,
:;ii
if'!lJi:;:''&&:!:;:il;s!i:
!5ii;!!
i^m. '^T".^'^v,-a.ik
<
r
*
'*o^T1>C*
HD28
.M414
51
DEWEY
i\o
n
ALFRED
P.
WORKING PAPER
SLOAN SCHOOL OF MANAGEMENT
Firm Diversification and
CEO Compensation:
IVIanagerial Ability or Executive
Nancy
Entrenchment?
L. Rose, Center for Advanced Study
Behavioral Sciences, MIT and NBER
Andrea Shepard, Stanford and NBER
April
1994
MASSACHUSETTS
INSTITUTE OF TECHNOLOGY
50 MEMORIAL DRIVE
CAMBRIDGE, MASSACHUSETTS 02139
in
the
Firm Diversification and CEO Compensation:
Managerial Ability or Executive Entrenchment?
Nancy
L. Rose, Center for Advanced Study
Behavioral Sciences, l\/IIT and NBER
Andrea Shepard, Stanford and NBER
April
1994
Working Paper No. 3684-94
in
the
'^M.I.T.
ImRAhiES
3
Firm Diversification and
CEO Compensation:
Managerial Ability or Executive Entrenchment?
Nancy
L.
Rose
Center for Advanced Study in the Behavioral Sciences,
MIT and NBER
Andrea Shepard
Stanford University and NBER
April
1994
Data for a sample of 558 CEOs over 1985-1990 suggest substantial
compensation premia for managers of diversified firms. The CEO of a firm with two
distinct lines of business averages 1
to 12 percent more in salary and bonus and 1
to 17 percent more in total compensation than the CEO of a similar-sized but
undiversified firm, all else equal. This corresponds to average 1990 salary gains of
$115,000 to $145,000 per year for our sample. Diversification may raise pay
because the CEO's job requires higher ability or because it is associated with CEO
entrenchment. If ability explains the correlation, we would expect the diversification
premium to be invariant to tenure. Entrenchment models suggest higher premia for
more experienced (more entrenched) CEOs, and an increase in compensation when
the CEO diversifies the firm. The data support an ability model over an entrenchment
explanation. The diversification premium is unaffected by tenure, and increasing
diversification reduces compensation for incumbent CEOs, all else equal.
Abstract.
We
are grateful to the National Science Foundation, the MIT Center for Energy and Environmental Policy
Research, and the Alfred P. Sloan Foundation for financial support of this project.
This research was
undertaken while Rose was a fellow at the Center for Advanced Study in the Behavioral Sciences. We
thank Severin Borenstein, Ben Hermalin, Paul Joskow, Richard Lambert, Edward Lazear, James Poterba,
Garth Saloner and Mark Wolfson, and participants in the Stanford Brown Bag Lunch and the NBER Industrial
Organization program for helpful discussions, and Justin Adams for data assistance.
I.
INTRODUCTION
The economics and
management
strategic
literatures of the past
decade have grown
increasingly skeptical about the value of combining several lines of business within a
The
single firm.
positive literature has noted the negative response of the stock
markets to diversifying mergers
and the
relatively
The normative
literature
effort to
in
the
1
970s
to a
diversified
understand
(e.g.,
Lang and
1
990)
Stuiz, 1993).
more tempered discussion of how corporate
its
parts
why
diversification failed to
information on
produce the performance gains
1960s conglomerate merger wave has produced two sharply
enterprises,
connection between
find
Morck, Shieifer and Vishny,
whole more valuable than the sum of
divergent classes of explanations.
We
(e.g.,
1985).
anticipated during the
diversified
980s
1
has shifted from an enthusiasm for portfolio management
managers might make the
The
the
poor performance of diversified firms
within diversified firms
(e.g., Porter,
in
the
One focuses on
other on
managerial entrenchment.
CEO compensation and
558 CEOs
in
418
the difficulty of managing
firms over
firm diversification.
1985-1990
Both imply a
This study uses
to explore this connection.
evidence of significant compensation premia for managers of diversified
We
enterprises.
use
patterns
in
these
premia to
explanations from those based on managerial
some type
is
undertaken
if
it
entrenchment
ability.
Ability-based explanations focus on the difficulty of
Diversification
distinguish
managing a
diversified firm.
increases the expected value of the firm through
of synergy or spillover across the lines of business. Realizing the potential
value of these synergies depends on the ability of top management.
If
diversification
increases the value the firm places on managerial talent, more diversified firms
hire
more able CEOs. Because
talent,
CEOs
ability
at diversified firms will
counterparts at less diversified firms.
premium
will
be invariant to
earns a premium
in
the market for corporate
be more highly compensated than are
Ability
will
models imply that the
their
diversification
CEO tenure, and that incumbent CEOs who diversify their
2
firms will be paid less than
CEOs who
matched
are optimally
to the firm's
new, higher
level of diversification.
Proponents of managerial entrenchment explanations argue that diversification
frequently
is
undertaken by self-serving managers to increase the value of
compensation packages, even when
One simple
story
is
their
diversification reduces the value of the firm.
that diversification
is
way
an easy
to increase firm size
when
antitrust constraints restrict acquisitions within a firm's existing lines of business.
Boards of Directors reward CEOs for firm
diversify
even when
does not contribute to shareholder wealth.
it
argument, and the one on which
scope of the firm to match
we
focus here,
their particular talents.
good match between top executives and the
be captured by the
CEO
compensation increases
correlation
for
diversification premia for
ability
variables
known
These
which are dominated by
statistically significant diversification
diversify their firms, a positive
This
in
is
confirmed
and smaller
in
CEO
in total
and other
our sample of
the position for more than
interfirm variation,
premia
in
two
480
years.
suggest economically and
compensation. Firms with two
on average, 10 to 12 percent more
Compensation increases
into additional business
These models predict
explanations imply a positive cross-sectional correlation
CEOs- those who have been
firms.
may
than for incumbent CEOs.
to affect compensation.
and 13 to 18 percent more
change the
firm diversification, holding constant firm scale
distinct lines of business pay,
subtle
creates managerial rents that
diversification in the cross-section,
"experienced"
results,
it
A more
diversification creates a uniquely
who
incumbent CEOs
new CEOs
between compensation and
firm,
If
incentive to
that top executives
is
and Vishny, 1989).
(Shieifer
between compensation and
Both entrenchment and
CEOs may have an
size,
If
compensation than do
in
salary and
bonus
similar but undiversified
further, but in smaller increments, with diversification
segments.
3
The power to distinguish between
and entrenchment explanations comes from
ability
examining diversification premia for
new CEOs and from
specifications to isolate the response of compensation to
initiated
of
172
by incumbent CEOs.
first-year
CEOs.
The
increase
CEOs
in
premium
diversification
for our
sample
roughly equal to that earned by the panel of experienced
is
compensation.
all
changes
We find that the diversification
first-difference results
diversification,
estimating first-difference
suggest that increasing diversification
Indeed,
else equal.
suggest a
data
the
premium
These patterns provide strong support
for
fails
to
reducing
for ability-based
explanations over managerial entrenchment explanations of diversification premia.
The paper
organized as follows:
is
In
section
III.
II.
discuss the
CEO compensation and
explanations of the connection between
The data and
we
II,
two competing
firm diversification.
empirical specification used to test these views are described
Results are presented
DIVERSIFICATION
in
in
section
section IV, followed by a brief conclusion.
AND EXECUTIVE COMPENSATION
There are two general views of the relationship between diversification and
compensation. The
first is
CEO
align the interests of the
compensation
will
reflect
diversification affects
own
may
compensation, or
based on
this
it
is
this,
The opposing view
diversification
CEOs have
If
because diversification changes the value
is
general perspective as managerial
that
CEOs have
considerable latitude
These
objectives at the expense of shareholders' interests.
include
if
Given
the CEO's contribution to the value of the firm.
compensation,
input or "ability" models.
objectives
with those of the firm's stakeholders.
