Does Performance-Based Managerial Compensation Affect Corporate Performance?
Author(s): John M. Abowd
Source: ILR Review, Vol. 43, No. 3, Special Issue: Do Compensation Policies Matter? (Feb.,
1990), pp. 52S-73S
Published by: Sage Publications, Inc.
Stable URL: http://www.jstor.org/stable/2523571
Accessed: 06-04-2018 11:01 UTC
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide
range of content in a trusted digital archive. We use information technology and tools to increase productivity and
facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
http://about.jstor.org/terms
Sage Publications, Inc. is collaborating with JSTOR to digitize, preserve and extend access to
ILR Review
This content downloaded from 147.251.55.59 on Fri, 06 Apr 2018 11:01:31 UTC
All use subject to http://about.jstor.org/terms
DOES PERFORMANCE-BASED MANAGERIAL
COMPENSATION AFFECT CORPORATE
PERFORMANCE?
JOHN M. ABOWD*
The author, using 1981-86 data on more than 16,000 managers at
250 large corporations, investigates whether the sensitivity of managerial
compensation to corporate performance in one year is positively related
to corporate performance in the next year. Accounting-based measures
of performance yield only weak evidence of such an association, but
economic and market measures yield stronger evidence. Payment of an
incremental 10% bonus for good economic performance is associated
with a 30 to 90 basis point increase in the expected after-tax gross
economic return in the following fiscal year; and payment of an
incremental raise of 10% following a good stock market performance is
associated with a 400 to 1200 basis point increase in expected total
shareholder return.
JN order to quantify the potential gains
from performance-based managerial
compensation, in this study I specify and
estimate two related models of the connec*
John M. Abowd is Professor of Labor Economics
and Management at Cornell University and Research
Associate at the National Bureau of Economic
Research. The author is grateful to George Milkovich
and a major compensation consulting company
for providing the data for this study. The author also
thanks Ronald Ehrenberg, Robert Gibbons, Edward
P. Lazear, and Kevin J. Murphy for extensive
comments on earlier drafts. The Alfred P. Sloan
Foundation and the National Science Foundation
(grant number SES-8813847) provided financial
support. The data were converted to research files
with the support of the Center for Advanced Human
Resource Studies at Cornell University.
A data appendix describes the procedures used to
assemble the managerial compensation data base
used in this study. The compensation data are
confidential but the author's access is not exclusive.
The executive compensation data used by Jonathan
Leonard (this issue) are substantially the same as
those used in this study. The financial data used in
this study may be licensed from Standard and Poor's
COMPUSTAT service.
tion between increased performancesensitivity
in compensation and increa
subsequent corporate performance. These
models control for the historical levels of
both compensation and performance so
that it is possible to focus on the extent to
which changes in the correlation between
current compensation and current performance
affect future performance.
The first model, a discrete formulation,
focuses on the conditional probability of
good future corporate performance given
current corporate performance and the
current association between pay and performance.
In this model increased performance-sensitivity
is accomplished by increasing
the probability of high pay when
there is high performance and low pay
when there is low performance.
The second model, a continuous formulation,
focuses on the conditional expectation
of future corporate performance
given an elaborate, nonlinear function of
current performance and compensation.
Industrial and Labor Relations Review, Vol. 43, Special Issue (February 1990). C) by Cornell University.
0019-7939/90/43SP $01.00
52-S
This content downloaded from 147.251.55.59 on Fri, 06 Apr 2018 11:01:31 UTC
All use subject to http://about.jstor.org/terms
PERFORMANCE-BASED COMPENSATION 53-S
In this model performance-sensitivity in
the compensation system can be varied
continuously. The effects of this performance-sensitivity
are captured by two
interaction terms that measure the association
between future performance and
current compensation when current performance
is below average and when
current performance is above average.
The models are specified and interpreted
using a general principal-agent
framework in which the stockholders'
compensation contract with the managers
varies across companies and years in the
extent of performance-sensitivity. The
data used to test the models, obtained
from a survey conducted by a major
compensation consulting firm, cover some
25,000 managers at 600 corporations for
the period from 1981 to 1986.
The Essential Features of
Principal-Agent Models
Performance-based managerial compensation
has recently attracted considerable
attention in the professional literatures of
economics, accounting, and human resource
management. (See Ehrenberg and
Milkovich 1987 for a comprehensive review.)
This interest is justified by the belief
that contingent, performance-based compensation
provides a viable solution to the
problem of aligning the interests of managers
with those of the owners of the
corporations that they manage. In the
conventional agency cost formulation, the
shareholders of the corporation are the
principals and the managers are their
agents. Manipulating the degree of performance-sensitivity
in the manager's compensation
contract is the principal's method of
controlling the tradeoff between better
management and increased compensation
costs.
Ross (1973) first posed the basic principal-agent
problem as it pertains to corporate
managers. In his formulation the
compensation contract is chosen so as to
elicit actions by the agent that maximize
the principal's utility subject to the constraints
of a reservation utility level for the
agent (feasibility) and private optimality of
the agent's action (incentive compatibility).
Becker and Stigler (1974) and Lazear
(1979) recast the problem in terms of
implementing a long-term implicit contract
in which the manager's compensation
over time provides the correct incentives.
Holmstr6m (1979) and Grossman and
Hart (1983) analyzed in considerable
detail the theoretical structure of performance-based
compensation systems designed
to mitigate single-period principalagent
problems. (See Hart and Holmstrom
1987 for a comprehensive review.)
The Grossman-Hart model solves the
agency problem by showing that there are
two conceptual steps to an optimal program.
First, the principal chooses a different
contingent compensation plan relating
pay to performance outcomes for each
action the agent might take. Second, by
choosing a particular plan, the principal
chooses an optimal action for the agent
that maximizes the principal's utility net of
the expected cost of the chosen pay plan.
It is the Grossman-Hart formulation of
the problem that makes clear the fundamental
tradeoffs involved in performancebased
compensation. Actions that the
agent dislikes (relative to their alternatives)
are more costly to implement and
require a greater degree of performancesensitivity
in the compensation plan. The
expected cost of a compensation system
must increase as it becomes more performance-based-expected
payroll costs and
the degree of performance-sensitivity in
the compensation plan are positively
correlated. By the same token, the expected
performance of the corporation
also increases as the degree of performance-sensitivity
in the compensation plan
increases-expected corporate performance
and the slope of the payfor-performance
relation are positively
correlated. Because expected payroll costs
and expected managerial performance are
both increasing as the degree of performance-sensitivity
in the compensation contract
increases, the solution to the principalagent
problem generally occurs when the
incremental payroll costs just equal the
value of the incremental performance
This content downloaded from 147.251.55.59 on Fri, 06 Apr 2018 11:01:31 UTC
All use subject to http://about.jstor.org/terms
54-S JOHN ABOWD
gains associated with the chosen level of
performance-sensitivity.
In the financial economics literature,
Jensen and Meckling (1976) and Fama
and Jensen (1983a, 1983b) demonstrated
that the agency costs associated with
running a large corporation are intrinsic
to organizations in which ownership and
control are separated. Furthermore, it is
efficient to encourage the separation of
ownership and control because the limited
wealth of all individual investors prevents
undertaking large positive net present
value investments unless investors pool
resources, thus creating organizational
control problems. Agency costs reduce the
gains from the separation of ownership
and control but do not eliminate them.
The principal-agent literature, in spite
of its apparent applicability to the design
of compensation systems, has not been
easy to translate into empirically tractable
models. There are two basic problems.
Lazear (1986) showed that the appropriateness
(optimality) of contingent performance-based
managerial compensation
contracts depends critically on the assumption
that direct monitoring of the agent's
actions is prohibitively costly. When a
relatively inexpensive monitoring system
is available, both managers and principals
will prefer noncontingent (salary-based)
compensation systems with performance
appraisals based on the information generated
by the monitoring system. To the
extent that monitoring costs vary across
firms the predicted positive relation between
strong incentive pay and corporate
performance may not hold. The present
study cannot control for differential monitoring
costs in any meaningful way.
A second problem in testing principalagent
models of managerial compensation,
identified by Miller and Scholes
(1982) and also studied in Lewellen et al.
