more. When we use either measure in the index alone, we find
qualitEktively similar results. We have also estimated a regression in which we include all four governance measures
large shareholders. This
in
27
For accounting measures of performance (columns
(9)
and
(10)),
we again
find that
luck diminishes with the governance, wliile pay for performance does not change.
however,
10%
only significant at the
is
level.
To gauge
and
~
^^^
^^^^ ^^
^^^ P^y ^^^
(12)), increase in the
When we
^^*'^ coefficient.
bigger.
A
Finally,
is
that
5%
level.
for firm size.
affects
effect for a
pay
governance "effect."^^
for luck
One might worry
for luck sensitivities, the estimates
A
by 26%.
that large firms have
7,
we address
this
is
problem by controlhng
for size interacted
*
may pay
(5)
through
(8) of
Table
4.
We see
the filtering of accounting rates of return in fact strengthens
versus —1.48).
The
effects
We
per fa. Omi measure of size in these
average log real assets of the firm over the period.
compared to columns
less for luck.
with performance.
We
investigate
measures: large shareholders on the board and the governance index.^^ Colunms
are to be
size
concrete version of this problem might arise in the risk sharing
re-estimate equation 6 but this time include a term Size
regressions
If this
above might confuse this
hypothesis above. If larger firms are simply more able to bear risk then they
In Table
for
Moreover their magnitude
quite different pay for performance sensitivities than small firms (Baker and Hall, 1999).
also translates into different
pay
The primary concern one might have
investigate the robustness of our findings.
we have not adequately controlled
to a
use market measures (columns (11)
one standard deviation increase in governance decreases pay
we
coefficient,
Such an increase leads
2.
governance index greatly reduces pay for luck but hardly
performance. In this case, the coefficients are significant at the
is
for
the magnitude of these effects, consider
a one standard deviation increase in the governance index, about
~~^23*^
The
pay
when
two governance
(1)
through
that the effect of governance
(4)
on
these controls are added (—2.23
on market performance measures, however, weaken
slightly (.059 versus
(and their interactions) together. These regressions shewed all the governance measures entering with the same sign
large shareholder variables being statistically significant.
We have also attempted other robustness checks. We checked whether jtay for luck happened over longer time
and only the
horizons by aggregating our data over several yeeirs as well as looking at lags.
We also allowed for
interactions between
performance and year in our regressions. These modifications did not alter our qualitative findings.
'^We have reestimated all the previous tables with these controls and found similzhr results.
28
.076).
Columns
(5)
through
through (12) in Table
6.
(8)
use the governance index and are to be compared to columns
(9)
This comparison shows that the results weaken very slightly (—.197 versus
—.216) for accounting measures as well as for market measures (—.027 versus —.033). For both
governance measures, however, the results remain economically significant.
(colxmm 4 and to some extent column
6)
,
In two of the cases
they remain statistically insignificant. In the other two,
they remain statistically significant.
3.3.4
Summary
The pay
of
Pay
for
for luck regressions
Luck Findings
have helped establish
five
the direct finding of a pay for luck relationship. First,
are rewarded for general performance changes.
measures, such as
oil
movements. Third,
price changes
CEOs
are as
main
CEOs
facts.
The
first
three facts relate to
are rewarded for luck as
much
as they
Second, the pay for luck occurs for various luck
and exchange rate movements, and not only
much rewarded
for luck
for
mean
industry
through changes in salary and bonxis as
they are through changes in the value of options granted.
The
last
two facts relate to the
effects of
governance on the pay for luck relationship. Foiu-th,
the presence of large shareholders (and to a lesser extent, smaller boards or less insiders on the
board) reduces the extent of pay for luck.
An
additional large shjireholder
by 23 to 33%, depending on the performajice measure. Fifth,
pay
for luck
pay
for luck increases
but only when there
is
no
on the board reduces the
eis
large shareholder present
29
CEO teniure increases,
on the board.
4
Charge For Options
The
turn to a second prediction of the skimming view.
We now
aromid the
appear on balance sheets.^^ Under the skimming
fact that stock option grsmts never
model, this makes options grants an attractive
the balance sheet bottom
line,
idea for this second test centers
way
to skim.
stock options might aJlow the
If
shareholders primarily focus on
CEO
to take
money out
of the firm
without attracting shareholder attention. In skimming models, therefore, options represent "free
Because
money."
make money.
If it
CEOs
are not charged for them, they cannot lose. If the stock does well, they
does not, they are no worse
off.
In contrast to this prediction, agency models predict that
for their options.
options
it
CEOs
should be completely charged
Consider a firm that for an exogenous reason increases the amount of stock
gives to its
CEO. Further suppose
dollars according to Black-Scholes.
an increase in his overall
utility.
that this increase in options
is
worth one million
While these extra options incentivize the CEO, they are
In expected value, he
is
been given one million
dollars.
also
Agency
theory predicts, therefore, that other components of pay would be adjusted downwards in order to
leave the
CEO indifferent between the new
averse, the million dollars in options
terms. So other components of
is
and old compensation packages. Since the
worth
less to
him than one
pay would go down by
Examining whether CEOs are charged
less,
CEO
is
risk
million dollars in expected utility
but they would go
for options could therefore provide
down
nevertheless.
another test of skimming
against agency models.
^'Accounting rules for stock options issued to employees were decided by the predecessor of the Financial Accounting
Standards Board, the Accounting Principles Board. They date back to 1972. See Murphy (1999) for more details.
30
4.1
Empirical Methodology
this test, consider the following regression:
To implement
'
^
Othera
where Otheru
is
the level of
= ai + at + ax
CEO
option grants, Oj are firm fixed
characteristics
and Optionu
is
compensation
effects,
+ P * Optionn +
for firm
in year
i
t,
new
tu
(7)
excluding the value of the new-
Xn
at are time fixed effects,
the level of the
CEOs
Directly testing whether
Xit
*
are firm
/3i,
are being optimally charged
when they
Under the optimal contracting model, we would then expect the
to be negative but larger than —1. Moreover, the higher the
smaller the absolute value of
CEO's
are granted
by a desire
to give the
CEO
other components of pay will also
rise,
coiild
even be positive.