We refer to models taking this
of managerial input.
to pursue their
predicated on a belief that market mechanisms effectively
if
it
enables
a taste for diversification
view as "entrenchment" models.
CEOs
itself.
We discuss
to
We
increase
refer to
their
models
below the implications of
these competing explanations for the relationship between executive compensation
and firm diversification.
A common
CEO
and
Diversification
Ability
justification for diversification
is
potential synergies or spillovers across
Broadly interpreted, these include spillovers from
lines of business.
R&D
activities,
synergies from capabilities applicable to multiple lines of business, economies of scope
from slack resource constraints or amelioration of market power
downstream markets through vertical
integration.
These
by an increased load on limited managerial inputs.
CEO
managerial input at the
may
Diversification
Hambrick,
1
level is likely to
upstream or
potential gains
may be offset
In particular,
in
it
the
demand
(Finkelstein
more than one industry requires the CEO to understand
A CEO
of a firm
in
Managing diverse
and
different
lines of
industry
business
structures,
and
different
may require deploying a broad
several,
multiple lines of
business must evaluate competitive strategies for product lines that
customers,
and
increases the complexity of the resource allocation
potentially quite disparate, product markets.
different
for
increase with firm diversification.
increase the complexity of the CEO's job
989); at a minimum,
decision. Operating
in
may have
competitors.
variety of resources
capabilities. Finally, realizing potential synergies involves facilitating coordination
and communication across business groups within the
firm. All
these factors suggest
that the firm's marginal return to executive talent will increase with diversification.
In
an
market for managerial
efficient
compensation
at
more
compensation can be
talent, this higher return will lead to higher
Equilibrium effects of diversification on
diversified firms.
illustrated in a simple
higher ability are matched with positions
higher (Rosen 1982,
productivity of
all
managers make a
ability realized
Waldman
in
In
workers below them
which the marginal return to
is
ability
is
these models, managers contribute to the
in
the firm's hierarchy, and more able
larger contribution than less able
by the firm
of higher ability
1984).
in
matching model, where managers with
managers. The marginal return to
the increment to productivity caused by hiring a person
a given position
and
is
a characteristic of the firm. Characteristics
5
that increase the marginal return to ability
effects:
in
the ability of those selected to
fill
connpensation associated with the position
This
argument
and
diversification
index
in
for
CEO
positive
a
will
is
equilibriunn
be greater, and the
will
be higher.
correlation,
between
paribus,
ceteris
firm
compensation, and a positive coefficient on a diversification
Since the
ability of
the
CEO
in
not a function of tenure, the predicted diversification premia are the
new and
experienced CEOs.^
Matching models provide
diversification
have two
will
the position
cross-sectional compensation regressions.
these models
same
implies
the position
little
guidance on the expected impact of changes
on the compensation of incumbent CEOs.
costless, these models imply that
replace the current
CEO
with one
scope. Because
CEO
to retaining the
incumbent
turnover
suggests that changes
in
changes
is
CEO
whose
in
If
CEO
ability optimally
matches the
as long as the mis-match
may
replacement were
diversification should lead the firm to
costly to both the firm and the
diversification
is
lead to bargaining
who have
firm's
CEO, there
new
is
value
not too severe.
This
between the CEO and
As
the firm, with both parties sharing the costs of the mis-match.
compensation of incumbent CEOs
in
a result,
the
increased their firm's diversification
should be less than the compensation that would be earned by a new, optimally-
matched CEO.
their firm's
Similarly, the
compensation earned by incumbents
compensation should be higher than that of the
matched CEO. This suggests that the
and changes
in
compensation
between compensation and
for
correlation
incumbent CEOs
diversification
who have
less able, but optimally
between changes
will
observed
reduced
in
diversification
be smaller than the correlation
in
the levels regressions.
Compensation empirically tends to increase slightly with tenure in the position,
which may reflect human capital accumulation. Unless the ability to manage a
'
diversified firm increases differentially with tenure, this should not affect the
diversification
premium.
6
Diversification
An
and Management Entrenchment
between
alternative explanation for the relation
compensation
is
A
managerial entrenchment.
popular press accounts of top
possible misalignments
management
growing academic
for aligning incentives
CEO
literature as well as
large public corporations focuses
in
between shareholders'
The primary mechanisms
firm diversification and
interests
--
on
and the CEO's objectives.
pay-for-performance, monitoring by
shareholders or their representatives, and the market for corporate control
--
might not
be sufficiently powerful or precise to prevent self-serving decisions that are good for
managers but not
insufficient
for
owners. Executive compensation schemes,
(e.g.,
).
The market
mismanagement,
incurred
argued, provide
Jensen and Murphy, 1990). Boards of Directors may have
inadequate information and incentives to monitor and control
971
is
performance incentives to make shareholder wealth maximization the top
executive priority
1
it
in its
which may
for corporate control,
is
at best an imperfect
CEO
behavior (Mace,
discipline particularly egregious
mechanism given the transactions costs
use.
Absent effective controls on executive behavior, CEOs may pursue a variety of
objectives (including, perhaps, increased diversification), regardless of whether they
are consistent with shareholders' objectives.
decisions,
there
are
a
compensation might be
change the
of
linked.
Some
The
and
first is
match by
risk are held
a variant
on the
which
risk.
There
diversification
directly affect
direct influence of diversification
arguments that imply a relationship between
size
enforce self-serving
of these are indirect:
firm's size or risk profile, both of
the presence of controls for scale and
when
CEOs can
channels through which diversification
variety
These explanations suggest no
If
are,
and
may
compensation.
on compensation
in
however, two entrenchment
diversification
and compensation even
constant.
ability
diversifying the firm.
matching model
Shieifer
in
which the CEO creates a good
and Vishny (1989) argue that CEOs may use
7
diversification to increase their value to the firm
if
create an industry mix for which his/her managerial
increasing the value to the firm of his/her specific
negotiate a higher
their firms
of
much
smaller for
manage the
This
will
and compensation
diversification
could have diversified their firms)
qualified to
first
between
link.^
in
capital, the
if
tend
for
difference specification, changes
to
CEO
CEOs who
this is the
is
able to
diversify
dominant source
generate
experienced
positive
CEOs
(those
the levels equations. Compensation premia
new CEOs, who by assumption
firm than are the
to
are uniquely well-suited. By
human
should be rewarded with higher compensation
correlations
should be
skills
the Board of Directors.^ Incumbent
diversification-compensation
a
who
wage with
CEO
diversification enables the
CEOs who
are discretely less well-
created the diversification.
in diversification
In a
and compensation should be
correlated.
A second argument
for a linkage
managers have a taste
who
are
for both
more powerful
between compensation and
more
diversification
vis-d-vis their
diversification
is
that
and higher compensation.* CEOs
Boards are better able to implement
their
preferences, leading to both greater diversification and higher compensation at these
firms.
This
implies
diversification,
a
positive
cross-sectional
correlation
and a positive correlation between changes
in
between pay and
diversification
and
^ The compensation increase will depend on the difference between the value to the firm of the
incumbent CEO and the value of the next best qualified manager. This can be thought of as the
rent value of the incumbent CEO's unique talents. Note that this argument implicitly assumes that
the CEO cannot seek employment with a firm already active in the areas to which she/he is
uniquely well-suited and that it is easier for the CEO to affect the scope of the firm than the level of
CEO compensation.
^ If the rent value of a CEO's human capital were maximized by specializing the firm, this model
suggests that entrenched managers would reduce diversification. In this case, compensation would
be negatively correlated with diversification in both cross-sections and first-differences.
*
The compensation preference seems
CEOs may
like diversification because it is a
GE's 1976 acquisition of a billion dollar
mining company, for example, was attributed by the business press to a desire by the CEO to make
a 'lasting imprint on his firm" (Fortune, August 1977).