(1987), is that many apparently performance-based
compensation systems are
actually designed to minimize the total tax
burden of the principal and the agent.
Hence, the measured performancesensitivity
in the compensation system is a
veil for tax avoidance. Contingent, performance-based
compensation that is specifically
designed to increase the joint tax
liability of the corporation and the managers
may be explained on incentive grounds
alone. Most contingent compensation systems,
however, reduce the joint tax liability
of the corporation and its managers, so
that the tax consequences of the compensation
system must be controlled before
the incentive effects can be determined.
The present study considers only annual
corporate performance and annual cash
compensation (including amounts that the
manager elects to defer). This focus
reduces the potential for tax-related considerations
to confound the results.
Most studies of managerial compensation
have investigated the empirical relation
between the level or rate of change of
managerial compensation and corporate
financial, economic, and market performance
indicators. (See, for example, Lewellen
1968; Lewellen and Huntsman 1970;
Masson 1971; Murphy 1985, 1986; Antle
and Smith 1986; Jensen and Murphy
1987, 1988; Baker et al. 1988; Leonard,
this issue; and Gibbons and Murphy, this
issue.) When current compensation is
shown to be sensitive to these performance
indicators, the system is declared
performance-based, whether or not an
explicit formula exists that links current
compensation to the indicators. Disputes
arise as to whether or not the observed
degree of sensitivity of compensation to
performance is adequate to solve the
principal-agent problem between the owners
and the managers. (See Baker, Jensen,
and Murphy 1988.)
In contrast to the fairly large number of
studies investigating the sensitivity of
managerial compensation to corporate
performance measures, comparatively little
research exists on the efficacy of
performance-based compensation systems.
The notable exception is Larker's
(1983) study of the returns on investment
decisions made by managers paid with
different types of executive compensation
systems. Masson (1971) also attempted to
address this question as a part of his
analysis of executive compensation and
common stock performance. In financial
economics the event study methodology
This content downloaded from 147.251.55.59 on Fri, 06 Apr 2018 11:01:31 UTC
All use subject to http://about.jstor.org/terms
PERFORMANCE-BASED COMPENSATION 55-S
has been used to assess the performance
effects of some kinds of contingent compensation.
(See Bhagat et al. 1985; Brickley
et al. 1985; Coughlan and Schmidt
1985; Lambert and Larker 1985; and
Tehranian and Waegelein 1985.) Recent
work in compensation (Gomez-Mejia et al.
1987; Rabin 1986, 1987), accounting (Lambert
and Larker 1987), and financial
economics (Lewellen et al. 1987) also
begins to address these issues. Healy
(1985) considered the agency cost-related
problem of manipulation of accounting
quantities when contingent compensation
is based on accounting performance measures
rather than on economic or market
measures.
Tests of Compensation System
Design and Effectiveness
When the sensitivity of compensation to
performance measures is increased, under
what conditions does the subsequent performance
of the corporation improve,
worsen, or remain the same? Since an
optimal compensation system balances the
gain from additional performancesensitivity,
which takes the form of incremental
corporate performance, against
the cost of additional performancesensitivity,
which takes the form of higher
average compensation, the answer to this
question is at the heart of the study of the
design and effectiveness of compensation
systems. The gains are achieved because
the manager's extra effort induced by the
greater return to performance in the
compensation system increases the probability
of favorable corporate outcomes.
The costs are incurred because a feasible
performance-based compensation system
must deliver greater expected total compensation
the greater the effort level the
system tries to induce from the managers.
Neither the expected total cost of the
compensation system nor the expected
corporate performance improvement can
be calculated without quantitative measures
of the relation between the performance-sensitivity
of compensation and
future corporate outcomes.
It may be that the apparent complexity
of determining the optimal compensation
design and then validating that design by
quantifying the improvements in performance
that it caused has obscured the
important and practical implications of
quantifying the relationship between characteristics
of the current managerial pay
system and subsequent corporate performance.
If the existence of this relationship
and some estimate of its magnitude could
be inferred from the sensitivity of corporate
performance to the pattern of performance-based
contingencies in a sample of
corporate compensation plans, then compensation
designers could use the estimated
change in corporate performance
from such a sample to justify a modification
of the structure of a particular plan.
The ability to evaluate existing compensation
plans using formulas that reflect the
consequences of incentives would greatly
facilitate the comparison of such plans.
Statistical Models for Contingent
Compensation Effects on
Performance
Any statistical model of the relation
between compensation and current performance
must begin with the equation
describing pay for the individual manager.
To specify that function, define:
yijt = the compensation measure for manager
i in company j for year t;
xijt = the personal characteristics of manager
i (including job level at company
1) for year t.
Since corporate performance does not
vary for managers within a particular
company, a general form for the compensation
equation is:
(1) Yijt = Olit + (Xiyt + Uijt
where oijt = the effects on compe
of being at company j for year t, called the
company * year effects on compensation; f
= the effect of xiit on compensation; uijt =
the statistical error term associated with
(1); and x = the grand mean of xijt. The
individual characteristics are expressed as
deviations from the grand mean in order
This content downloaded from 147.251.55.59 on Fri, 06 Apr 2018 11:01:31 UTC
All use subject to http://about.jstor.org/terms
56-S JOHN ABOWD
to force the estimated otjt through the
grand mean of yjit.
Because all the information about the
link between corporate pay and performance
is contained in the company year
effects otjt, I specify a statistical model
relating these effects to annual performance.
Let qjt = the estimated otj from
equation (1) and pat = the performance of
company j for year t. Then the implications
of principal-agent models for compensation
design and annual performance
reviews are:
(2a) q= at + b tpt + vjt
(2b) PJt+1 = % + Olajt + 02bjt
+ 03Pjt + ejt+I
where aj, = the intercept of the compensation-performance
relation (2a); bjt = the
slope of the compensation-performance
relation (2a); vjt = the statistical error in
equation (2a); ok = the parameters of the
future performance relation (2b); and
e t+ I = the statistical error in equation
('2b). A pay system is performance-based if
the slope bjt is positive. A performancebased
system is consistent with a solution
to the principal-agent problem if, as b-t
increases, the intercept ajt falls. The payoff
to low performance outcomes must be
lower and the payoff to high performance
outcomes must be higher the greater the
work effort the compensation contract
induces. The performance-based pay system
is valid if 01 < 0 or 02 > 0, indicating
that increasing the performance-sensitivity
in compensation does increase subsequent
performance.
It is not possible to test equation (2b)
directly, because we do not have repeated
observations on the pjt and qat variables for
each company year; hence, equation (2a)
cannot be estimated. The predictions
regarding the system in equations (2) can
be explored under the maintained hypothesis
that comparison of the pair (pjt, qjt)
with the median outcomes for performance
and pay for all companies in year t
provides information about the values of
ajt and bjt.
Figure 1 illustrates this maintained
hypothesis. In the figure, performance is
o0 (C) (A)
can Both PerformanceC) 'and Compensation
E above Media (II)
. (I
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~.c.....:."...:.::,'I
>I''-'~~~~~~~~~.'":.'.................................'-,
Both Performance
and Compensation
below Median
(B) (D)
Median Performance
Figure 1. An Illustration of the Maintained
Hypothesis About the Location of More vs.
Less Performance-Sensitive Outcomnes.
plotted along the horizontal axis and
compensation along the vertical axis. The
graph is divided into four regions by the
dashed lines at the medians of performance
and compensation. Region (A)
contains points that are above both medians.
Region (B) contains points that are
below both medians. Regions (C) and (D)
contain points that are above one median
but below the other. The maintained
hypothesis is that systems in which compensation
is more performance-based are
more likely to produce outcomes in regions
(A) and (B).
The argument, which can be made
rigorous for certain conditional probability
distributions of the error term 15t i
illustrated by the lines (I) and (II) and the
shaded areas surrounding them. Line (I)
represents a pay-for-performance relation
that is not very sensitive (high
intercept aft and low slope b1t); line (II
pay-for-performance relation that is very
sensitive (low intercept, high slope). In
both cases the shaded area around the
lines contains the likely scatter of outcomes
around the basic relation. The
proportion of the outcomes for line (II)
that lie in regions (A) and (B) is greater
than the similar proportion of outcomes
for line (I). Hence, an outcome in region
(A) or (B) is more likely to have come from
This content downloaded from 147.251.55.59 on Fri, 06 Apr 2018 11:01:31 UTC
All use subject to http://about.jstor.org/terms
PERFORMANCE-BASED COMPENSATION 57-S
a highly performance-sensitive compensation
system than is an outcome in region
(C) or (D).