We
more
overall
rise.
required to estimate equation (7)
we can
not on the direct
''in practice,
accounting for
when we
is
CEO
in options grants are
compensation (perhaps because
This suggests that when
compensation.^^
As with the pay
Optionsu on Otheru but on how
directly regress the level of option graints
when we add
will
be biased
to erroneoxisly reject the hypothesis
The
exogeneity assumption
for luck coefiicient,
tliis
effect varies
we can
focus
with the level of
on the other components of compensation, only
CEO tenure and CEO age in addition to the firm and year fixed effects,
.046. Interestingly,
/3
for all
too severe.
get around this problem.
effect of
Optioriit,
Because we cannot perfectly control
would then tend
that individual rationality constraint holds in
Fortunately,
on
level of risk aversion, the
the factors that might induce a rise in overall compensation in Xn, the estimated
upward and
stock
CEO compensation
coefficient
Changes
of changes in the executive's performance or outside opportunities).
options
new
/3.
In practice, however, such a regression would be mispecified.
at least partly driven
specific
stock options granted in that year.
options would require an exogenous shock to Optiouit, the use of stock options in
packages.
CEO
and
we
find
a
positive coefficient of
various firm performance measures (sales, income to assets ratio, level of shjireholder
wealth) and also allow for industry specific shocks (interaictions of time
coefiicient on Optiona goes up, not down (.051).
31
dummies with
2-dlgit
SIC
industries), the
We
governance.
-
Otheru
where Govn
thus propose to estimate instead the following regression:
is
Oi
As
*
+ "G
^it
*
+ P*
Govu
Optionn
+7*
{Optionn
*
Govit)
+ en
(8)
the governance variable under study (ownership concentration for example) ajid
Optioriit * Govit
granted.
+ oit + ax
is
the interaction of the governance variable with the value of the
before, diflFerentiate
both sides with respect
dOtherit
—__
=
/3
+7*
new options
to Optioriit to get:
Govit
oUptionit
for options
showing that the charge
the coefficient
7
is
depends on the
negative indicating that
level of
CEOs
are
governance.
more
We propose to test
cheirged for options
whether
when govemiince
increases.
Could the same biases that
variables problem affects
It is
only
7<
0.
an upward
bias of /? lead to
a bias of 7? In short,
While
firms are hss affected
it is
typically quite hard to imagine
why
scenario where
Suppose following an increase in reservation
occxu-.
well governed firms tended to
it
valid test.
estinaate
a
this particular omitted variable
why
might
the omitted
by the bias that we would spuriously
bias structure shoiild axise (or
it
if
both well governed and poorly governed firm equally, we have a
when well governed
coefficient of
led to
should be similar across governance measures), there
utility or
is
one
performance,
pay out a greater fraction of the subsequent pay increase aa options.
In other words, suppose better governed firms used more options in their pay packages for the
marginal increase in compensation. This difference would lead to an apparently smaller coefficient
/3
for
them and lead
We
can actually
to a bias
test
downward on
whether this
is
7.^^
the case by computing whether in the data stock option
grants are a bigger fraction of total compensation for better governed firms.
appendix shows that, except
^'^For details,
an appendix
is
for the presence of
available
Table A3 in the
any large shareholder, the relative use of stock
from the authors.
32
options grants
is
among
not larger
the firms with better governajice. In
the firms that have at least one large shareholder on their board than
fact, it is
among
4%
smaller for
the firms that do
not-^^
Large Shareholders and the Charge for Options
4.2
In Table
we use
8,
large shareholders as our
that are monitored by large shareholders
(1)
through
(4)
use
all
measure of governance and examine whether the
Eire
more charged
for the options they axe given.
large shareholders as the measure, while
those on the board. In
aJl
regressions, the
dependent variable
columns
is
(5)
through
(8)
CEOs
Columns
use only
the real value of compensation
excluding the value of option grants. Included as regressors in each regression are year fixed effects,
firm fixed effects, a quadratic in age, a quadratic in experience
and performance
income to assets ratio and shareholder wealth. Also included
real sales,
controls such as
are, of covu-se,
the real
value of options grants, the corporate governance variable under study and the interaction of the
governance variable with the real value of options grants.
The
first
row of column
(1) tells
us that, in the absence of any large shareholders, each extra
dollar of option grants leads to another 5 cents in non-option based compensation. This positive
coefficient
emphasizes the omitted variable bias that we
cally significant coefficient
lead
is
CEOs
small.
to be
We
for the options for they are given.
Each additional large shareholders leads the
is
The second row
indicates a statisti-
on the large shareholders interaction term, suggesting
more charged
dollar in options he
face.
CEO
lairge
The economic
shareholders
size,
however,
to be charged an extra cent for each
given.
verify the robustness of our findings to allowing the coefficients
on the governance
variable
In fact, this pattern makes more sense from an agency standpoint. Assuming that governance and pay-forperformance are substitute incentive mechanisms, better governed firms would use less options. It also makes more
sense from a skimming standpoint. Poorly governed CEOs would skim more, since they are charged less for them
and so would have options be a bigger part of
total
compensation.
33
and on the options grants variable to change over time. For example, one might be concerned that
the relative use of options has trended
similar trend. Allowing for year
this concern.
3),
dummies
interacted with the options grants variable takes care of
Allowing the coefficient on large shareholders to change over time (colimin
not affect the basic finding.
(column
up over time and that the governance variable displays a
When we
edlow the coefficient
does
2)
on option grants to be time dependent
the effect of large shareholders on the charge for options, while qualitatively unchanged,
The same pattern
drops in both economic and statistical significance.
on option grants and the
allow both the coefficient
coefficient
holds in column 4 where
we
on large shareholders to chajige over
time.
In columns
(5) to (8),
we focus on
The —.05
stronger results.
large shareholders that sit on the board
coefficient in
column
the charge on options by 5 cents on the dollar.
a
CEO who gets
we
becomes
To
assess the
on the
yesir. If
we add a
(7)
and
large shareholder
50, 000 dollars.
dollar or 30, 000 dollars in our example)
allow for the coefficient on option grants to be time-dependent, but
(colmnns
somewhat
magnitude of these findings, consider
on these options would increase by
slightly smaller (about 3 cents
find
suggests that one extra shareholder increases
one million dollars worth of options in each
to the board of his firm, the charge he faced
effect
(6)
and
it
on
The
when
stays strongly significant
(8)).
In summary, the results in Table 8 confirm oiu: previous finding of an importajit role of large
shareholders in increasing the rationality of
CEO compensation.
holders sitting and not sitting on the board
is
reward
for luck.
When
large shareholder that
is
it
is
comes to
also
stronger here than
financially charging
CEOs
a director appears much more
not represented on the board.
34
The
it
contrast between large share-
appeared in our analysis of the
for the options
effective
they are gramted, a
than a large shareholder that
In the reward for luck test,
we examined
CEO
entrenchment by comparing the
tenure on pay for luck in firms with and without large shareholders on their board.
this exact
(2)
we
same
examine the
exercise in Table 9 in our test of the charge for options grants.
a function of whether there
effect as
is
of
effect
We
CEO
replicate
Columns
(1)
and
a large shareholder present. As before,
separate firms into two groups based on whether they have or not a large shareholder on their
board.