Roll (1986) argues that acquisition
decisions in general may be the result of CEO 'hubris', a belief that they are uniquely well qualified
relatively
easy
way
to evaluate and/or
clear.
to affect the direction of the firm.
manage the
target firm.
8
changes
in
compensation,
entrenchment.
Entrenchment
more opportunity
a constituency
both
caused
is lil<ely
by
unobserved
variations
because the CEO has
to increase with tenure
to influence the composition of the Board of Directors
among managers within
CEO
in
the firm. This suggests that
and to
build
new CEOs should
average smaller estimated diversification premia than those of experienced CEOs.^
III.
AN EMPIRICAL MODEL OF COMPENSATION
We
explore the effects of diversification on
measure of
(see
CEO compensation by embedding
diversification in a standard empirical
Joskow, Rose, and Shepard, 1993).
a
model of executive compensation
Compensation
is
a function of firm
characteristics such as size, financial performance, and the extent of diversification;
CEO
characteristics such as age and tenure; and industry
shifts in
norms and economy-wide
managerial compensation, captured by industry and year fixed effects. The
data and specification of the compensation equation are described below.
Data
This study combines information on
Forbes' annual
CEO compensation
CEO compensation and
characteristics from
survey with information on firm characteristics
from Standard and Poor's Annual and Industry Segment
Center for Research on Security Prices (CRSP) returns
financial services or regulated industries, leaving
match across our four data sources.^
COMPUSTAT
files.
We
files
and the
delete firms
1830 CEO-years over 1985-90
The analysis
of "experienced"
^
Sample selection biases also may induce a positive correlation between tenure and
unobserved entrenchment: as, for example, if more entrenched managers are able to stay
CEO position longer than the average CEO.
that
CEOs
in
in
is
the
' Firm deletion is based on their primary industry assignment, as recorded by COMPUSTAT.
Regulated industries during our sample period include electric or gas utilities, gas pipelines, and
telephone service. Regulated and financial services industries are omitted due to noncomparability
in firm data and differences in the structure and level of compensation across these and the
remaining industries in our sample (Joskow, Rose, and Shepard, 1993). The short sample period is
due to the restricted span (7 years maximum) of COMPUSTAT's Industry Segment Hie.
9
restricted to those with
more than two years of tenure
yields a panel data set of
1
of
new CEOs
Table
1
through
is
reports
3;
505 observations on 480 CEOs
based on 172 CEOs
summary
summary
in
whom we
observe
statistics for the first-year
403
CEO
This
position.
firms.
in their first
experienced
statistics for the
in
CEO
the
Our analysis
year of office.
panel
in
sample are reported
in
columns
column
1
4.
Variable descriptions follow below.
Compensation Measures.
salary and
We
use two measures of
bonus (SALARY), whether current or deferred.
compensation
is
relatively well-defined
inclusive
This
first is
component
of
and consistently measured across firms and
The second,
over time for our sample period.
most
CEO compensation. The
total
compensation (TOTAL),
measure of compensation reported by Forbes.
It
is
the
includes benefits,
long-term and contingent compensation, and net gains from the exercise of stock
options, stock appreciation rights and stock accrual rights (collectively referred to
TOTAL mismeasures
below as "options").
First,
it
understates current compensation
current compensation for
in
years that
options grants; the ex ante value of options
Second,
it
overstates current compensation
the entire ex post gain
measure of
The mean
$1.9
total
real
Table
million (see
slightly
years that options are exercised,
in
It is,
however, the best
our data set.'
in
our sample
1,
column
1).^
is
$1.1 million;
CEOs
mean TOTAL compensation
at undiversified firms
(column
is
2) earn
(column
3),
although the differences are not statistically significant.
Forbes does not provide sufficient information on options awards to support a more
if one could solve the valuation problem for unexercised options.
sophisticated treatment, even
"
when
lower average salaries and total compensation than their counterparts at
diversified firms
'
compensation
receive substantial
not recorded as compensation.
attributed to current compensation.
is
SALARY
in
is
CEOs
two reasons.
Dollar
amounts
are
all
inflated to
1
990 constant
dollars using the implicit
GNP
deflator.
10
Firm Characteristics.
We
identify three firm characteristics as potentially important
determinants of executive compensation levels: diversification,
size,
and
financial
performance.
Diversification. Data
on a
firm's major business activities are reported in the business
segment section of the Securities and Exchange Commission 10-K reports beginning
in
The 1976
1977.
Financial Accounting Standards Board standard for reporting
these data allows firms considerable latitude
in
defining business segments:
[N]o single set of characteristics is universally applicable in determining
the industry segments of all enterprises, nor is any single characteristic
Consequently,
determination of an
enterprise's industry segments must depend to a considerable extent of
the judgement of the management of the enterprise. (SFAS No. 14,
determinative
in
all
cases.
Paragraph 12)
While the standard rejects
common classification schemes as inadequate,
encourages identifying those businesses operating
Once segments
segments.
data for
all
therefore,
is
at least
required to report data on
language
product markets as
companies are required
are defined,
segments contributing
in distinct
its
to report
segment
10 percent of the firm's revenue. No
firm,
more than ten segments.
We obtain business segment data from COMPUSTAT's Industry Segment Files, which
record seven years of segment data for each covered company.
a primary 4-digit SIC code to each reported
reported
segments.
We
use
these
COMPUSTAT assigns
segment and records revenues
data
to
construct
two
indices
for
most
of
firm
11
similar results for both
The simplest
we
While they have different characteristics,
diversification.^
measures.
diversification index
segments reported
for
obtain qualitatively
is
a count of the
number
of unique 4-digit SIC
each year (NUMSEG). This index weights
Distributions of firms and observations over
NUMSEG
all
segments
are described
in
code
equally.
Table
2.
Approximately one-third of our observations are for firms reporting only one segment,
and multi-segment firms report 3.4 segments on average.'"
A more complex
diversification index,
segments.
distribution of
DIVERSE
is
DIVERSE, incorporates information on the
defined as
NUIklSEG
DIVERSE
where segment sales are the
sales
the
is
sum
segment
of
=
1
-
52
1
segment
sales,
company
sales
sales reported for the business segment,
sales for the firm.
This measure, which
and company
is
one minus
Herfindahl index for the firm's business segments, increases (nonlinearly)
number
of segments, holding constant the variance of
segment
the variance of segment sales shares, holding constant the
^
We
size
size,
number
in
a
the
and declines
in
of segments."
measure of diversification, which ranked as more diverse
segments that were 'distant' from each other, similar in spirit to Gollop
measured distance between a pair of industry segments by the
also experimented with a third
firms operating in industry
and Monahan (1991).
We
probability of observing the given pair of industries within a single firm, using the universe of
business segment data to construct probabilities.
small, this
'"
a
measure had
The number
mean
change
of 2.47
in
in
little
power
to distinguish
Because most of the
among
probabilities
were quite
diversified firms.
segments declines slightly over time, from a mean of 2.79 in 1985 to
1990. This decline may reflect both a decline in firm diversification and a
of reported
reporting behavior (see Lichtenberg, 1991).
Similar measures have been used in a few earlier studies of diversification; see Ravenscraft
and Scherer (1987), Gollop and Monahan (1991), and Comment and Jarrell (1992).
^^
12
Thus, a firm with two equal-sized segments
is
ranked as more diversified than a firm
with two unequal segments. As a result, this index can be sensitive to the fact that
a firm with
two
equal-sized segments
which the
two-segment
firm in
construction,
DIVERSE has
upper bound of 0.90
segment).
a lower
is
0.48.
distribution of business
require different managerial inputs than a
segments represents 90 percent of
bound of zero
(for single
segment
for
DIVERSE
Because
it
segments, our
is
is
0.32, and the
mean
firms)
initial
analysis focuses on
and an
in
each
for firms with
number and
sensitive to both the
By
sales.
ten-segment firm with 10 percent of sales
(for a
The sample mean
segments
multiple
larger
may
size
DIVERSE.