The Discrete Model
The following model incorporates the
maintained hypothesis in a test of the
efficacy of certain pay-for-performance
systems. The test has the advantage of
being simple and direct. It has the
disadvantage of being difficult to translate
into an estimate of the magnitude of the
effect of increasing performance-sensitivity
on future performance. This problem
is handled in the continuous model.
Let:
p*. 1 when pj, > median overj for year
1t t of all Pj1; 0 otherwise.
q*j 1 when qjt > median overj for year
a 1t of all qjt; O otherwise.
[1 when the total assets of company
= Jj for year t exceed the median total
it- assets for year t on the New York
Stock Exchange; O otherwise.
The test of the average sensitivity of
compensation to performance is based on
the following log linear model for the
probabilities:
(3) log Pr{q*1t = I| p* t, X*jt}
= 40 + (XPiP*jt
+ +2X*jt.
Equation (3) is called the compensation
equation in the discrete model. The
parameter 4) measures the average sensitivity
of compensation to the particular
performance measure specified. Since I
cannot verify that every performance
measure I consider is appropriate, the
compensation equation is used as a test
that the average effect of a particular
performance measure on compensation is
actually positive. Such a test is germane
because the average bjt from equation (2a)
over all companies and years must be
positive, which implies that the parameter
M1 will be positive. The variable x*,t is
included to control for sample design
problems in the statistical analysis.'
The test of whether greater sensitivity
of compensation to performance is associated
with increased future performance is
based on the following log linear model
for the conditional probability of next
year's performance given this year's compensation,
performance, and size control.
(4) log Pr{p*jt + 1 =1 p*jt, q*jt x}
=80 + 5iP*jt + 82P*itq*jt
+83(1-p*jt) . (l-q*jt) + 84X*jt
Equation (4) is called the performance
equation in the discrete model. The
parameter 82 should be positive because it
captures the effect of being above the
median for both performance and compensation
in the current year (region (A)
in Figure 1). The parameter 53 should also
be positive, since it captures the effect of
being below the median for both performance
and compensation in the current
year. The parameter 81 will generally be
positive for most accounting, economic,
and market performance measures. It is
included in the model because performance
measures, especially those based on
accounting data, are known to possess
positive serial correlation for a variety of
reasons unrelated to the compensation
system design (Foster 1986). The sign of
84 is unknown a priori, since the variable
x*ft is included as a control for sampling
design problems in the analysis sample.2
The Continuous Model
In order to calculate an estimate of the
1 This variable is introduced because the analysis
sample is not a random sample of companies from
the comparison group. Larger companies are more
apt to be in the sample. In a discrete model, inclusion
of this size variable adequately controls for the
selectivity bias created by this sampling plan.
2 It is important to note that equation (4) is
saturated in the p*.t and q*j effects as specified
because there are only four possible combinations of
outcomes for these variables and three independent
effects (81, 82, and 83) are included. The remaining
outcome (P*jt =0, q*jt = 1) is the reference point for
the contrasts.
This content downloaded from 147.251.55.59 on Fri, 06 Apr 2018 11:01:31 UTC
All use subject to http://about.jstor.org/terms
58-S JOHN ABOWD
magnitude of the effect of increasing the
sensitivity of managerial compensation to
performance, I specify the following system,
which also incorporates the maintained
hypothesis. The model is specified
in terms of the conditional expectations of
the estimated company * year effects in
compensation and performance next year,
given the appropriate controls.
The compensation equation for the
continuous model is:
(5) E[qit I pjt, Xjt] = 40 + PlPjt + 422Xjt
where Xft is total assets for company j in
year t, and the other variables are defined
above. The parameters Pk have the same
interpretation as in equation (3), so I have
not used new symbols. In particular, 4ij >
o is required for compensation to be
performance-based, on average.
The performance equation for the
continuous model is:
(6) E[p1t+1 | Pit, qt, Xjt] = bo + 51pjt
+ 82T+(pit-mt)qjt + 83T (Pit - mt)qjt
+ 84Xjt + 855ft
where mt is the median of pjt overj for year
t; the function T+ (z) = z if z > 0 and 0
otherwise; and the function T- (z) = z if z
< 0 and 0 otherwise. The interpretation of
the parameters is similar to the interpretation
in equation (4), so I have not changed
the symbols. In particular, if performance
is above the median, then increasing
compensation is associated with increasing
the sensitivity of pay to performance
(region (A) in Figure 1); therefore, 82
should be positive. If performance is
below the median, then decreasing compensation
is associated with increasing the
sensitivity of pay to performance (region
(B) in Figure 2); therefore, 83 should also
be positive (because T- (pjt - mt) < 0 in this
case).3
3 Notice that since equation (6) is not saturated by
the inclusion of 81, 82, and 83, it is possible for
current compensation to have an independent effect,
85, that is not modeled.
The Managerial Compensation and
Financial Data
The managerial compensation data used
in this study were derived from the annual
cash compensation survey of a major
compensation consulting firm. The data
cover approximately 75 top management
employees for the period from 1981 to
1986 for each of about 600 corporations.
The company, executive, and position are
all identified in the survey data. Therefore,
it is possible to follow both individuals
and positions across years within a
single company. All financial data used in
this study were derived from Standard
and Poor's COMPUSTAT data base
(1988). The data appendix contains a
detailed description of the methods used
to create the analysis file.
Two important selection rules were
applied to the companies in the original
survey to derive the analysis file. First, a
company must appear in the compensation
survey at least three years (not
necessarily consecutive) to have sufficient
data for my analysis. Second, I used only
publicly held U.S. companies for which
the COMPUSTAT financial data and the
respondent's self-reported financial data
matched exactly. The strenuous requirement
of an exact match on total assets was
imposed to guarantee that the financial
data used were always from the most
recently completed fiscal year prior to the
March 1 survey reference date and to
guarantee that the position of the managers
within the corporate hierarchy was
comparable across companies (see the data
appendix). Only about 225 companies and
99,200 executive years were used in each
of the basic statistical analyses, with slight
variations depending on the particular
analysis. Fewer executive years but the
same number of companies were available
for the analyses involving changes.
Since the data were originally collected
by a compensation consulting firm, it is
appropriate to discuss the implicit sampling
frame used to generate the survey
responses. The original data were collected
from client submissions to the
consulting firm. A human resource manThis
content downloaded from 147.251.55.59 on Fri, 06 Apr 2018 11:01:31 UTC
All use subject to http://about.jstor.org/terms
PERFORMANCE-BASED COMPENSATION 59-S
agement professional employed by the
respondent company completed the survey
form for each executive the company
wished to have appraised. Unlike normal
social science surveys, but consistent with
industry practice in compensation surveys,
the respondent company paid a fee to be
included in the survey. The company
controlled how often it participated and
which executive positions were submitted
in a given year. The consulting company
had an active client relationship with many
of the participating companies; the primary
product of the survey, however, was
an analysis of the competitive position of
the respondent company's managers with
respect to the managers of a comparison
group of companies. The consulting company
did not design most of the compensation
systems in the survey.
Because the managers and companies in
the sample are self-selected, it is important
to know how representative they are of
various populations. The companies in the
sample I analyzed, all of which had public
financial data, are primarily large U.S.
corporations. On average, they have total
assets 2.7 times greater than the average
New York Stock Exchange company followed
in the COMPUSTAT files (in
1986). For this reason, I included a
company size control based on total assets
in all analyses.