The
patterns in Table 9 strikingly match the patterns in T^ble
CEOs
shareholders on the board,
Each extra year
among
is
no evidence that longer tenure
they are charged for
new
one just beginning with the firm
(1) for his
(column
The entrenched
options
2),
if
options as their tenure increase.
on "value of options grants" by
we
is
associated with
find that the
.4
and
.8 cent,
present to limit board
more uncharged
more experienced
options. In fact,
CEOs
are the
more
options grants.
the magnitude of this effect,
million dollars.
less for their
In the absence of large
On the contreiry, when large shareholders are
the firms with such active boards,
To gauge
and
of tenure increases the coefficient
depending on the specification.
capture, there
get charge less
5.
there
is
no
let's
compare two CEOs, one with 9 years of tenure and
(0 years of tenure),
CEO
Suppose both are given options grants of one
would be charged 80,000 dollars
large shjireholder present. If there
is
less
according to column
a large shareholder present
he would be charged 40, 000 dollars more. Thus the presence of a large shareholder
costs 120,00 dollars
more
to the entrenched
CEO.
Other Governance Measures and Robustness
4.2.1
As with
the pay for luck test,
(1) to (4) of
we
Table 10 use board
replicate our results with other governance measures.
size.
We
find that larger boards charge
CEOs
Columns
less for the
options
they grant. Consider again a grant of one million dollars worth of options. Relative to a board
with
6 directors,
a board with 10 directors charges
35
its
CEO between 36,000 and
72,000 dollars
less
(columns
(1)
and
In colurans
(4) ).
we examine
(5) to (8),
sign on the interaction term that
in
columns
(5)
and
eflfect
contrary to our
When we
While
it
remains of the wrong sign,
dollcir
magnitude smaller than the
to arise primarily
We
then ask
initial
member
to
expectation and that
for
effects
not statistically significant and
it is
we found
Thus the
earlier.
it is
an insider in a 10 person board reduces
results in
frciction insiders
As
before,
we construct
this index
This
columns
is
(5)
an order of
and
(6)
seem
and time.
by taking the unweighted average of
the standardized values of four governance variables: the number of
of large shareholders in the board, minus boaird size
all
is in fact significant
a composite index of governance explains variation in the financial charge
for options across firms.
In
find a
the other governance measures,
option grant by only four thousand dollars.
from a spurious relationship between
how
on the board. Here we
allow the effect of options grants to vary by year, the effect
economically very small. Converting a board
the charge for a million
of insider presence
However, unlike what we observe
(6).
this result is not robust.
greatly diminishes.
is
the
specifications (columns (9) to (12)),
we
all
shareholders, the
number
and the share of outsiders on the board.^^
find that better general corporate governance
associated with a higher charge for options grants.
The magnitude
standard deviation increase in the governance index (about
2) leads to
is
again moderate.
A
is
one
an extra charge of between
36,000 and 56, 000 dollars for a million dollar grant in options.
Finally,
assuming that the govemcuice
firm size, one might wonder whether
we
Vciriables
we have
isolated in our data are correlated to
are able to explain any of the variation in the charge for
options grants beyond the variation that
is
7 but with an additional control for Size
* Opticniit,
related to firm size.
where
We,
size is
therefore, reestimate equation
measured as
in T^ble
7:
average
'"The variables are standardized by demeaning and dividing by standard deviations. Note the same caveat as
Large shareholders on the board count twice in this measure. As before, if we include only large
shareholders on the board, we find similar results.
before holds.
36
log assets of the firm over the
coefficients of interest at all.
large shareholders
column
The
results are
(8) in
all
Again the
coefficients are
the specifications in Table 11,
we
(3)
and
does not chemge the
in Table 11 for 2 of the governance measures:
Table 8 shows that the coefficient
—.046 and —.031 versus —.030). Columns
in Table 10.
shown
size control
on bocird and the governance index. Comparing column
column
(2) to
sample period. The addition of a
(4)
is
(1) to
column
(6)
and
roughly the same (—.040 versus
can be compared to colimins (10) and (12)
roughly similar and stay highly significant. Note that for
find
some evidence that
large firms charge less for the options
they grant.
To summarize, the
test.^''
CEO
results in this
second
test closely
Firms with large shareholders charge their
tenure seems to decrease the charge for options
but perhaps even increase
for the
We began
CEO
there
is
We
any.
CEOs more
if
for the options they are granted.
there are no large shareholder on the board,
find qualitatively similar results for board size
governance index. Finally, our results do not seem to be simply driven by firm
Summary and
5
it if
mirror our findings in the pay for luck
this
paper in an attempt to "distinguish" between the agency and the skimming view of
they are charged for options. For each of these specific
we found a
significant
amount of pay
this generalizes the findings of the relative
RPE
CEOs
teats,
are rewarded for luck and whether
we learned some
for luck for
lessons. In the
a variety of measures. As noted,
not a specific result due to strategic competitive efiects in the product market.
of
also
found that better governed firms use significantly
The spurious estimates on
fraction insiders
is
pay
performance evaluation Uterature and suggests that the
Icick
is
size.
Conclusion
compensation. To this end, we examined whether
for luck test,
and
the exception.
37
less
pay
for luck
We
than worse governed firms.
Adding a
large shareholder
on the board,
for
example, decreased the pay
This finding weakens two prominent explanations of pay for luck: "Paying
"Filtering out luck
is
dollars
more
is
for luck
by 23 to 33%.
is
optimal" and
too hard."
In the chcirge for options
A CEO who
for luck
test,
we found that better governance increased the charge
for optiona.
given one million dollars in options would be charged between 30 and 50 thousand
for these options if his firm
dependence of the charge
for options
had a
large shareholder present
on the boaxd.
on governance suggests that poorly governed
CEOs
This
are not
being charged enough for their options. Though we do not have direct evidence, the fact that options
do not appear on balance sheets
discussion
may
well be the culprit. Small shareholders, not involved in board
and simply tracking the bottom
line of their
company's balance
sheet,
may
simply not
be aware of new options grants. These findings suggest that a simple policy intervention, forcing
firms to put options on the balance sheets, could result in a curtailing of
More
pay.
some skimming
broadly, though, the results in this paper encourage a revision of our views
In contrast to the view painted in the introduction, the
of pay.
on
CEO
skimming and agency models are
not to be "distinguished" but rather serve as complementaxy views. Well governed firms conform
to the predictions of the agency theory, while poorly governed firms
of the
skimming
theory. In other words, principal agent
individuals around to act as principals.