Firm Size. Measures of firm scale are included to account for the well-documented
relationship
total
between compensation and
company revenues (SALES); we
firm size.^^
Most studies measure scale by
also consider controlling for the
number
of
employees (EMPLOY). These two measures are highly correlated and, as the data
reported
in
Table
1
show, the firms
Annual revenues average over $6
firm has
40,000 employees.
in
our sample are quite large
billion in
both dimensions.
constant 1990 dollars, and the average
Diversified firms are larger
firms for both scale measures.
in
on average than undiversified
^^
Firm Finar)cial Performance. While the theoretical and empirical literature suggests a
positive correlation
between
firm financial performance
and
CEO
pay, there
is little
consensus on the functional form of the relationship or the correct measures of
^^The positive relationship between firm size and executive compensation is one of the more
robust empirical regularities in the large literature on top executive pay- The formal ability
matching models, beginning with Rosen (1982), have been motivated in part by a desire to explain
this cross-sectional relationship. Others (e.g., Baker, Jensen and Murphy, 1988) have argued that
the tendency of Boards of Directors to reward firm size may be an artifact of rule-of-thumb
heuristics for determining CEO pay popularized by executive compensation consultants.
^^ All
noted differences
in
raw means between the
diversified
significant at the .01 level or better, unless otherwise noted.
and undiversified subsamples are
13
Fortunately, our empirical analysis suggests that the diversification
profitability.
how we
estimates are relatively robust to
We
control for performance.
consider two measures of firm profitability:
common
equity
(MKTROR) and
(ACCROR). Despite the
the stock market rate of return on
the accounting rate of return on book equity
similarity in overall
mean
returns (.18 for market and .15 for
accounting), the correlation between these variables
is
While
only .24.
we
from several performance specifications, the data appear to prefer a
results
specification that includes current and lagged market and accounting returns
determination of
CEO
compensation. Following Joskow and Rose (1994),
one- and two-year lags
we can
implies that
who have
The
full
in
market return and one-year lags
estimate the
held the office for
full
model only
in
compensation due to
return variance
may
for
in
in
the
we include
accounting return. This
"experienced" CEOs,
i.e.,
those
more than two years.
return specification includes a
differences
Lower
report
measure of stock
price volatility to control for
differential riskiness of firms' profit streams.
lessen the probability of large negative returns that could
lead the Board to replace the
CEO (Amihud and
Lev,
1
981 or lead to bankruptcy that
)
terminates employment (Rose, 1992). Diversification also
may reduce agency
costs
by improving the information available to the Board of Directors about the CEO's
quality, therby reducing the
1993).
If
CEO's compensation
risk (Aron,
1
988, Hermalin and Katz,
the base salary includes an "insurance" component that reflects risk-return
trade-offs,
should
it
fall
with reduced variance.
Financial risk
is
indexed by the
standard deviation of the firm's weekly stock market return over the
fiscal year,
SDRET.
Consistent with previous research
Stuiz, 1993),
average
The average market
versus
16%
profitability
return
is
six
(e.g.,
is
Montgomery and Wernerfelt,
1
higher at the undiversified firms
988; Lang and
in
our sample.
percentage points higher at undiversified firms (22%
for diversified firms); the
average accounting return
is
two percentage
14
(16% versus 14%
points higher for these firms
for diversified firms).
The higher mean
SDRET
returns at undiversified firms are associated with higher return variance.
averages 4.6 percentage points for undiversified firms, compared to 4.1 percentage
points for diversified firms.
CEO Characteristics.
Prior
work suggests
that
have significant effects on compensation
levels.
variables recording the chief executive's age
CEO (CEOTEN),
tenure as a
a
We
dummy
may
use Forbes data to construct
when appointed CEO
if
variable equal to
the
(AGE), years of
CEO was
one
if
the
hired from
CEO was
the
(FOUNDER).^*
There are notable differences
diversified
age, tenure, and background
dummy variable equal to one
outside the firm (OUTSIDE), and a
firm's founder
CEO
and undiversified
firm, diversified firms are
(14% versus 18%).^^
they assume the
CEO
of this age difference,
in
the distributions of
firms.
even
While most
CEOs
CEO
characteristics
between
are promoted from within the
less likely to hire outsiders than are undiversified firms
Diversified firms'
CEOs
are on average four years older
when
position than are those at undiversified firms. Perhaps because
CEOs
have a shorter average tenure as
at diversified firms also
CEOs (10.4
years versus 12.7 years for undiversified firms).'® There are very few
founders
the sample, but the probability that the
in
segment
single
commitment
firms.
This
may
reflect
more
CEO
is
a founder
is
specialized expertise,
to the original line of business by founding
larger for
a greater
CEOs, or differences
in
firm
age.
'*
OUTSIDE
is
^^
'*
which
one for a CEO who was employed by the firm fewer than four years
CEO and who was not the firm's founder. The results are not sensitive to
set equal to
before appointment as
the four year cutoff.
Significantly different at the .09 level.
Recall that these
is
means
on the CEO having more than 2 years of tenure,
the experienced CEO panel.
are conditional
a criterion for inclusion
in
15
CEO compensation
Industry and Time Effects.
norms and economy-wide movements
Shepard, 1993).
in
real
We
compensation
compensation
industries into "industry groups" (see
{6^)
to control for aggregate shifts
To accommodate differences
over time.
levels across industries,
strongly influenced by industry
executive pay (see Joskow, Rose, and
in
introduce year fixed effects
levels
is
we
first
classify similar two-digit SIC
Appendix Table A1).
We
in
that year.
weighted average of effects
for
The
fixed effect for
mean
code
then calculate the
share of the firm's revenues derived from each industry group for which
business segment sales
in
it
reports
each observation
is
the
each industry group, using these revenue shares as
weights:^'
K
WTGROUPj,
=
J^a^SHARE^.
k-1
where SHARE|^j^
in
is
industry group k
only
in
industry
the proportion of sales firm
in
year
t,
and
a^^ is
j
has attributed to business segments
the industry fixed effect for a firm operating
k.
Econometric Specification
For the
full
panel of experienced
CEOs we
estimate the following basic compensation
equation:
Some
previous work has assigned each observation to a single industry group based on the
primary SIC code for the firm (as recorded by CRSP or COMPUSTAT). If compensation at a
diversified firm is based on the set of industries in which it operates rather than its primary industry
only, this procedure could mis-estimate the diversification coefficient. When we estimate our
model with only a primary industry effect, the diversification coefficients are slightly higher than,
^^
but otherwise quite similar to, those
we
report below.
,
16
ln(CEO COMPENSATION,^
where
denotes the CEO,
i
= p,
j
DIVERSE,,
+
P4SDRET,,
+
P7OUTSIDE,
+
WTGROUPp
denotes the
subscript has been added because
RETURN
is
This specification
consistent
appointment,
is
(<5j)
+ Pa
assumes a constant
with
and
+ PftAGE,
FOUNDER,
+ 6, + €„
and
firm,
observe
elasticity of
most previous
financial
shift the
P3RETURN,,
t
denotes the year.
nnultiple
CEOs
The CEO
for sonne firms.
(ACCROR, MKTROR and
lags of
the corresponding parameter vector.
compensation with respect to firm
DIVERSE, CEO tenure, age
studies.
performance
proportional impact on compensation.
effects
P5CEOTEN,,
a vector of firm performance measures
these variables); P^'
size,
we
+
+ Pg '"(SALES,,) +
all
are
assumed
to
at
have a constant
OUTSIDE, FOUNDER, WTGROUP, and year
compensation curve up or down. The reported regressions adjust
the standard error estimates for possible heteroskedasticity and within-CEO correlation
in
the error,
IV.