The compensation survey includes salary
and annual bonus. The annual bonus
was defined as any cash payment earned
during the previous twelve months that
was based exclusively on performance
during a single year. Cash bonuses for
meeting multiyear performance goals were
not included in the annual bonus. (The
information is actually collected in a
different survey.) Cash that was received
during the last twelve months but that was
earned during an earlier period (with
payment deferred) was not included in the
bonus. Hence, the bonus data really are
for annual performance. Long-term incentive
pay was not available. For this reason,
I tried to design the empirical analysis so
that it focuses on annual performance
criteria. To the extent that annual performance
influences long-term incentive pay,
I am not able to capture the effect of
performance on compensation. To the
extent that there is a substantial difference
between the performance-sensitivity of
annual pay and long-term incentive pay,
the annual pay for performance analysis is
inappropriate.
Consider next the dating of the financial
and compensation variables. In the
theoretical framework, current compensation
is based on current performance, that
is, current compensation is paid at the end
of the current period when current performance
can be observed. In the empirical
analysis, current performance is defined
as the financial data for the most recently
completed fiscal year prior to March 1 of
the survey year, the reference date for the
base salary. Future performance is defined
as the performance during the fiscal
year that includes March 1 of the survey
year. Current compensation is defined as
the base salary as of March 1 of the survey
year and the most recently awarded bonus
prior to March 1. I have made every effort
to ensure that the bonus used in the
statistical analysis was determined when
the results of the most recently concluded
fiscal year prior to March 1 were known.
The data appendix describes the method
for checking the dating of the financial
information vis-a'-vis the compensation
data.
I used four distinct financial performance
variables. Two of the performance
measures, after-tax return on assets and
after-tax return on equity, are conventional
accounting measures of asset profitability,
generically called return on investment.
Actual definitions of these ratios
differ greatly from one application to the
next. Since there is no commonly agreedupon
method for calculating the ratios, I
used the formulas in Bernstein (1983,
Chapter 19).
The numerator of after-tax return on
assets is net income plus the interest
expense, adjusted to an after-tax basis at
the marginal corporate tax rate. The
denominator is average total assets over
the fiscal year.
After-tax return on equity was defined
for common stock equity, adjusted for
This content downloaded from 147.251.55.59 on Fri, 06 Apr 2018 11:01:31 UTC
All use subject to http://about.jstor.org/terms
60-S JOHN ABOWD
unconsolidated minority interests. The
numerator of after-tax return on equity is
net income less income to minority interests
less preferred stock dividends paid.
The denominator is average common
stock equity over the fiscal year.
The third performance measure I used
is a measure of gross cash flow, net of
taxes, divided by an estimate of the
replacement cost of total assets. This ratio
is called after-tax gross economic return.
The numerator, operating income less
income taxes, corresponds approximately
to the after-tax cash flow into the business.
The denominator, an estimate of the
current replacement cost of total assets at
the beginning of the fiscal year, corresponds
approximately to the wealth tied
up in the business at the beginning of the
fiscal year. The data appendix contains a
detailed description of the calculation of
this variable.
The final performance measure is a
market measure-total shareholder return,
which is the calendar year holding
period return per share of common stock.
The numerator of total shareholder return
is dividends per share earned over
the calendar year plus the capital gain per
share between the end of last year and the
end of the current year. The denominator
is the price per share of common stock at
the end of the previous calendar year.
Stock prices and dividends per share were
adjusted to reflect the effects of stock
splits and stock dividends during the
calendar year.
Table 1 contains a summary of all
variables used in my statistical analyses.
The table shows that the average executive
in the sample earned $106,689 per year in
total cash compensation over the period
from 1981 to 1986. The executives were
employed by companies that had an
average of $3,334 million in total assets
and earned an average of 6.7% per year in
after-tax return on assets. The shareholders
of these companies earned an average
of 17.3% per year total return. The notes
to the table contain the short definitions of
all variables; the data appendix contains
long definitions.
Statistical Results
The first requirement of the empirical
analysis is to estimate the company year
effects in equation (1) for each of the
compensation variables used in the analysis.
Table 2 contains a summary of the
results for the four compensation measures-log
of total salary, percent increase
in total salary, log of base salary, and
bonus as a percent of base salary.4 The
table contains no surprises and is presented
to show that the adjustment to the
various compensation measures is consistent
with analysis of individual compensation
data from many sources.5
Table 3 presents the results of the
discrete model using the annual percentage
increase in total salary as the compensation
measure and using after-tax return
on assets (ROA), after-tax return on equity
(ROE), after-tax gross economic return
(ERET), and total shareholder return (TSR)
as the performance measures. The compensation
equation clearly shows that
compensation is performance-based with
respect to each of the performance measures,
on average.
The performance equation gives mixed
results for the two accounting measures
(ROA and ROE), indicating that increased
performance sensitivity is not always associated
with increased performance. The
coefficients on the "Current Performance
and Current Compensation Both Above
Median" variable for the two accounting
measures are both positive, although the
coefficient in the ROE equation is impre-
4Throughout the paper, the exact form of the
percent increase in total salary is 100 (log(total
salary current year) - log(total salary previous
year)); only consecutive years are used. The exact
form of the bonus-to-base ratio is 100 log(1 +
bonus/base). This form was chosen because it makes
the decomposition of the logarithm of total salary
into the logarithm of base salary plus the bonus/base
ratio exact, except for the multiplication by 100,
which facilitates comparison of statistical models
using the bonus-to-base ratio with models using the
percentage change in total salary.
5 The company year effects from the "Log of
Total Salary" and "Log of Base Salary" columns are
never used as dependent variables in a compensation
equation. They are used as control variables in the
continuous model.
This content downloaded from 147.251.55.59 on Fri, 06 Apr 2018 11:01:31 UTC
All use subject to http://about.jstor.org/terms
PERFORMANCE-BASED COMPENSATION 61-S
Table 1. Summary of the Managerial Compensation and Corporate Performance Data for All
Firms Used in the Analysis (1981 to 1986).
Definition Std. Sample Definition Std. Sample
of Variable Mean Dev. Size of Variable Mean Dev. Size
Individual Data
Base Salarya 85,599 61,513 99,219 Total Shareholder
Logarithm Returnt 17.3% 35.1% 1,114
of Base Salary 11.2 .5 99,219 Total Assets at
Bonus Paymentb 21,090 39,321 99,219 Beginning of
Total Salaryc 106,689 94,826 99,219 Yearu 3,334 6,814 1,117
Logarithm of Log of Total Assets
Total Salary 11.4 .6 99,219 (Beg. of Year) 7.2 1.3 1,117
Percentage Increase Proportion of Comin
Total Salaryd 9.2% 1.4% 60,227 panies with Adj.
Bonus as a Percent Average Increase
of Basee 16.0% 15.1% 99,219 Above Median .497 na 857
Years of Educationf 16.4 1.8 99,219 Proportion of ComYears
of Labor Force panies with Bonus
Experienceg 26.2 9.0 99,219 as a Percent of
Years at Employerh 14.7 10.4 99,219 Base Above
Percentage Job Median .498 na 1,107
Level P 1.2% na 99,219 Proportion of ComPercentage
Job panies with AfterLevel
2J 6.6% na 99,219 Tax Return on
Percentage Job Assets Above
Level 3k 19.9% na 99,219 Median .554 na 1,107
Percentage Job Proportion of ComLevel
41 30.6% na 99,219 panies with After-Tax
Return on Equity
Company Data Above Median .505 na 1,104
Average Adjusted Proportion of
Log Total Salarym 11.4 .3 863 Companies with
Average Adjusted After-Tax Gross
Increase in Economic ReTotal
Salaryn 8.8% 10.5% 863 turn Above
Average Adjusted Median .560 na 1,052
Log Base Salary' 11.2 .3 1,114 Proportion of ComAverage
Adjusted panies with Total
Bonus/Base Shareholder
(percent)P 15.5% 10.7% 1,114 Return Above
After-Tax Return Median .493 na 1,114
on Assetsq 6.7% 5.7% 1,107 Proportion of ComAfter-Tax
Return panies with Total
on Equityr 10.8% 13.7% 1,104 Assets (Beginning
After-Tax Gross of Year) Above
Economic Returns 11.4% 5.6% 1,052 Median .873 na 1,107
a Base Salary is the annual salary (exclusive of bonus and l
March 1 of the survey year.
b Bonus is the most recent payment (prior to March 1 of the survey year) determined by an annual review
cycle. Bonus payments determined on review cycles longer than one year are considered long-term incentive
compensation and are not included.
c Total salary is the sum of base salary and bonus.
d Percentage increase in total salary is 100 * (log(total salary as of March 1 of the survey year) - log(total
salary as of March 1 of the previous year)), for consecutive surveys only. The variable is only available when
same executive is surveyed in two consecutive years.