38
conform to the predictions
models work best when there are in
fact
References
Aggaxwal, Rajesh and Andrew Samwick, "The Other Side of the Trade-Off:
The Impact
of Risk
on Executive Compensation," Journal of Political Economy, 1999a.
Aggarwal, Rajesh and Andrew Samwick, "Executive Compensation, Strategic Competition, and
Relative Performance Evcduation: Theory and Evidence," Journal of Finance, 1999b.
Baker, George, Robert Gibbons and Kevin Murphy, "Subjective Performance Measures in Optimal
Incentive Contracts," Quarterly Journal of Economics, 1994.
Baker, George and Brian HaJl, "Understanding
CEO
EflFort,"
Top Management
Incentives:
Firm
Size,
Risk and
mimeo, Harvard Business School, Harvard University, 1998.
Bertrand, Marianne and Sendhil Mullainathan, "Executive Compensation and Incentives:
Impact of Takeover Legislation,"
NBER Working
Paper no. 6830, 1998.
Blanchard, Olivier, Florencio Lopez-de-Silanes, and Andrei ShJeifer,
Cash Windfalls?," Journal of Financial Economics,
Crystal, Graef, In Search of Excess:
Co.,
New
The
"What Do Firms Do with
1992.
The Overcompensation of American Executives (W.W. Norton
York, N.Y.), 1991.
Diamond, Peter, "Managerial Incentives:
nal of Political Economy, 106
(5),
On
the Near Lineaxity of Optimal Compensation," Jour-
1998.
Garen, John, "Executive Compensation and Principal-Agent Theory," Journal of Political Econ-
omy, 1994.
Gibbons, Robert and Kevin Murphy, "Relative Performance evaluation for Chief Executive
cers," Industrial
Offi-
and Labor Relations Review, 1990.
Gibbons, Robert and Kevin Murphy, "Optimal Incentives in the Presence of Career Concerns:
Theory and Evidence," Journal of
Political
Economy, 1992.
Haid, Michael, Incentive Compensation and the Market for Corporate Control: Substitutive Forces
to Discipline
From
the Oil Industry
Hall, Brian,
12(2),
Hall, Brian
Management
1977
of Publicly Held Organizations in the U.S.: Empirical Evidence
-
1994 (Haupt, Bern), 1997.
"The Design of Multi-Year Stock Option Plans," Journal of Applied Corporate Finance,
Summer
and
1999a.
Jeffrey
Liebman, "Are
CEOs
Really Paid Like Bureaucrats?," Quarterly Journal
of Economics, 1998.
Holmstrom, Bengt, "Moral Hazard and Observability," Bell Journal of Economics, 1979.
Holmstrom, Bengt and Paul Milgrom, "Aggregation and Linearity
in the Provision of Intertemporal
Incentives," Econometrica, 1997.
Hubbard, Glenn and Darius
Palia, "Executive
Pay and Performance: Evidence from the
US Banking
Industry," Journal of Financial Economics, 1994.
Janakiraman, Surya, Richard Lambert and David Larcker, "An Empirical Investigation of the
Relative performance Evaluation Hypothesis," Journal of Accounting Research, 1992.
39
Jensen, Michael and Kevin Murphy, "Performance Pay
of Political
and Top-Management
Incentives," Journal
Economy, 1990.
Joskow, Paul, Nancy Rose and Andrea Shepard, "Regulatory Constraints on
CEO
Compensation,"
Brookings Papers on Economic Activity, 1993.
Milgrom, Paul and John Roberts, "The Efficiency of Equity in Organizational Decision Processes,"
American Economic Review, 1990.
Murphy, Kevin, "Corporate Performance and Managerial Remmieration:
An
Empirical Analysis,"
Journal of Accounting and Economics, 1985.
Murphy, Kevin, "Incentives, Learning, and Compensation:
tion of Managerial
A
Theoretical and Empirical Investiga-
Labor Contracts," Rand Journal of Economics,
1986.
Murphy, Kevin, "Executive Compensation," in O. Ashenfelter and D. Card,
eds.,
Handbook
of
Labor Economics, North Holland, 1999.
Shleifer,
Andrei and Robert Vishny, "Large Shareholders and Corporate Control," Journal of Po-
litical
Economy, 1986.
Tufano, Peter,
"Who Manages
Risk?
An
Empirical Examination of Risk Management Practices in
the Gold Mining Industry," Journal of Finance, 1996.
Yermack, David, "Do Corporations Award Stock Options Effectively?" Journal of Financial Economics, 1995.
Yermack, David, "Higher Market Valuation of Companies with a Small Board of Directors," Journal
of Financial Economics, 1996.
40
Table
1:
Pay
for
(Luck Measure
is
Dependent Variable:
Specification:
Luck
for Oil
CEOs
Crude Oil)
Ln (Total Compensation)"
log Price of
General
Luck
General
Luck
(1)
(2)
(3)
(4)
.82
2.15
—
—
(.16)
(1.04)
—
—
Ace. Rate
of Return
Ln(Sh. Wealth)
Age
.38
.35
(.03)
(.17)
.05
.07
.05
.05
(.02)
(.03)
(02)
(.02)
-.04
-.05
-.04
-.04
(.02)
(.02)
(.02)
(.02)
.01
.01
.01
.01
(.01)
(.01)
(.01)
(.01)
-.03
-.03
-.03
-.03
(.02)
(.01)
(.02)
(.02)
Firm Fixed Effects
Year Quadratic
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Sample Size
Adjusted i?^
827
827
827
827
Age2
100
*
Tenure
100
Tenure=^
.70
.75
"Notes:
1.
Dependent v&ri&ble is the log&ntbm of tot&] compensation. PerFonn&nce measure is tvccountiDg rate of return in columns (1) and
log&rithm of shareholder we&ltb in columns (3) and (4). All nominal variables are expressed in 1977 dollars.
2.
Summary
3.
The
statistics for the
eample of
luck regressioo (colunmB 2
and
4)
oil
firms are available in
In 1977 dollars.
E^ach regression Includes firm fixed efTecis
6.
Standard errors are
in
and a quadratic
in y«u-.
parentbeses.
41
and tbe
Appendix Table Al.
instrument for perfbrmaoce with tbe logarithm of the price of a barrel of crude
4.