€,^.^^
RESULTS
Results for several variants of the basic compensation equation are reported below.
We
first
describe
observations on
models of compensation estimated on our panel of 1505
480 experienced CEOs over 1985-1990.
We
then investigate
'* Both
the ability model and some versions of the entrenchment model imply that the error
contains a CEO-effect that is correlated with both diversification and compensation. In the case of
ability models, this CEO-effect ('unobserved ability') is what drives the predicted positive cross-
sectional correlation.
Entrenchment effects ('CEO power') play a similar role in the "taste for
The first-difference specifications can be thought of as implicitly
diversification' explanations.
conditioning on (or removing) the CEO- or firm-specific component of the error. The implications of
this conditioning vary across the ability and entrenchment explanations as described in section II.
This
is
what gives the data power
to distinguish empirically
between the explanations.
17
we
whether the patterns
CEOs
in their first
observe
panel of experienced
in this
CEOs
year of tenure. This allows us to distinguish diversification effects
due to job characteristics from those due to actions a CEO undertakes
management
the top
position.
compensation to changes
their
in
Our
third set of estimates
measures of
after
analyzes the response of
diversification of their firms. Finally,
diversification
assuming
These examine whether CEOs can
diversification.
compensation by increasing the
alternative
also apply to
to
the
identify
we explore
form
precise
raise
the
of
compensation-diversification relationship.
Panel Estimates of Diversification Premia
Table 3 reports parameter estimates for variants of the compensation equation using
our panel data set of
1
505 CEO-years.
In
specifications reported here and the broad
range of alternative specifications explored
in
preliminary analysis,
increases with the level of diversification reported by the firm.
statistically
and economically significant. Consider,
for
CEO compensation
This effect
is
both
example, the effect of moving
from an undiversified, single segment firm to the same size firm with two equal-sized
business segments, a
standard deviations.
SALARY
and to
by about
raise
move
An
12%
that increases
DIVERSE from 0.0
increase of this magnitude
by
13%
'"
$250,000
sample means. ^°
To
in
The
calculate the effect of this
is
estimated to raise
3%; columns 2 through
(standard error,
proportions correspond to an increase of roughly
at the
DIVERSE
(standard error of the estimate,
TOTAL compensation
increase of over
in
$140,000
5%; column
in
move
in
percentage terms,
5).^^
DIVERSE
all
coefficient generally
we compute exp
(
0.5
/^diverse
our log-linear model of compensation is normally distributed, predicted
be log-normal, and calculated as exp(X^ + WTGROUP + 6, +
variance(e)/2). The corresponding compensation increases for a one standard deviation (.27)
increase in DIVERSE are $75,000 to $82,000 in SALARY and over $143,000 in TOTAL
If
the error
compensation
in
in dollars will
These
else constant
1.0.
^°
4),
predicted salary and an
predicted total compensation, holding
point estimates of the
two
to 0.5, or about
'
18
are stable, although they tend to increase slightly with nnore complete controls for firm
performance; compare column
financial
with column 2, which adds current
1
accounting return and lagged return variables to the model.
The other
coefficients reported
largely unaffected
Table 3 are consistent with earlier studies and
in
compensation with respect to firm
SALARY and
.33 (.02) for
additional scale variable
in
is
measured by SALES,
size,
TOTAL
compensation.
column 4 adds
on the estimated diversification
premium
The
by the inclusion of diversification measures.
little
in
about .25 (.01)
for
Including employees as an
explanatory power and has no effect
This suggests the diversification
coefficient.^^
not due to simple differences
is
elasticity of
labor scale or intensity across diversified
and undiversified firms.
CEO pay
varies directly with both accounting and market returns, and the estimated
sensitivity of
pay to firm performance
included, consistent with results
example, increasing market return
a
SALARY
year,
increase of
in
in
in
much
larger
once lags
Joskow and Rose
(1994).
in
performance are
In
column
in
the current year,
the third year.
An
1.7% (0.5%)
in
SALARY and
5.7% (1.6%)
3.5%
SALARY. The estimated performance
is
.4%) increase
greater for
to that for
in
next year's
TOTAL compensation
(column
5),
the next
increase of 10 percentage points
current accounting return implies a
(1
3, for
the current year by 10 percentage points implies
1.2% (0.5%)
and 1.2% (0.4%)
is
increase
in
current
in
a
sensitivity
but the coefficient pattern
is
similar
SALARY.
compensation.
We have experimented with specifications that replace SALES with book assets and
employees, as in Joskow, Rose, and Shepard (1993). The size elasticity of .25 divides between
assets and employees in these equations, and the DIVERSE coefficient tends to be slightly smaller
and noisier, though in general within one standard error of the results in Table 3. We also have
experimented with specifications that allow the coefficient on In(SALES) to change as SALES
increases. This yielded an almost constant elasticity of .25 across all SALES ranges and no change
^^
in
the coefficient on DIVERSE.
19
The
compensation also increases with the
results indicate that
performance measures. Both
variability of firm
SALARY and TOTAL compensation
increase with the
An
standard deviation of weekly market returns for the firm, SDRET.
SDRET from
in
four percentage points to five percentage points (one standard deviation
increase from
TOTAL
increase
sample mean) raises predicted
its
by almost
6%
(1.6%). This
bear more compensation
risk
is
SALARY
by almost
5%
CEOs who
consistent with a pay premium for
through greater
(1%) and
performance-based
variability in their
compensation or through an increased probability of termination resulting from very
low performance
SDRET
including
diversification,
Because SDRET
realizations.
increases
the
is
negatively
magnitude
estimated
with
correlated
the
of
diversification effect slightly.
Finally,
compensation varies
in
expected ways with
modest age and tenure premiums. An
raises
SALARY
by
(0.2%; column
underlying upward trend
The
and founders
in
The
somewhat
results
1980s
for executive talent,
^^
captured
in
TOTAL compensation equation exhibit a similar pattern
3 generally are consistent with both ability-based
compensation
assuming the marginal return to
is
is
addition to the
the
1
entrenchment-based explanations of executive compensation.
Compensation
that
raises
likely to
These are reported
but
larger in magnitude.
Table
in
late
in
CEO
CEO
5% SALARY premium relative to internal
realize SALARY discounts of about 1 3% relative to non-founders.
point estimates for the
tend to be
age before becoming
additional year of tenure as
compensation over the
There are
characteristics.
These increments are
3).
Outside hires earn a
fixed year effects."
hires,
additional year of
0.4% (0.2%) and each
SALARY by 0.7%
CEO
in
will
ability
In
an
efficient
and
market
be higher at more diversified and larger firms,
increases with diversification and firm size.
increase with firm financial performance and with the risk
Table
A2
for the specifications in Table 3,
columns 3 and
5.
20
of executive compensation streams, reflecting the
contracts.
positive
Finally,
outcome
of efficient incentive
age and tenure effects may be associated with the
accumulation of human capital by top executives.
Positive diversification coefficients also
would be expected
by investment undertaken by self-serving CEOs to increase
if
powerful
CEOs have
a taste for both diversification
diversification
if
is
caused
their value to the firm, or
and compensation.
In
both
these cases, the positive tenure coefficient might be interpreted as returns to
managerial entrenchment that increases with tenure."
Compensation Equations
for First-year
CEOs
Information on compensation during the CEO's
first
year of office
may
allow us to
distinguish premia for higher ability from pay surpluses resulting from managerial
entrenchment.
by CEOs
If
the diversification premium
in their first
for diversification
is
and
all
subsequent years of
less plausibly explained
due to CEO-specific investments, as
in
diversification
program
alter their firm's
in their first
her/his
office.