2 Bonus as a percent of base is 100 . log(1 + Bonus/Base).
f Years of education is imputed for survey year 1986 using the history of the executive, when available, or the
vaiue 16.4 if there is no history.
g Years of labor force experience is Current age - Years of Education - 5.
h Years at employer is the executive's actual tenure with the surveyed company.
(table notes continue)
This content downloaded from 147.251.55.59 on Fri, 06 Apr 2018 11:01:31 UTC
All use subject to http://about.jstor.org/terms
62-S JOHN ABOWD
'Job level 1 is the highest position in t
as reported on the survey.
Job level 2 reports to the CEO (usually President and Chief Operating Officer).
kJob level 3 reports to the level 2 position.
'Job level 4 reports to the level 3 position. All other positions are level 5 or below.
m Average adjusted log total salary is the estimated company year effect from the regression of log total
salary on years of education, labor force experience, labor force experience squared, years at employer,
indicators for job levels 1, 2, 3, and 4 (level 5 and above is the reference group), and unrestricted year within
company fixed effects. The estimated company * year effects were forced through the grand mean of log total
salary. See Table 2.
n Average adjusted increase in total salary is the estimated company * year effect from the regression of the
percentage increase in total salary on the variables listed in note m (except labor force experience squared). See
Table 2.
? Average adjusted log base salary is the estimated company * year effect from the regression of log base
salary on the variables listed in note m. See Table 2.
P Average adjusted bonus/base (percent) is the estimated company * year effect from the regression of
100.log(1 + bonus/base) on the variables listed in note m. See Table 2.
q After-tax return on assets: 100 (Net Income + Interest (1 - Tax Rate)) divided by (Beginning Total Assets
+ Ending Total Assets)/2.
'After-tax return on equity: 100 . (Net Income) divided by (Beginning Shareholder's Equity
Shareholder's Equity)/2.
s After-tax economic return is operating income less taxes as a percentage of beginning of period total assets,
valued at replacement cost: 100 (Operating Income - Taxes) divided by (Beginning Total Assets, valued at
replacement cost).
'Total shareholder return: 100 (Dividends per beginning share + Capital gain per beginning share)
divided by (Beginning price per share).
U Total assets at the beginning of the year is the book value of all assets at the end of the previous fiscal year
(in millions of dollars). In all cases, this value is for a fiscal year that ended prior to March 1 of the
compensation survey year.
Sources: (1.) All financial data are from Standard and Poor's COMPUSTAT service for fiscal years 1980 to
1986 based on the September 1988 annual industrial tape. Only the 2,423 New York Stock Exchange
companies available during this period were used for comparisons. (2.) All compensation data are from the
annual surveys of a major compensation consulting firm (1981 to 1986 surveys). Surveys were conducted in
March and April of the survey year.
cise. The coefficients on the "Current
Performance and Current Compensation
Both Below Median" variable for the two
accounting measures are both negative,
although both coefficients are imprecise.
The results are also mixed for total
shareholder return (TSR), although, given
the rapidity with which the stock market
moves, it is always possible that the effect
of the incentive compensation was capitalized
during the current year and not
during the next year. The performance
equation indicates a very substantial effect
of increased performance-sensitivity, in
the predicted positive direction, when the
performance measure is after-tax gross
economic return (ERET).
Table 4 presents a parallel analysis of a
discrete model using as the compensation
measure the bonus as a percent of base.
The results are not substantially different
from the results in Table 3. Compensation
is performance-related for all performance
measures. Only the after-tax gross
economic return shows evidence of a
performance improvement when performance-sensitivity
in the compensation equation
increases.
Table 5 shows the results for a continuous
model in which the compensation
measure is the percentage increase in total
salary. All compensation equations indicate
that pay is performance-related, on
average. The performance equations for
after-tax return on assets and after-tax
gross economic return give mixed evidence
for an effect of increased performance-sensitivity
on future performance.
There is no evidence that sensitivity to
after-tax return on equity affects future
performance, but substantial evidence that
total shareholder return has such an
effect.
In the total shareholder return performance
equation in Table 5, the estimated
82 is .0109 (?.0034) and the estimated 83 is
This content downloaded from 147.251.55.59 on Fri, 06 Apr 2018 11:01:31 UTC
All use subject to http://about.jstor.org/terms
PERFORMANCE-BASED COMPENSATION 63-S
Table 2. Summary of the Regression Models Used to Adjust the Various Compensation
Measures Defined at the Company Level Estimated by Least Squares with Fixed
Company* Year Effects.a
Percent
Log of Increase Log of Bonus as
Total in Total Base a Percent
Independent Variableb Salary Salary Salary of Base
Years of Education .0429 -.2260 .0377 .5136
(.0006) (.0278) (.0006) (.0209)
Years at Employer .0033 -.0356 .0016 .1620
(.0001) (.0059) (.0001) (.0044)
Years of Experience .0379 -.1774 .0325 .5323
(.0006) (.0069) (.0005) (.0195)
Years of Experience Squared -.479 10-3 na -.397-10-3 -.0082
(.116- 10-4) na (.965- 10 -5) (.0004)
Job Level 1 1.9145 3.60 1.7126 20.18
(.0010) (.3765) (.0083) (.3109)
Job Level 2 1.0478 3.98 .9014 14.65
(.0048) (.1930) (.0040) (.1511)
Job Level 3 .5505 2.34 .4647 8.58
(.0033) (.1352) (.0027) (.1020)
Job Level 4 .2737 1.28 .2284 4.54
(.0027) (.1166) (.0023) (.0857)
Standard Error of Equation .333 10.86% .277 10.40%
R 2 .697 .395 .713 .528
Degrees of Freedom for
Company Year Effects 1,151 896 1,151 1,151
Sample Size 99,219 60,227 99,219 99,219
a Coefficients are shown with standard e
b Table 1 contains variable definitions and
Source: All data are from the annual surveys of a major compensation consulting firm (1981 to 1986 surveys).
.0095 (?.0058). These coefficients translate
into rather substantial performance
effects. A one standard deviation increase
in the raise (an extra 10.5% of last year's
salary) delivered when current total shareholder
return is one standard deviation
above the median (3,510 basis points
above the median) yields a 400 basis point
expected increase in next year's total
shareholder return. The estimated effect
is only 11% of the standard deviation of
total shareholder return, which would be
difficult, but not impossible, to detect in a
sample of total shareholder returns for
which there was only compensation announcement
information. The expected
effect of delivering the same incremental
raise when total shareholder return is
three standard deviations above average is
1,200 basis points of additional total
shareholder return, which is about onethird
of the standard deviation. These
results suggest that general managerial
compensation policy may affect the stock
market value of a company on a yearto-year
basis. This conclusion is surprising
in view of the timing difficulties associated
with measuring the effects of managerial
and other compensation policy changes on
stock returns (see Abowd, Milkovich, and
Hannon, this issue); it is not inconsistent,
however, with efficient capital markets. If
the compensation policies are announced
after the current fiscal year results, the
reported effects could legitimately be
associated with the performance-sensitivity
of compensation.
Table 6 reports the results for a
continuous model in which the compensation
measure is the bonus as a percent of
base. The compensation equations indicate
that pay is performance-related, on
This content downloaded from 147.251.55.59 on Fri, 06 Apr 2018 11:01:31 UTC
All use subject to http://about.jstor.org/terms
64-S JOHN ABOWD
Table 3. Summary of the Statistical Analysis of the Discrete Model Using the Annual Percentage
Increase in Total Salary as the Compensation Measure and a
Variety of Performance Measures.