(2)
oil in
that year, expresaed
Table
Full
2:
Summary
Yermack
CEO
Statistics
Sample
Mean
Age
CEO
6.84
9.10
8.08
901.69
795.15
6.62
.60
1595.85
3488.32
Ln(Total Compensation)
6.98
.81
Number
of Large Shareholders (All)
1.12
1.42
Number
of Large Shareholders on
.24
.74
13.45
4,54
.42
.19
CEO
Salary and
Bonus
Ln(Salary and Bonus)
Totd Compensation
Board
Size
Freiction of Insiders
on Board
"Notes:
1.
Sample period
All
is
D.
57.42
of
Tenure of
2.
S.
1984-1991.
nominal variables &re expressed in thousands of 1961 dollars.
42
Board
Table
Cash Comp
Dep. Var.:
General
for
Luck"
Ln(Tot Comp)
Ln(Cash)
Luck
General
Spec:
Pay
3:
Luck
Panel A: Luck Measure
Income
.17
.35
(.02)
(•16)
IncoTTie
AssetB
Ln(Shareholder
Wealth)
—
—
Luck
General
is
W
1737
1737
—
.75
—
—
—
—
2.13
2.94
2.36
4.39
(-16)
(1.28)
(-28)
(2.17)
—
—
—
—
1729
.75
1729
—
1722
.58
Panel B: Luck Measure
Income
Income
Asaete
Ln (Shareholder
.21
.34
(.02)
(.10)
—
—
—
—
is
1722
—
Mean
Adjusted
H?
4684
4684
2.18
4.02
2.07
4.00
(.12)
(.53)
(.21)
(.86)
—
—
—
—
4648
4648
.81
.77
4624
.70
Luck
General
Luck
—
—
—
^
~
.22
.32
.31
.57
(.02)
(.13)
(.03)
(.23)
1713
1713
1706
1706
.75
—
.59
—
Industry Performance
Wealth)
Sample Size
General
Ln(Tot Comp)
Exchange Rate Shock
-
Sample Size
Adjusted
Ln(Cash)
4624
—
—
—
—
—
.20
.22
.25
.29
(.01)
(.12)
(.02)
(.19)
4608
4608
4584
4584
.82
—
.71
"Notes:
Dependent variable is the level of SAiBry and boous in columnfl (1) and (2), the logarithm of salary and bonus in columna (3), (4), (7) and
(8) and the logarithm of total compensation in columns (fi), (C), (9) and (10). Performance measure is operating income before extraordinary
items in cotumne (I) and (3) (la roillionB). operating income to total asaets in columns (5) to (6) and the logarithm of shareholder wealth in
columns (7) to (10). All Dominal variables are expressed in real dollars.
In the luck rcgreBsions in Panel
A, the performance measure Is lnBtrument*d with current and lagged appredation and depre<ciation dummies
rate index growth. First-stage regressions are presented in Appendix Table A2.
and current and tagged exchange
In the luck regressions In Panel B, the perfonneknce measure is instrumented with the total assets-weighted average perfbrmance measure in
the firm's 2-digit industry (the firm itself is excluded from the mean calculation).
Each regression includes firm
Standard errors are
fixed effects, year fixed effects
and demographic controls (quadratics
in parentbeoes.
43
in age
and tenure).
Table
4:
Large Shareholders and Pay for Luck
(Luck Measure is Mean Industry Performance)
Dependent Variable: Ln(Total Compensation)"
Large Shareholders
Governance Measure:
Spec.:
Income
Assets
Governance*
Income
AsBets
General
Luck
General
Luck
General
Luck
General
Luck
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
2.18
4.59
2.14
4.49
(.238)
(.912)
(.882)
-.094
-.416
—
(.217)
-.181
-1.48
—
—
(.094)
(.204)
(.176)
(.396)
—
ln(Shareholder Wealth)
—
Governance*
Large Shareholders on Board
—
ln(Shareholder Wealth)
.249
.383
.258
.318
(.018)
(.219)
(.017)
(.199)
.001
-.066
-.019
-.076
(.007)
(.036)
(.016)
(.053)
—
—
-.009
.018
-.017
.411
-.006
.084
.100
.480
(.011)
(.018)
(.049)
(.240)
(.021)
(.033)
(.108)
(.356)
Sample Size
4610
4610
4570
4570
4621
4621
4581
4581
Adj. i?2
.695
Governance
.706
.706
.694
'Notes:
1
DependeDt
variftble is the logarithm of tot&l compenaattoD.
nre expreBsed io re&l dall&rB.
Performance measure
is
oper&ting income to totaJ assets. All DOTniaal v&riablee
tn all the luck regreeaioas, both the performance coe&sure and the interaction of the performance meaaure with the governance measure
are instrumented- The instruments are the asBct-weigbted average performance in the 2-d.ig)t industry and the iDteractions of the industry
performance with that governance measure.
common
"Large SharehoIderB" iodicates the number of blocks of at least
five
a director. "Large Shareholders on Board" indicates the number
of blocks of Bt least five percent of the firm's
percent of the Hrm's
shares, whether the block holder is or is not
common shares that are held by
directors of the board.
Each
regression includes
Standard errors arc
Brm
fixed effects, year fixed effects, a guadrfttic in age
in parentheses.
44
and a quadratic
in tenure.
Table
Tenure, Large Shareholders and Pay for Luck
5:
(Luck Measure
is
Mean
Industry Performance Mean Industry Performance)
Variable: Ln(Total Compensation)"
Dependent
Any Large
No
Spec.
Income
Aseete
CEO
Assets
Ln
Yes
Luck
General
Luck
General
Luck
General
Luck
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
—
—
—
—
2.14
3.35
1.28
2.47
(.30)
(.96)
(.65)
(2.60)
.00
.13
.063
-.006
(.02)
(.05)
(.045)
(.131)
—
—
—
—
(Sh. Wealth)
—
CEO
Tenure*
Ln (Sh. Wealth)
CEO
No
Gen
Tenure*
Income
Shtireholder on the Board?
Yes
Tenure
—
—
—
.24
.26
.27
.53
(.02)
(.24)
(.05)
(.32)
.003
.009
-.005
-.013
(.001)
(.005)
(.003)
(.010)
.01
-.00
.010
.016
-.002
-.045
.044
.084
(.00)
(.01)
(.011)
(.016)
(.01)
(.04)
(.020)
(.059)
Sample Size
3884
3884
740
740
3841
3841
743
743
Adjusted B?
.7030
.757
.715
.700
"Notes:
1.
Dependent van&ble
2.
Id all the luck regreaalooB,
The
is
the tog&ritbm of total compeneatian. All noiniaai variables &r« expreased in real dollarg.
CEO
tenure are instrumentedboth the performaDce measure and the interaction of the performance measure with the
iDBtrutnents are the asoet-weighted average performance in the 3-digit industry and tbo interactions of the industry pcrformaDce with the
tenure.