A positive
is likely
to require
if
fit
should be earned
first-year
the
full
premium
some
lead time,
it
is
panel result
But because a
unlikely that
new CEOs
business mix rapidly enough to affect the compensation they
year as CEO. Instead, the
new CEO
will
be compensated based on
Since the successor
" The
positive correlation of pay with firm performance neither supports nor rejects
entrenchment hypotheses. Although most managerial entrenchment models suggest reduced payfor-performance sensitivity relative to optimal contracts, they need not imply zero correlation
between pay and firm
performance.
^*
We
have estimated equations that take
this statement literally, by replacing the current value
with its value during the last year of the preceding CEO's tenure. The
coefficient on DIVERSE increases slightly with this substitution, providing even further support for
of
DIVERSE
for
is
A CEO may be able
her/his unique abilities.
with the firm as shaped by the preceding CEO.^*
fit
it
by the effects of managerial entrenchment
the Shieifer-Vishny model.
to alter the business mix over time to
earn
a return to ability,
First-year diversification premia should be smaller
models.
can
is
new CEOs
21
must be
distinctly less well-suited to that business
mix
for this strategy to raise the
incumbent's compensation, the successor's premium from diversification should be
much
smaller.
arises
from the Board's
First-year premia also are likely to be smaller
compensation.
CEOs than
inability to control a
The Board
relative to
is
likely to
CEO who
last
have more bargaining power
CEOs who have
column of Table
in their first full
1
panel result
held that office long
among
reports descriptive statistics for the
relative to
enough
new
to affect the
other top managers.
1
72 CEOs we observe
year of office. As would be expected, they earn less
in
SALARY and
compensation than do experienced CEOs. They also tend to be employed by
total
slightly larger
This
panel.
and more
diversified firms than the
firms
in
a
average
consistent with shorter average tenure
is
(compare columns 2 and 3 of Table
new CEO
1),
CEO
among
experienced
CEOs
A move from
in
for
Table
DIVERSE
3,
ability
firms
which implies over-representation of these
are quite similar to those
in
Table 4.^^ The
reported for the
although the results are noisier for this smaller sample.
an single-segment firm to the same size firm with two equal business
segments implies an increase of
year
diversified
sample.
estimates
coefficient
our experienced
in
Estimated compensation equations for this sample are presented
first
full
prefers diversification and high
composition of the Board or to build a base of support
The
the
if
TOTAL
premium
compensation,
11%
all
(5%)
in first
else equal.
year
SALARY and 16% (7%)
This provides strong support for an
interpretation of the diversification effect.
These
results provide
support for entrenchment explanations that are based either on actions the
undertakes or on managerial power that
is
in
an increasing function of tenure
in
no
CEO
position.
ability interpretations.
First, CEO tenure
is based on column 3, Table 3, with three exceptions.
and therefore omitted. Second, only current performance measures are used
because first-year CEOs have no prior performance. Third, the founder and SDRET have been
dropped because they are quite noisy and generally estimated as close to zero.
*^
The
a constant
specification
(1
)
is
22
Compensation Changes
We
can
for
CEOs
directly test the claim that
examining the effect of changes
CEOs
CEOs
in
diversify to improve their
diversification
on changes
diversify to increase their compensation, these
correlated.
If
CEO
the panel and first-year
in
compensation by
compensation.
changes should be
results represent
between
To
test these
first
levels of
we
competing hypotheses,
We
differences.
to ability,
CEO
should be
compensation and diversification
1005 CEO-years
for
panel. ^^
Table 5 reports
estimate our basic compensation model
of the
summary
1505 CEO-years
first
These include changes
or nearly constant for a
difference equations. ^^
decreases, and no changes
due to movements
in
in
CEO
in
^^
80%
in
our experienced
CEO
compensation, DIVERSE, In(SALES), and
in
the levels equation are either
over time and therefore are excluded from the
The sample divides
fairly
evenly
among
DIVERSE. While most of the changes
segment shares
the 147 observations with changes
more than
in
statistics for the variables included in the first
the rate of return measures. The remaining variables
literally
.
have data on compensation, return, sales and diversification
changes
difference model.
positively
payment
however, the correlation between these changes for an incumbent
smaller than the correlation
If
rather than
in
of the total variation
changes
the reported
in
in
the
in
increases,
DIVERSE
are
number of segments,
number of segments account
for
DIVERSE.
Estimated compensation level equations for this sub-sample are quite similar to those for the
1505 observation dataset, reported in Table 3. The coefficient on DIVERSE
the 1005 observation sub-sample, ranging from .24 (standard error, .08) to .28
full
specifications, but within
one standard
is
slightly larger for
(.07) in the salary
error of the results reported in Table 3.
" The model includes year effects but not weighted industry effects. Including changes in the
weighted industry fixed effects in the model does not affect the qualitative results but substantially
increases the noisiness of the estimates. The differenced data do not appear to have sufficient
power
to estimate industry effects as well as the reported coefficients.
23
Results for the first-differenced models are reported
and 3
(Total)
report results for
first
in
Table
Columns
6.
1
(Salary)
The
differences of our basic specification.
estimated compensation effect of increasing DIVERSE
is
negative and relatively
substantial for both compensation measures. This suggests that the Boards rewarded
specialization
and penalized diversification during our sample period. The estimates
imply that splitting an undiversified firm into
increase of 0.5
in
compensation by
two
equal-size business
DIVERSE) would reduce SALARY by
17%
8%
level for the
SALARY coefficient and 0.15
even a conservative interpretation of these estimates
is
we
correlation
between
for
all
first
difference
although the precision of the estimated coefficients
ask more of the data. For example,
diversification
compensation to respond to
TOTAL. But
else constant.
robust to a broad variety of alternative
we have explored,
declines as
TOTAL
rejects the interpretation that
increasing diversification will increase compensation, holding
specifications
(4.6%) and
(an
(12%). The point estimates are noisy, and can be bounded
from zero only at the 0. 1
This pattern of results
segments
size
we explored whether the
and compensation changes was due to
changes when both
changed, or due to greater rewards for CEOs
who
diversification
focus the firm.
To
negative
failures of
and sales
test this
we
allowed the coefficients on diversification and sales changes to depend on whether
diversification increased, decreased, or
in
columns 2 and 4 of Table
5.
remained constant. These results are reported
While the point estimates are noisy, they provide
support for these alternative explanations.^^
2),
for
In
the
SALARY
regressions (column
example, diversification changes appear to be associated with somewhat
smaller elasticities of compensation with respect to sales.
and reward
for focusing
magnitude from the
^^
Point estimates
in
the firm are completely symmetric and unchanged
results in
the
The penalty for diversifying
TOTAL
column
1,
however.
regressions are more sensitive to this split and
consistent with the alternative explanations. Their substantial standard errors
difficult,
little
however.
may
makes
in
part
intepretation
in
24
These
results are difficult to reconcile with managerial
entrenchment explanations of
diversification. In contrast to the predictions of the Shieifer-Vishny model, there
premium
CEO from
to the
"tailoring" the firm to
fit
no
is
the CEO. Changing the mix of the
firm without substantially increasing the scale of the firm reduces rather than
increases the CEO's compensation.
It is
more difficult
to empirically falsify the "taste"
version of the entrenchment-based explanations, but our results suggest
costly for a
CEO
to indulge such a taste.
The implication of these
results for overall
compensation movements depends critically
on what other changes accompany changes
increases
diversification,
in
compensation declines
5 indicates,
if
accompanied by
by
The negative DIVERSE
less than
not
adds a second segment of equal
in
Column
experiment for newly-hired
increase. That
is,
more than a CEO
a
CEO
hired
for
CEOs who reduce
1
is
is
size, increasing
CEOs
growth
in
more
rewarded
SALARY
(Table 4,
segment
the
than
when
the
firm that
by
In(sales)
1.
increment of 13%. The same thought
Column
1)
would imply a
by a firm with two equal-sized segments
larger
increase
diversification.