(Standard Errors in Parentheses)a
Performance Measure ROAb ROEc EREVT TSRe
Compensation Equationf
Performance Measure .614 .701 .777 .908
(.140) (.139) (.145) (.140)
Total Assets at .598 .579 .655 .488
Beginning of Year (.219) (.220) (.234) (.222)
Intercept -.876 -.866 -1.027 -.866
(.226) (.222) (.243) (.218)
Performance Equationg
Current Performance 1.704 1.857 2.931 .534
Measure (.233) (.238) (.287) (.221)
Current Performance &
Current Compensation .437 .218 .624 - .098
Both Above Median (.208) (.217) (.266) (.203)
Current Performance &
Current Compensation -.404 -.479 .365 .277
Both Below Median (.245) (.241) (.273) (.196)
Total Assets at -.375 - .360 -.593 .247
Beginning of Year (.248) (.251) (.304) (.214)
Intercept - .665 - .723 - .994 - .512
(.292) (.291) (.355) (.248)
Number of Observations 857 853 814 863
Number of Firms 227 228 214 228
a The reported results are maximum likelihood estimates of the logistic regression coefficients for an
equation that estimates the conditional probability of a one for the dependent variable, given the variables
shown in the rows of the table. Summary statistics are reported in Table 1.
b ROA is after-tax return on assets, defined in Table 1.
c ROE is after-tax return on equity, defined in Table 1.
d ERET is after-tax gross economic return (cash flow) as a percentage of beginning of period total
valued at replacement cost, defined in Table 1.
e TSR is total shareholder return, defined in Table 1.
f The dependent variable in the compensation equation is based on the average adjusted value of
100 * (log(total salary year t) - log(total salary year t-1)), called the average adjusted increase in total salary.
These are the company * year effects implied by the "percent increase in total salary" column of Table 2. If the
average adjusted increase in total salary for a particular company exceeds the median average adjusted increase
in total salary for all the firms in the sample, then the compensation measure is one for that company;
otherwise, zero. The performance measure on the right-hand side of the equation is described in note g. The
variable total assets at beginning of year equal one if total assets for the company exceed the median of total
assets for the New York Stock Exchange companies listed on COMPUSTAT; otherwise, zero.
g The dependent variable is a performance measure based on the next fiscal year's value of the performance
variable indicated by the column heading relative to the median of the New York Stock Exchange companies
listed on COMPUSTAT for that year. If the performance variable for a particular company exceeds the annual
median performance on the NYSE, then the performance measure equals one for that company; otherwise,
zero. Current performance and current compensation both above median equals one when the performance
measure is above the median for NYSE companies in the current fiscal year and the compensation measure is
above the median for the compensation sample companies for the current year; otherwise, zero. Current
performance and current compensation both below median is defined similarly when both measures are below
the appropriate medians.
Sources: (1.) Annual median ROA, ROE, ERET, and TSR, Total Assets at Beginning of Year, and all other
financial data are from the 2,423 New York Stock Exchange companies in Standard and Poor's COMPUSTAT
for the fiscal years 1980 to 1986 (September 1988 annual industrial tape). (2.) All compensation data are from
the annual surveys of a major compensation consulting firm (1981 to 1986 surveys).
This content downloaded from 147.251.55.59 on Fri, 06 Apr 2018 11:01:31 UTC
All use subject to http://about.jstor.org/terms
PERFORMANCE-BASED COMPENSATION 65-S
Table 4. Summary of the Statistical Analysis of the Discrete Model Using the Bonus as a
Percentage of Base Salary as the Compensation Measure and a
Variety of Performance Measures.
(Standard Errors in Parentheses)a
Performance Measure ROAb ROEC ERE' TSRe
Compensation Equationf
Performance Measure .817 .938 .882 .489
(.126) (.126) (.130) (.126)
Total Assets 1.191 1.185 1.250 1.090
(.203) (.204) (.216) (.200)
Intercept -1.518 -1.510 -1.611 -1.186
(.212) (.208) (.226) (.198)
Performance Equationg
Current Performance 1.796 1.993 2.753 .065
Measure (.214) (.221) (.252) (.182)
Current Performance &
Current Compensation .045 .164 .533 .029
Above Average (.191) (.199) (.228) (.173)
Current Performance &
Current Compensation - .470 - .389 .012 - .071
Below Average (.210) (.212) (.245) (.172)
Current Total Assets -.487 -.351 -.654 .255
(.221) (.220) (.262) (.185)
Intercept - .236 - .652 - .805 - .272
(.265) (.264) (.321) (.215)
Number of Observations 1107 1104 1052 1114
Number of Firms 228 227 216 229
a The reported results are maximum likelihood estimates of the logistic regression coefficients for an
equation that estimates the conditional probability of a one for the dependent variable, given the variables
shown in the rows of the table. Summary statistics are reported in Table 1.
b ROA is after-tax return on assets, defined in Table 1.
c ROE is after-tax return on equity, defined in Table 1.
d ERET is after-tax gross economic return (cash flow) as a percentage of beginning of period total assets
valued at replacement cost, defined in Table 1.
e TSR is total shareholder return, defined in Table 1.
f The dependent variable in the compensation equation is based on the average adjusted value of 100 * log
+ Bonus/Base), called bonus as a percent of base. These are the company -year effects from the "Bonus as a
Percent of Base" column of Table 2. If the adjusted average bonus as a percent of base for a particular company
exceeds the median adjusted average bonus as a percent of base for all the companies in the sample, then the
compensation measure is one for that company; otherwise, zero. The performance measure on the right-hand
side of the equation is described in note f. The variable total assets at beginning of year equal one if total assets
for the company exceed the median of total assets for the New York Stock Exchange companies listed on
COMPUSTAT; otherwise, zero.
g The dependent variable is a performance measure based on the next fiscal year's value of the performance
variable indicated by the column heading relative to the median of the New York Stock Exchange companies
listed on COMPUSTAT for that year. If the performance variable for a particular company exceeds the annual
median performance on the NYSE, then the performance measure equals one for that company; otherwise,
zero. Current performance and current compensation both above median equals one when the performance
measure is above the median for NYSE companies in the current fiscal year and the compensation measure is
above the median for the compensation sample companies for the current year; otherwise, zero. Current
performance and current compensation both below median is defined similarly when both measures are below
the relevant medians.
Sources: (1.) Annual median ROA, ROE, ERET, and TSR, Total Assets at Beginning of Year, and all other
financial data are from the 2,423 New York Stock Exchange companies in Standard and Poor's COMPUSTAT
for the fiscal years 1980 to 1986 (September 1988 annual industrial tape). (2.) All compensation data are from
the annual surveys of a major compensation consulting firm (1981 to 1986 surveys).
This content downloaded from 147.251.55.59 on Fri, 06 Apr 2018 11:01:31 UTC
All use subject to http://about.jstor.org/terms
66-S JOHN ABOWD
Table 5. Summary of the Statistical Analysis of the Continuous Model Using the Annual
Percentage Increase in Total Salary as the Compensation Measure and a
Variety of Performance Measures.
(Standard Errors in Parentheses)a
Performance Measure RoAb ROEc EREVd TSRe
Compensation Equationf
Performance Measure .353 .182 .306 .048
(.061) (.028) (.066) (.010)
Total Assets .654 .518 .601 .561
Beginning of Year (.267) (.265) (.290) (.267)
Intercept 1.547 2.909 .891 3.898
(2.061) (1.985) (2.335) (1.992)
Standard Error of Eqn. 10.367 10.261 10.513 10.388
R 2 .042 .051 .029 .032
Performance Equationg
Current Performance .463 .551 .685 -.203
Measure (.036) (.042) (.033) (.044)
Current Adjusted Log 1.046 3.854 1.562 -7.109
of Total Salary (.866) (2.008) (.648) (5.363)
(Current Performance
Above Median, if > 0) .0066 -.0020 .0043 .0109
* Current Increaseh (.0031) (.0024) (.0022) (.0034)
(Current Performance
below Median, if < 0) - .0140 .0012 - .0065 .0095
* Current Increase' (.0043) (.0037) (.0042) (.0058)
Total Assets -.373 -.772 -.404 -2.219
Beginning of Year (.209) (.478) (.160) (1.289)
Intercept -6.406 - 34.761 -11.727 87.484
(8.799) (20.388) (6.509) (54.682)
Standard Error of Eqn. 5.354 12.597 3.947 34.694
R 2 .228 .235 .516 .028
Number of Observations 857 853 814 863
Number of Firms 227 228 214 228
a The reported results are least squa
the conditional expectation of the dependent variable, given the variables in the rows of the table. Summary
statistics are reported in Table 1.
b ROA is after-tax return on assets, defined in Table 1.
c ROE is after-tax return on equity, defined in Table 1.
d ERET is after-tax gross economic return (cash flow) as a percentage of beginning of period total assets,
valued at replacement cost, defined in Table 1.
e TSR is total shareholder return, defined in Table 1.
f The compensation measure is the average adjusted increase in total salary based on the company year
effects implied by the "percent increase in total salary" column of Table 2.
g The performance measure is the value of the variable indicated by the column heading for the next fiscal
year.
h The current performance above median compensation interaction is (the value of the performance variable
minus the annual median of New York Stock Exchange companies for this performance variable, if this is
positive; zero, otherwise) times the adjusted average increase in total salary.