CEO
3.
in columns (1), (2), (C) and (6) is the fret of firm-yeax obaervatiooB for which there is do large shareholder sitting on the board of
directors; sample in columns (3), (4), (7) and (8) is the set of Brm-year observations For which there is at least one large shareholder sitting
on the board of directora.
Sample
4.
Each regression includes 6rm fixed
6.
Standard errors are
effects, year fixed effects,
in parentheses.
45
a quadratic in ago and a quadratic in tenure.
Corporate Governance and Pay
for Luck
Industry Performance)
Dependent Variable: Ln(Total Compensation)"
Table
6:
(Luck Measure
Mean
Board Size
Governance Measure:
Spec.:
is
General
Luck
General
Luck
General
Luck
General
Luck
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
2.61
5.19
2.27
—
(1.62)
(.453)
(1.24)
-.045
-.093
—
—
2.30
(.558)
—
—
-.482
4.51
(.043)
(.094)
(.853)
(2.69)
fncome
Aeeets
Governance*
Income
AsBets
ln(Sh. Wealth)
—
Governance*
ln(Sh. Wealth)
Governance
Sample Size
Adj. R^
—
.099
(.034)
(.210)
.002
.013
(.002)
(.006)
—
—
—
—
—
—
.241
.241
(.029)
(.215)
.027
.126
(.05)
(.190)
.015
-.013
-.080
.158
-.315
-.066
-.742
(.005)
(.007)
(.016)
(.041)
(.129)
(.271)
(.407)
(1.29)
4624
4624
4584
4584
4624
4624
4584
4584
.706
.695
.695
.706
Govemcince Index
General
Luck
General
Luck
(9)
(10)
(11)
(12)
—
—
Income
Asseta
Governance*
Income
AflBBlS
2.07
4.23
(.210)
(.865)
.007
-.216
(.057)
(.134)
ln(Sh. Wealth)
Governance*
ln(Sh. Wealth)
Governance
Sample Size
Adj.
.216
.
.012
Governance Measure:
Spec.
Fraction Insiders
n?
—
—
.249
.252
(.016)
(.232)
-.003
-.033
(.004)
(.015)
-.016
.000
.010
.210
(.007)
(Oil)
(.027)
(.103)
4610
4610
4551
4551
.695
.705
"Notes:
1.
Dependent variable is the logarithm of total compensation. All Dominal variables are deflated. Each regression includes firm
fixed effects, a quadratic >q age and a quadratic in tenure.
2.
"Board Siae" indicates the number of members of the board of directors, as listed In the proxy statement near the start of the fiscal jrear.
"FVaction Insiders" is the fraction of inside and "grey** directora on the board of directors. "Governance Index" is the unweighted average of 4
standardized governance vajiabtes (number of large abarebolder^, number of large sharebolders on board, minus board siie and one minus fraction
fixed effects, yt&r
insiders).
3.
In all the luck regressions, both the performance measure and the interaction of the peTforxDaiice measure i^'ith the govemaace measure are
iDstnimented. The instruments are the asset- weighted average performance in the 2-digit industry and the interactions of the induBtry performance
with that governance measure.
4.
Standard errors are in parenCbcses.
46
Corporate Governance and Pay for Luck
Robustness Checks
(Luck Measure is Mean Industry Performance)
Table
7:
Dependent
Income
Assets
Governance*
Income
Aeaeta
In(SharehoIder Wealth)
Governance*
Firm Size*Performance?
Sample Size
Adj.
^
Governance Index
Gener£il
Luck
General
Luck
Genercil
Luck
General
Luck
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
.288
36.9
12.3
—
(14.6)
(1.31)
(6.17)
-.118
-2.23
—
—
-.570
(1.14)
—
—
.056
-.197
—
—
(.18)
(.52)
(.061)
(.126)
—
—
—
—
ln(Sharebolder Wealth)
Governance
Ln(TotaI Compensation)"
Large Shareholders
on Board
Governance Measure:
Spec:
Variable:
.223
-.136
(.086)
(.334)
-.018
-.059
(.016)
(.053)
—
—
—
—
.251
-.194
(.098)
(.345)
-.003
-.027
(.004)
(.012)
-.010
.127
.094
.365
-.019
-.002
.007
.170
(.021)
(.038)
(.109)
(.357)
(.007)
(.010)
(.029)
(.085)
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
4621
4621
4581
4581
4610
4610
4551
4551
.695
—
.706
.695
.705
—
"Notes:
1.
Dependent variable ib tbe logarithcn of totaJ compeng&tioc. Performance meftaure is operating income to total aasetB- All nominal variables
are eKpressed in real dollars. Each vegreBaion includcB firm fixed effects, year fixed effects, a quadratic in age and a quadratic in tenure-
2.
In all the luck regressions, both tbe perforniance measure and the interaction of tbe performance measure with the governance measure
are instrumeated. The lostrucaents are the asset -weighted average perfoniLaiice in the 3-digit industry and the interactions of the industry
performaDcc with that goveroance measure.
3.
4.
"Large Shareholders on Board" Indicates tbe number of blocicg of at least five percent of the firm's common ebares that are held by directors
of tbe board. 'Governance Index" is tbe unweighted average of 4 Btandardized governance vanables (number of large shareholders^ number
of large shareholders on boardt minus board size and one minus fraction inBiders). *Firm Size" Indicates average log assets over the sample
period.
Standard errore are
in parenthefies.
47
i
Table
8:
How Much Are CEOs
Charged
for
Options Grants?
The Role of Large Shareholders
Dependent Variable: Cash Compensation"
Large Shareholders
Governance Measure:
Large Shareholders
on Board
All
(2)
(3)
(4)
.055
.055
—
—
(.006)
(.007)
(1)
Value of Options Grants
(5)
(6)
(7)
(8)
.046
.046
—
—
(.003)
(.003)
Value of Options Grants*
-.011
-.011
-.007
-.007
-.050
-.051
-.029
-.030
Governance
(.005)
(.005)
(.006)
(.006)
(.008)
(.008)
(.010)
(.010)
Governance
-20.5
—
-25.0
—
11.9
—
-3.84
—
(13.1)
Sales
Income
ASBCtB
Sh.
Wealth
(billions)
Year F.E.*GovErnance?
Year F.E.*Value of Options Grants?