DIVERSE by .50 and
by a firm with one segment and
much
actual
As Table
CEOs who
Alternatively, consider a single
imply a
hired
DIVERSE)
CEO
is
32%
will
salary
earn
32%
half the sales. ^^ This pattern
provides additional support for ability-based explanations.
for diversification
for
for
secondary businesses (which increases DIVERSE),
in
controlling for overall firm size.
coefficients
imply
necessarily
sufficient increases in firm size.
reduces
(which
business
primary
does
it
The penalty
coefficient indicates that disproportionately high
disproportionately high growth
The
diversification.
does
example,
for
in
compensation increases on average even
diversification, but
firm's
may be
it
We
find that the
premium
optimally-matched to the firm
(i.e..
'* Similar calculations can
be made for the reverse move: focusing the firm. A CEO who sells
one of two equal-sized business segments reduces DIVERSE and In(SALES) by 0.5. The results in
column 1 imply that the incumbent CEO's salary would fall by 2%. The results for new CEOs in
table 4 imply a decline of 24%.
25
when
when
the estimates allow ability to vary with the level of diversification) than
the incumbent
we would
CEO
expect
retained
is
if
(i.e.,
holding the ability of the
ability-matching
is
CEO
fixed). This
is
what
the dominant source of the diversification
premium.^"
The
negative
compensation
correlation
is
late
changes
in
diversification
and
executive
also consistent with market preferences for industrial focus or
specialization over our
diversification
between
sample period Recent empirical 'esearch suggests that greater
.
was on average associated
with reduced shareholder wealth during the
1980s. Studies have concluded that diversified firms tend to be less profitable
than comparable undiversified firms (Lang and Stuiz,
Wernerfelt,
1993; Montgomery and
1988); that the stock market reacts less favorably to diversifying
acquisitions than to acquisitions of business closely related to the firm's existing line
of business
includes a
(Morck, Shieifer and Vishny,
1990).^^
If
executive compensation
component that rewards improvements to and penalizes degradation
performance beyond the response to changes
in
return,
expected outcomes of adding or subtracting diverse
pay might adjust to
lines of business.
imply a compensation penalty to increased diversification, as observed
This result echoes Lambert and Larcker
(1
compensation through acquisitions only
987),
if
who find
that
CEOs can
of firm
reflect the
This would
in
Table
6.
increase their
the acquisitions increase the shareholder
value of the acquiring firm.
^° For example, because the incumbent CEO has higher ability than would be optimal for a
newly focused firm, she/he is paid more than the alternative, well-matched and lower-ability
candidate would be. Because the incumbent CEO has lower ability than would be optimal for a
newly diversified firm, she/he is paid less than the alternative, well-matched and higher-ability
candidate would be.
''
There is conflicting evidence on the effect of diversification on firm performance and on the
stock market's reaction to diversification over time. For example, Matsusaka (1993) finds that the
stock market responded positively to diversifying mergers in the 1 960s, and Morck, Shieifer and
Vishny report that the negative reaction to diversifying mergers is more pronounced in the 1 980s
than in the 1 970s. Studies finding a positive effect on firm performance from diversification also
tend to be using data for periods prior to 1 980. For our sample period, the most common finding is
a negative share price reaction to diversification announcements.
26
The Form of the
The
Diversification
specifications estimated
Premium
above assume that compensation increases proportionally
with DIVERSE. While the qualitative patterns are robust to alternative measurements
depend on the
of diversification, the numerical estimates
particular index
we
We
use.
have explored many formulations of the diversification measure and report results
two
additional
parts:
one
reflecting the
The second measure
size.
reported
effect
in
Tables
7A and
first
may
measure decomposes DIVERSE
segments and another the variance
diversification premia
provide
segments and the
some
insight
segment
based on these measures are
the basic compensation equation, which are
We explore this
of segments, the
size distribution of
how the
on
segment share concentration,
number
two
DIVERSE have no substantive
7B. These substitutions for
in
in
into
number of distinct business segments the
a count of the
investigate whether compensation responds similarly to the
with diversification.
the
first
the table size.
limit
different business
of
is
of
on the remaining coefficients
suppressed to
This
number
(NUMSEG). Estimated
firm reports
We
measures below. The
for
segments within the
of
firm.
complexity of the management task varies
question by decomposing DIVERSE, a measure
into
two terms. The
second decreases
DIVERSE
number
= -
1
+
(
in
1
first
increases (nonlinearly)
the variance of segment size
-
no^
in
:
),
n
where o^
Table
7A
is
the variance of segment revenue shares
reports results for this decomposition
compensation specifications.
drives the entire
DIVERSE
in
in
the basic
The estimates suggest
result,
the firm.
SALARY and TOTAL
that the
number component
with coefficients on the number component that are
27
identical to
The
3).
those on DIVERSE
coefficient
firm, holding
its
all
CEO 12% more on
else constant.
Table
in
two
average than an undiversified, single-
Each additional segment
progressively smaller increments to compensation, with an overall
for a
in
on the number component implies that a firm operating
industry segments pays
segment
(compare Table 7A to columns 3 and 5
itself
is
associated with
SALARY premium
10-segment firm of 21 %. The data do not suggest any effect of the variance
component on compensation." This seems
increases the complexity of the
to imply that adding business
management task even
if
the
segments
new activities
are small
relative to the original line of business.
We
further explore the nonlinearity of
segments
in
table 7B.
columns
In
1
compensation with respect to the number of
(SALARY) and 3 (TOTAL), we
report separate
premia, relative to compensation at undiversified single-segment firms, for each value
of
NUMSEG
(2
through 10 segments). Because
the number of segments rise (see Table
across similar
NUMSEG
cell
we
2),
counts drop off quite sharply as
also consider estimates that pool
These estimates are reported
categories.
in
columns 2
(SALARY) and 4 (TOTAL).
While the relationship between diversification and
it
differs
seem
in
Table 7B
is
nonlinear,
from the form imposed by the construction of DIVERSE. Diversified firms
to partition roughly into three categories based
two segments, three through
five
segments, and
categories form the basis of the estimates
single-segment to a two-segment firm adds
3-5 segment category adds an additional
firms, but
"
SALARY
no incremental premium
in
on reported segment counts:
six or
column
2,
more segments.
Table 7B.
Moving from
0% to SALARY. A further move into the
7% premium relative to single segment
segments within
this category.
sample
is
subdivided
quite stable across tenure categories.
in this
CEO
tenure, although the point estimates
way. The coefficient
CEOs
component may
become
on the numbers component is
Detailed investigation of this decomposition suggests that the variance
quite noisy as the
a
1
for increased
influence compensation during the early years of
These
28
of the
most
diversified firms
premium, implying
a
SALARY
of undiversified firms,
The pattern
for
all
that
TOTAL compensation
more segments. The premium
estimated at
almost
is
17%,
larger than
increases by another
flat
thereafter (column 4).
to the
a third
segment
is
This suggests a smaller
SALARY compensation premium
at
TOTAL compensation are substantially noisier
SALARY
compensation, and the point estimates generally are within
we
find in the
SALARY
regressions.
suggest that the distinction between diversified and undiversified
results
more
most
significant than
distinctions within the class of diversified
firms, with the possible exception of those
few
more business segments). This pattern
required managerial inputs
when
diversified. This could reflect
a firm
changes
for a substantively different
results for the
most
in
firms with highly diffuse operations
is
consistent with a discrete jump
changes from specialized (one-segment)
type of chief executive, or both.
heavily-diversified firms, while based
distinct business
segments.
in
to
the hierarchical organization of the firm, the
companies, suggest a further discrete change
numbers of
3% when
diversification effects for
a standard error of those
need
the corresponding premium for
not overemphasize this distinction, however.
than those for
(6 or
CEOs
One should
the most diversified firms.
is
for
(columns 3 and 4) seems to imply only two
TOTAL compensation premium compared
enterprises
SALARY
adding a second segment to an undiversified firm
for
substantially
added, but remains relatively
These
higher than the
those with two segments and those with three or
SALARY. TOTAL compensation
The estimated
30%
11%
else equal.
categories of diversified firms:
is
through 10 segments) earn an additional
(6
in
The SALARY
on small numbers of
the complexity of managing large
29
V.