' The current performance below median . compensation interaction is (the value of the performance variab
minus the annual median of New York Stock Exchange companies for this performance variable, if this is
negative; zero, otherwise) times the adjusted average increase in total salary.
Sources: (1.) Annual median ROA, ROE, ERET, and TSR, Total Assets at Beginning of Year, and all other
financial data are from the 2,423 New York Stock Exchange companies in Standard and Poor's COMPUSTAT
for the fiscal years 1980 to 1986 (September 1988 annual industrial tape). (2.) All compensation data are from
the annual surveys of a major compensation consulting firm (1981 to 1986 surveys).
This content downloaded from 147.251.55.59 on Fri, 06 Apr 2018 11:01:31 UTC
All use subject to http://about.jstor.org/terms
PERFORMANCE-BASED COMPENSATION 67-S
Table 6. Summary of the Statistical Analysis of the Continuous Model Using the Bonus as a
Percentage of Base Salary as the Compensation Measure and a
Variety of Performance Measures.
(Standard Errors in Parentheses)a
Performance Measure RoAb ROEc EREt TSRe
Compensation Equationf
Performance Measure .540 .304 .577 .029
(.051) (.024) (.053) (.008)
Total Assets 2.303 2.138 2.433 2.179
Beginning of Year (.221) (.217) (.240) (.233)
Intercept -5.354 - 3.762 -8.848 - .901
(1.698) (1.620) (1.937) (1.730)
Standard Error of Eqn. 9.660 9.533 9.871 10.238
R 2 .154 .195 .158 .081
Performance Equationg
Current Performance .563 .518 .635 - .201
Measure (.039) (.046) (.035) (.040)
Current Adjusted Log - 1.321 - 1.364 .732 -.298
of Base Salary (.805) (1.950) (.638) (5.523)
(Current Performance
Above Median, if > 0) .0075 -.0020 .0050 .0005
Current Bonus/Baseh (.0025) (.0025) (.0017) (.0028)
(Current Performance
Below Median, if < 0) -.0217 .0048 .0031 .0187
Current Bonus/Basei (.0030) (.0051) (.0036) (.0047)
Total Assets -.095 -.181 -.301 .160
Beginning of Year (.165) (.402) (.136) (1.143)
Intercept 17.489 20.934 -2.265 25.907
(8.197) (19.842) (6.482) (56.158)
Standard Error of Eqn. 4.918 11.961 3.834 34.526
R 2 .301 .241 .538 .036
Number of Observations 1,107 1,104 1,052 1,114
Number of Firms 228 229 216 228
a The reported results are least squares
the conditional expectation of the dependent variable, given the variables in the rows of the table. Summary
statistics are reported in Table 1.
b ROA is after-tax return on assets, defined in Table 1.
c ROE is after-tax return on equity, defined in Table 1.
d ERET is after-tax gross economic return (cash flow) as a percentage of beginning of period total assets,
valued at replacement cost, defined in Table 1.
e TSR is total shareholder return, defined in Table 1.
f The compensation measure is the average adjusted bonus as a percent of base implied by the "Bonus as a
Percent of Base" column in Table 2.
g The performance measure is the value of the variable indicated by the column heading.
h The current performance above median . compensation interaction is (the value of the performance variable
minus the annual median of New York Stock Exchange companies for this performance variable, if this is
positive; zero, otherwise) times the adjusted average bonus as a percent of base.
'The current performance below median . compensation interaction is (the value of the performance variable
minus the annual median of New York Stock Exchange companies for this performance variable, if this is
negative; zero, otherwise) times the adjusted average bonus as a percent of base.
Sources: (1.) Annual median ROA, ROE, ERET, and TSR, Total Assets at Beginning of Year, and all other
financial data are from the 2,423 New York Stock Exchange companies in Standard and Poor's COMPUSTAT
for the fiscal years 1980 to 1986 (September 1988 annual industrial tape). (2.) All compensation data are from
the annual surveys of a major compensation consulting firm (1981 to 1986 surveys).
This content downloaded from 147.251.55.59 on Fri, 06 Apr 2018 11:01:31 UTC
All use subject to http://about.jstor.org/terms
68-S JOHN ABOWD
average. The performance equation for
after-tax return on assets gives mixed
evidence for an effect of increased performance-sensitivity
on future performance.
There is no statistical evidence that increasing
performance-sensitivity for after-tax
return on equity affects future performance.
The evidence that after-tax gross
economic return and total shareholder
return have an effect, however, is substan-
tial.
I will illustrate the magnitude for the
gross economic return. The estimated 82 is
.0050 (?+.0017) and the estimated 83 is
.0031 (?.0036). Only 82 is statistically
precise. The effect of increasing the
bonus-to-base ratio by one standard deviation
(10.7% of the base salary) when
after-tax gross economic return is one
standard deviation above the median (560
basis points) is an expected 30 basis points
of economic return. The expected effect is
5% of the standard deviation of economic
return. The expected effect from the
same change in the bonus-to-base ratio
when economic return is three standard
deviations above the median (1680 basis
points) is 90 basis points, which is 16%
of the standard deviation of after-tax
gross economic return. The estimated
magnitudes of the effects of increasing the
performance-sensitivity based upon the
economic return measure are slightly
smaller than the estimated effects for total
shareholder return after standardizing.
Given the variability of the estimates,
however, the results are basically compara-
ble.
Conclusions
I have specified an internally consistent
framework for measuring the degree of
performance-sensitivity in a compensation
system and assessing the validity of the
performance-base. The method shows
clearly that measuring the effects of a
change in the extent to which compensation
is related to performance requires an
analysis of the effects of interactions
between current performance and current
compensation on subsequent perfor-
mance.
The estimated models produce some
weak results and some fairly strong results.
It is perhaps surprising that the
accounting performance measures did not
fare as well as the economic measure or
the market measure. Accounting-based
performance measures are widely used in
businesses as a basis for compensation.
There are two reasonable explanations for
the weak results I obtained. First, most
corporate performance plans explicitly
using accounting data are multiyear plans,
which have been excluded from my data.
Second, since after-tax return on assets,
after-tax return on equity, and after-tax
gross economic return are all correlated
(in all cases > .6), using any one of these
measures as the basis for pay could
produce the desired results on after-tax
gross economic return, which is a better
measure of the profitability of the assets
than either of the usual accounting ratios.
This study suggests that pay-forperformance
systems based on after-tax
gross economic return and total shareholder
return may be effective, since I
find evidence that increasing the sensitivity
of compensation to either of these
measures may be associated with better
performance on that measure in the
future. None of the estimated equations
approached the degree of precision that
would warrant uncritical adoption of the
plans under study. The results do lend
credence, however, to claims that benefits
can be gained by increasing the payfor-performance
component of managerial
compensation.
Data Appendix
This appendix describes the sources and methods
used to assemble the managerial compensation and
corporate performance data used in this study. The
appendix discusses the variables derived from the
compensation surveys, Standard and Poor's COMPUSTAT
database service, the U.S. Department of
This content downloaded from 147.251.55.59 on Fri, 06 Apr 2018 11:01:31 UTC
All use subject to http://about.jstor.org/terms
PERFORMANCE-BASED COMPENSATION 69-S
Commerce Bureau of Economic Analysis time series,
and miscellaneous additional sources. To improve
the readability of this appendix I have not used
acronyms for the variables. Instead, variable names
are set in italics when first defined and capitalized
throughout.