Sample
Adj.
size
R^
(27.4)
(12.3)
(25.8)
.027
.027
.015
.015
.027
.027
.015
.016
(.006)
(.006)
(.006)
(.006)
(.006)
(.006)
(.006)
(.006)
1195.3
1203.1
1236.9
1246.2
1220.2
1222.4
1251.2
1259.4
(209.7)
(209.8)
(196.9)
(197.0)
(208.3)
(208.6)
(196.6)
(196.8;
.045
.045
.043
.043
.046
.045
.045
.045
(.006)
(.006)
(.006)
(.006)
(.006)
(.006)
(.006)
(.006)
No
No
Yes
No
No
No
No
Yes
Yes
Yes
Yes
No
No
Yes
Yes
Yes
5085
5085
5085
5085
5103
5103
5103
5103
.627
.627
.674
.674
.630
.630
.674
.674
"Notes:
1.
2.
3.
Dependent variable is the level of salary,
e^cpressed in real terms (1991 dollar^)-
AH Dominal
variables are
"Large Shareholders^ indicates the number of blocka oF at least f^ve percent of the firm's common shares, whether the block holder is or is not
a director; *'Large Shareholders on Board" indicates the number of blocks of at least five precent of the firm's common shares that are held by
directors of the board.
Each regression includes firm
CEO
4.
bonus and other compenBatioDS txcluding the value of options granted.
fixed effects
and year fixed
effects.
tenure.
Standard errors are
in.
parentheses.
48
Demographic controls include a quadratic
in
CBO
age and a quadratic in
Table
^
9:
Tenure, Large Shareholders and Charge for Options
Dependent Variable: Caah Compensation"
Any Large
Value of Options Grants
Value of Options Grants*
Tenure
Shareholder on the Board?
No
Yes
No
Yea
No
Yes
No
Yea
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
-.025
.019
-.025
.014
(.004)
(.015)
(.004)
(.015)
.008
-.004
.008
-.003
.004
-.006
.004
-.006
(.0004)
(.001)
(.0004)
(.002)
(.0008)
(.002)
(.0008)
(.002)
34.3
40.6
(8.2)
(14.7)
Performance Controls?
Year F.E.*Tenure?
Year F.E.*Value of Options?
Yes
Yes
No
No
No
No
Sample
4257
860
.674
.676
Tenure
size
Adj. i?2
.
29.6
47.7
(7.5)
(14.2)
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
No
No
No
No
Yes
Yes
Yes
Yes
Yes
4257
860
4257
860
4257
850
.674
.674
.682
.701
.682
.699
Yes
'
"Notes:
1.
Dependent v&ri&ble
expressed
2.
3.
in real
is
the level of sbl&ry, bonus and other compeDaations deluding the value of options granted.
All
nominal variables are
terme (1991 dollars).
E^h regrcaaion includes firm fiied effects and year fixed effects. Performance controls are
shareholder wealth. Demographic controls include a quadratic in CEO age and a quadratic in
Standard error? are
in parentheses.
49
sales,
CEO
income
tenure.
to aasets ratio
and the value of
Table 10:
How Much Are CEOs
Charged for Options Grants?
Governance
Variables
The Role of Other
Dependent Variable: Cash Compensation"
(1)
(2)
(3)
(4)
(5)
Board Size
Governance Measure:
Value of Options Grants
Value of Options Grants*
Governax»ce
-.214
-.214
(.012)
(.012)
Fraction Insiders
.
.226
.225
(.016)
(.017)
.018
.009
.009
-.27
-.27
-.04
-.04
(.0008)
(.001)
(.001)
(.02)
(.02)
(.03)
(.03)
6.24
—
283.7
—
190.6
—
Performance Controls?
Year F.E.*Governance?
Year F.E.*VaIue of Options Grants?
Yes
.
Sample
(157.6)
(5.67)
Yes
Yes
No
Yes
No
Yes
Yes
5104
5104
5104
5104
.636
.636
.673
.673
Yes
Yes
5104
5104
5104
.665
.678
.677
(10)
(11)
(12)
Yes
No
5104
.665
(9)
No
(150.0)
Yes
Yes
Yes
Yes
Yes
Yes
Yes
No
No
size
(8)
.018
(5.78)
Adj. R'^
(7)
.0007)
4.04
Governance
(6)
No
No
Governance Index
Governance Measure:
Value of Options Grants
-.008
-.008
(.005)
(.005)
—
—
Value of Options Grants*
-.028
-.028
-.018
-.019
Governance
(.002)
(.002)
(.003)
(.003)
Governance
-11.7
-16.16
(9.4)
(9.9)
Yes
Yes
Yes
No
Yes
Yes
Yes
Yes
5084
5084
5084
5084
.637
.639
.676
.677
Performance Controls?
Year F.E.*Governance?
Year F.E.*Value of Options Grants?
Yes
Sample
size
Adj. R"^
No
No
No
"Notes:
1.
Dependent v&ri*ble is the level of SftJ&ry, bo&ua and otbpr coroptfnSAtJonB excluding the v&lue of optiooB gruited.
expr«09ed in r«sJ terms (1991 dolLaiB).
All DomiDaJ ^'sxiables axe
2.
"Board Site' indic&tcB the number of memberB of the board of directors, &8 lieted in the proxy statement Dcaj the atart of the fiscal year.
"FVaction iDsldere" is the fraction of inside and "grey" directors on the board of directors. "Hjoveraauce Index" is the unweighted average of
4 standardized governance vaxiables (number of large sharebotders, number of large sbarebolders on boaxd, minus board size and one micus
insiders presence). Performance controls are sales, income to assets ratio and tbe value of shareholder wealth. Demographic controls include a
quadratic in
age and a quadratic in
tenure.
3.
Each regression includes firm
4.
Standard errors are
CEO
CBO
fixed efTecta
and year
fixed effects.
la parentheses.
50
T^ble 11:
How Much
Are CEOs Charged for Options Grants?
Robustness Checks
Dependent Variable: Cash Compensation"
Large Shareholders
on Bo^l^d
Governance Measure:
Value of Options Grants
Governance
Index
(1)
(2)
(3)
(4)
-.279
—
-.181
—
(.051)
(.052)
Value of Options Grants*
-.040
-.031
-.024
-.017
Governance
(.008)
(.011)
(.002)
(.003)
5.59
-4.33
-13.7
-17.1
(27.2)
(25.7)
(9.9)
(9.4)
Governance
Value of Options Grafts*
.039
.022
.021
.011
(.006)
(.006)
(.006)
(.006)
Performance Controls?
Year F.E.*Value of Options Grants?
Yes
Yes
Yes
No
Yes
No
Yes
Yea
Sample
5103
5103
5084
5084
.633
.6
.638
.680
Size
size
Adj. fi2
"Notes:
1.
2.