CONCLUSIONS
During the late 1980s, firm diversification appears to have been associated with
significantly higher
compensation
distinct lines of business earn
bonus paid to CEOs of
additional $1
CEOs
for the chief executive.
10%
on average
to
12% more
similar but undiversified firms.
15,000 to $145,000 per year
relative to the
of firms with
two
than the salary and
This corresponds to an
mean
salary and
bonus
for
our sample of top CEOs. Moreover, the effect of diversification on compensation
as strong for first-year
CEOs
CEOs who have been on
as for
premium
years. This suggests that the diversification
Its
demands, and not a
changes
result of
a characteristic of the job and
instituted
by incumbent managers to
inconsistent with
it
is
own
diversification during out
agendas.
sample period, holding
appear to reduce rather than increase compensation.
constant,
While
in
the job three or more
is
increase their value to the firm or to pursue their
For a given CEO, increases
CEOs undertaking
difficult to falsify
diversification to increase their
the hypothesis that
may
taste for diversification, indulging this taste
CEO
regressions suggest that the average
corporation without paying a price
be earned by increasing growth
in
in
CEOs.
If
diversify
This appears
compensation.
because they have a
be costly for the CEO. The change
has insufficient control to diversify the
diversification premia as rents earned by
diversification increases the marginal return to ability,
earn a higher compensation.
CEOs and
increments to changes
else
the core business.
labor market will allocate higher ability
experienced
CEOs
all
foregone compensation compared to what could
Our findings support an interpretation of
high-ability
is
This
CEOs
is
to
more
diversified firms
efficient
where they
will
consistent with the premium observed for
the equally high premium for first-year CEOs.
in diversification for
matching models with turnover costs.
an
The lower
incumbent CEOs are also consistent with
Because turnover
is
costly for both the firm
30
and the incumbent, bargaining produces an increment that
observed
While
If
this
the salaries of
study
diversification
smaller than would be
two well-matched CEOs were compared.
provides
direct
evidence only on the
and executive compensation,
of the disappointing performance of
critical
is
many
it
may
provide
relationship
some
diversified firms.
If
between
clues to the source
ability-matching
is
a
determinant of compensation patterns across firms, our results suggest that
diversified firms
may
simply be more
difficult to
mar.age successfully.
31
REFERENCES
Amihud, Y. and
Reduction as a Managerial Motive for Conglomerate
Mergers." Bell Journal of Economics, Vol. 12 (1981), pp. 605-617.
B. Lev. "Risk
Aron, D.J. "Ability, Moral Hazard, Firm Size and Diversification."
Rand Journal of
Economics. Vol. 19 (1988), pp. 72-87.
Baker, G.F., Jensen, M.C. and Murphy, K.J. "Compensation and Incentives: Practice
vs. Theory." Journal of Finance, Vol. 43 (1988), pp. 593-616.
Comment,
and
G. A. "Corporate Focus, Stock Returns, and the Market for
Corporate Control." Bradley Policy Research Center Managerial Economics Research
R.
Jarrell,
Studies Working Paper
Finkelstein, S.
MR
91-01, University of Rochester,
May 1992.
and Hambrick, D.C. "Chief Executive Compensation:
Intersection of Markets and Political Processes." Strategic
A Study
of the
Management Journal,
Vol.
10 (1989), pp. 121-34.
Monahan, J.L. "A Generalized Index of Diversification: Trends in U.S.
Manufacturing." Review of Economics and Statistics, Vol. xx (1991), pp. 318-330.
Gollop, F.M. and
Hermalin, B.E. and Katz, M.L. "Corporate Diversification and Agency," mimeo, 1993,
UC Berkeley.
Jensen, M.C. and Murphy, K.J. "Performance Pay and Top Management Incentives."
Journal of Political Economy, Vol. 98 (1990), pp. 225-64.
Joskow,
P.L.
and Rose, N.L. "Pay-for-Performance:
A Reassessment," mimeo,
1994,
MIT.
Joskow, P.,
Rose, N. and Shepard, A. "Regulatory Constraints on CEO
Compensation." Brookings Papers on Economic Activity: Microeconomics, No. 1,
1993,
pp. 1-58.
Lambert, R.A. and Larcker, D.F. "Executive Compensation Effects of Large Corporate
Acquisitions," Journal of Accounting and Public Policy. Vol. 6 (1987), pp. 231-242.
Lang, L.H.P. and Stuiz, R.M.
"Tobin's Q, Corporate Diversification and Firm
Performance." NBER Working Paper No. 4376, June 1993.
Lichtenberg, F.R. "The Managerial Response to Regulation of Financial Reporting for
Enterprise." Journal of Regulatory Economics, Vol 3 (1991 ),
Segments of a Business
pp.
241-249.
Mace, M.L. Directors: Myth and Reality. Division of Research, Graduate School of
Business Administration, Cambridge MA, Harvard University, 1971.
Matsusaka, J. G. "Takeover Motives During the Conglomerate Merger Wave," Rand
Journal of Economics, Vol. 24 (1993), pp. 357-379.
Montgomery, C.A. and Wernerfelt,
q."
Rand Journal of Economics,
B. "Diversification, Ricardian Rents,
Vol.
and Tobin's
19 (1988), pp. 623-632.
A. and Vishny, R.W. "Do Managerial Objectives Drive Bad
Acquisitions." Journal of Finance, Vol 45 (1990), pp. 31-48.
Morck,
R.,
Shieifer,
Ravenscraft, D. J. and Scherer, F. M. Mergers, Sell-offs, and Economic Efficiency.
Washington, D. C: The Brookings Institution, 1987.
"The Hubris Hypothesis of Corporate Takeovers." Journal of Business, Vol.
59 (1986), pp. 197-216.
Roll, R.
Rose,
D.C.
"Bankruptcy Risk, Firm-Specific Managerial Human Capital,
Review of Industrial Organization, Vol. 7 (1992), pp. 65-73.
and
Diversification."
Rosen, S. "Authority, Control and the Distribution of Earnings."
Economics, Vol. 13 (1982), pp. 311-323.
Bell Journal of
Rosen, S. "Contracts and the Market for Executives." In L. Wernin and H. Wijkander,
Economics. Oxford: Basil Blackwell, 1992.
eds.. Contract
and Vishny, R.W. "Management Entrenchment: The Case of ManagerSpecific Investments." Journal of Financial Economics, Vol. 25 (1989), pp. 123-39.
Shieifer, A.
Waldman, M. "Worker Allocation, Hierarchies and the Wage
Economic Studies, Vol. 51 (1984), pp. 95-109.
Distribution."
Review of
Table
1
:
Descriptive Statistics
Table
Number
digit
2:
Distribution of Diversification in
of reported 4-
SIC segments
Sample
Table 3:
Variable
Determinants of Compensation
N = 1505 CEO-years
Table 4:
Determinants of Compensation,
N = 172 CEOs
Variable
First- Year
CEOs
Table
Variable
5:
Descriptive Statistics, First Differences
Table
6:
Determinants of Changes
N = 1005
Variable
in
Compensation
Table 7A: Decomposition of DIVERSE Effects
N = 1505 CEO-Years
Diversification
Table 7B: Diversification Effects by Number of Segments
N = 1505 CEO-Years
Number
of
Segments
Appendix Table A1: Industry Classification and Fixed Effects
Category
Appendix Table A2: Year
Dependent Variable:
FixecJ Effects
MIT LIBRARIES
3
lOfiO
OOflqB'ifil
T
Date Due
m
LJb-26-67