DepeDdent variable
ifl the level of salary,
bonus frnd other compensfttions excluding tbe value of optioDs granted.
expressed in real terms (1001 dollars)-
All
nominal variables are
"Large Sbareholdere on Board'* indicates the number of blocks of at least five precent of tbe firm's comtnon sbares that are held by directors
of the board. "Governance Index' ie the unweighted average of 4 standardifted governance variables (number of la«Tge Bbareholders, number
of large shareholders on board, minuB board siae and one minus fraction insiders on board). **5)se** is time average of log real total assets for
each Brm.
3.
Each regression includes firm fixed
effects
and year
fixed effects.
shareholder wealth. Demographic controls include a quadratic in
4.
Standard errors are
in
parentheses.
51
Performance cootrolo are
age and ft quadratic in
CEO
sales,
CEO
income to assets ratio and tbe value of
tenure.
Appendix Table Al: Summary Statistics
50 Largest US Oil Companies CEOs "
Mean
Age
of
CEO
Tenure of
CEO
Total Compensation
Ln(Total Compensation)
S.
D.
58.562
7.892
10.181
9.781
608.269
697.194
6.125
.722
° Notes:
1.
2.
Data
set ia
60 of the 61 largest
US
oil
companiefl over the period 1977-1994.
Total Compensation is defined as the Bum of sai&iy and bonus (cub &nd stock bonus), company con<
tributions to thrift plans, other annu&l income and the value of the options granted to the CEO during
that year, in thousanda of 1977 dollars.
52
Appendix Table A2: Pay for Luck: First-Stage Regressions
(Luck Measure is Exchange Rate Shocks)"
Income
Dep. Var.
Inc. to
Ln(Sh. Wealth)
Assets
(1)
2% <
<4%
(2)
(3)
-56.588
-.006
-.039
(26.408)
(.004)
(.047)
-15.428
.004
-.027
(24.271)
(.004)
(.048)
-68.903
-.013
-.034
(Current)
(32.039)
(.005)
(.058)
Appr. > 4%
(Lagged)
-12.045
.006
.053
(30.646)
(.005)
(.055)
76.642
-.000
.153
(24.647)
(.004)
(.045)
85.858
.010
.114
(Lagged)
(25.942)
(.004)
(.047)
Depr. > 4%
(Current)
45.482
.007
.094
(27.761)
(.005)
(.050)
Appr.
(Current)
2% <
Appr.
<4%
(Lagged)
4%
Appr. >
2% <
Depr.
<4%
(Current)
2% <
Depr.
<4%
> 4%
Depr.
(Lagged)
Exch. Rate Index Growth
(Current)
Exch. Rate Index Growth
76.345
.017
.046
(29.791)
(.005)
(.054)
-19.273
-.000
-.077
(167.134)
(.030)
(.302)
216.140
.038
.237
(175.302)
(.031)
(.316)
Sample Size
1737
1729
1713
Adjusted
.622
.700
.873
(Lagged)
R"^
F-stat
(jxrob
>
F - etat)
3.48
2.6
2.47
(.000)
(.004)
(.006)
"Notes:
1.
Depeodeot v&riable is the level of income in column (I), tbe rfttio of operating income to totcJ BAdeta
column (2) and the log v&lue of Bh&reholder v^eAltb in column (3). Income and ehareholder ft-e&lth
are eirpressed to millioos of 1977 dollars. 2% < Aypr. < 4% is dummy variable that equals 1 if the
industry-specific exdiange rate index appreciated by more tban 3% and less than 4% since the previous
year. All the other appreciation and depreci&tion dummies are defined in a similar way.
in
2.
Each regression includes firm fixed effects axid year
CBO ftge and a quadratic in CEO tenure.
fixed effects. All regressioDB also include
a quadratic
In
3.
4.
The 3
regreaaions are the first-stage regressione associated with columns (3), (4), and (8) in Panel
Table
3.
Standard errors are in parentheses.
53
A
of
Appendix Table A3: Mean Ratios of Options Grants to Total
Compensation for High and Low Governance °
D
Governance Category
-lean
No
.16
.20
Lcirge Shareholder
.18
.22
Shareholders on Board
.18
.22
At Least One Large Shareholder on Board
.14
.22
Large Shareholders
At Least One
No Large
S.
Board
Size
Above Median
.18
.21
Board
Size
Below Medlam
.17
.22
Above Median
.17
.22
Fraction Insiders Below Median
.18
.21
Governance Index Above Median
.17
.22
Governance Index Below Median
.17
.21
Fraction Lisiders
"Notes:
1
"Large Shareholders" indicates the number of blocks of &t leaat 6ve percent of the Snn'e commoc sfa&res, whether the block holder is or is not
a director. "Large Shareholders oa Board^ iodicates the number of blocks of at least five precent of the firm's comincn shares that are held
by directors of the board. "FVaction Insider** is the fraction of inside and "grey" directors on the board of directors. "Governance Index" is
the unweighted average of 4 standardized governance variables (number of large shareholders, number of large shareholders on board, minus
board sise and one minus insider presence on board).
54
1977 Dollars per Barrel
I
I
I
I
I
I
I
L
21.0349
o
o
ID
O
CO
ir
5.47517-
T
I
I
I
1
I
I
~T
I
I
1
I
i
I
r
I
77 78 79 60 81 82 83 84 85 86 87 88 89 90 91 92 93 94
year
FIGURE
8.08356
1
:
Real Price of a Barrel of Crude
Oil
-
c
I.
3
9
D>
C
c
3
§
<
O
o
IS
2.75434
'
"T
I
I
I
I
I
I
I
r—
—
1
1
i
1
1
1
1
r
77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94
FIGUREJ^: Mean Rate
year
of
Accounting Return
s xaT a
[
Change
In
Total
Comp.
Change
In Oil
Price
.416984
C
S
o
Q.
.666937 1977
1993
1983
1991
1979
1981
1985
1987
1989
1994
1980
1990
1992
1978
1982
1984
1986
1988
Pay and Crude Oil Price
FIGURE^.: Oil Industry
CEO
sTaTa
531k
Date Due^^^
''""^
Lib-26-67
MIT LIBRARIES
3 9080 02237 2673
!
fiiffl
mm
#Wl'i;
Mm
I'll
M
'
m
^
„
(lif
'
llKaiJiWiifP
"
v
Vi.,; „"'!'<.
!'''!
I
'('''in
I
iikii
i!i
'mm
,(!,]!'
t
T
fi
'^
•ilk'
m
•'
'
pl^Bf
"li''
l"l!,
yMim
Aliliii:
'.III
6:/i!:i!i
i''|i'i
f
,
^iM'
IV r
'(I