Compensation Variables
The managerial compensation data were developed
using the cash compensation survey of a major
compensation consulting company. The survey collects
data for both the corporation and the individual
manager. Data for individual managers are identified
by company, person, and year. Thus, any particular
executive who appears in the survey more than once
can be followed over time. The respondent company
decides how many executives to include and how
often to participate. For this reason, sample mobility
of the managers does not reflect career mobility.
Exits from the sample do not imply either separations
or promotions. Data on individual managers
are available for survey years 1981 to 1986, inclusive.
Corporate data are identified by company and year,
making it possible to follow companies over time.
Variables from the Compensation
Survey Individual Data
Annual Base Salary: Salary (exclusive of long- and
short-term bonuses) as of March 1 of the survey year.
Bonus: The dollar amount of any short-term
incentive granted for the latest bonus period (prior
to March 1 of the survey year). The figure includes
any incentive awards based on one period's performance
regardless of whether the cash was paid in full
or deferred (completely or partially). Excluded from
the figure are cash bonuses that are dependent on
fulfillment of some future or longer-term organizational
performance objectives; cash payments paid
during the previous year for performance during an
earlier period; and sales commissions. (This definition
is a paraphrase of the survey instructions.)
Job Level: The position reporting level. The Chief
Executive Officer is reporting level 1. Reporting level
2 reports directly to the CEO. For managers in
divisions or subsidiaries, the divisional president or
general manager cannot have a reporting level
higher than 2. This definition is supposed to
guarantee that only the corporate-level CEO has
reporting level 1.
Years of Education, Birthday, and Date of Hire are
defined in the conventional manner.
Variables from the Corporate-Level
Compensation Data
Company Name: The name of the business participating
in the survey. This variable was used to build
the link to COMPUSTAT data.
Assets: Corporate total assets as reported by the
respondent to the survey (usually a member of the
corporate-level human resource management staff).
This variable was used to verify that the company
was participating in the survey at the corporate level.
Respondent companies from subsidiaries, divisions,
and separate business units of a corporation could
elect to participate as if they were stand-alone
companies. In this case, the variable Assets would
contain total assets for the relevant business unit, an
amount less than corporate total assets. Such companies
were excluded from the present study. The
Assets variable was also used to verify that the
relevant fiscal year closed prior to March 1 of the
survey year. Only companies with an exact match
between the Assets variable in the compensation file
and the Total Assets variable from COMPUSTAT
were included in the study. This procedure ensured
that the COMPUSTAT fiscal year information dated
1980 was available to make the compensation
decisions reported in the 1981 compensation data,
and so forth for the succeeding years.
Shareholders' Equity: Corporate common equity as
reported by the survey respondent. This variable was
used to check the match to COMPUSTAT based on
the company name and assets. (Definitions of
shareholders' equity are complicated by the treatment
of preferred stock, so an exact match was not
required.)
COMPUSTAT Variables
The descriptions below are based on COMPUSTAT
(Standard and Poor's 1988) documentation
of standard financial accounting concepts. The item
numbers refer to the variable locations on the annual
industrial files.
Net Sales: Sales revenue net of discounts and
returns (COMPUSTAT Item 012).
COGS: Cost of goods sold (COMPUSTAT Item
041).
Selling and Administrative Expenses: Selling, overhead,
and general administrative expenses (COMPUSTAT
Item 012 less Item 041 less Item 013).
Operating Income: Net Sales - COGS - Selling and
Administrative Expenses.
Interest Expense: Gross interest expense (COMPUSTAT
Item 015).
Income Taxes: Total income taxes (COMPUSTAT
Item 016).
Net Income: Income before extraordinary items
(COMPUSTAT Item 018) plus gain (or loss) on
extraordinary items (COMPUSTAT Item 048).
Minority Interest Income: The part of net income due
to unconsolidated minority interests in the company
(COMPUSTAT Item 049).
Preferred Stock Dividends: Dividends paid to holders
of preferred stock (COMPUSTAT Item 019).
Total Assets: End of fiscal year book value of all
assets on the balance sheet. This value must equal the
sum of all liabilities and shareholders' equity (COMPUSTAT
Item 006).
Inventory: Asset consisting of the value of raw,
intermediate, and finished goods inventory (COMPUSTAT
Item 003). The method of book valuation
is discussed below.
Inventory Valuation Method: Last-in-first-out (LIFO)
is distinguished from all other methods; other
methods are treated as first-in-first-out (FIFO). The
inventory adjustment method is based on COMPUSTAT
Item 059.
Current Assets: Cash and short-term investments,
This content downloaded from 147.251.55.59 on Fri, 06 Apr 2018 11:01:31 UTC
All use subject to http://about.jstor.org/terms
70-S JOHN ABOWD
accounts receivable, inventory, and other short-term
assets (COMPUSTAT Item 004).
Gross Property, Plant, and Equipment: Asset consisting
of undepreciated historical cost of property,
plant, and equipment (COMPUSTAT 007).
Net Property, Plant, and Equipment: Asset consisting
of depreciated historical cost of property, plant, and
equipment (COMPUSTAT 008).
Other Long-Term Assets: Total Assets less current
assets less net property, plant, and equipment
(COMPUSTAT Item 006 less Item 004 less Item
008).
Current Liabilities: Short-term debt, accounts payable,
and other short-term liabilities (COMPUSTAT
Item 005).
Current Debt: Short-term debt component of
current liabilities (COMPUSTAT Item 034).
Common Equity: End of fiscal year book value of
common stock equity (COMPUSTAT Item 060).
Gross Investment: Gross spending on new property,
plant, and equipment (COMPUSTAT Item 030).
Dispositions: Proceeds from the sale of property,
plant, and equipment (COMPUSTAT Item 107).
Cumulative Adjustment Factor: Restates common
stock data so that all previous fiscal years are on the
same basis as the most recent fiscal year in the file,
usually 1987 (COMPUSTAT Item 027).
Bureau of Economic Analysis Variables
The variables described below are based on the
National Income and Product Accounts (NIPA,
Survey of Current Business, monthly). NIPA variables
are referenced by their standard table numbers.
NIPA variables extracted from CITIBASE (Citicorp
Database Services 1978) are referenced by the
CITIBASE name as well. The Bureau of Economic
Analysis (BEA) in the Department of Commerce
maintains estimates of fixed reproducible tangible
wealth. Variables from this BEA data base are
referenced by their table numbers in Fixed Reproducible
Tangible Wealth in the United States, 1925-85
(called FRTW below). (See also Musgrave 1986.)
Equipment Proportion of Industry Fixed Nonresidential
Investment: Derived from the ratio of Fixed Nonresidential
Private Capital, Equipment Investment, in
millions of current dollars, to Fixed Nonresidential
Private Capital, Total Investment, in millions of
current dollars; FRTW Table B 1, by two-digit
Standard Industrial Classification. This variable was
supplied by Shapiro (See Brainard et al. 1988) as
extracted from the BEA Wealth tape, which contains
the FRTW data.
Structure Proportion of Industry Fixed Nonresidential
Investment: Derived from the ratio of Fixed Nonresidential
Private Capital, Structure Investment, in
millions of current dollars, to Fixed Nonresidential
Private Capital, Total Investment, in millions of
current dollars; FRTW Table B 1, by two-digit
Standard Industrial Classification. This variable was
also supplied by Shapiro. Equipment and structure
proportions of industry fixed nonresidential investment
sum to one.
Industry Implicit Price Deflator for Plant Investment:
Derived as the ratio of Fixed Nonresidential Private
Capital, Plant Investment (millions of current dollars)
to Fixed Nonresidential Private Capital, Plant
Investment (millions of 1982 dollars); FRTW Table
B1, by two-digit Standard Industrial Classification.
This variable was also supplied by Shapiro.
Industry Implicit Price Deflator for Equipment Investment:
Derived as the ratio of Fixed Nonresidential Private
Capital, Equipment Investment (millions of current
dollars) to Fixed Nonresidential Private Capital,
Equipment Investment (millions of 1982 dollars);
FRTW Table B1, by two-digit Standard Industrial
Classification. This variable was also supplied by
Shapiro.
Industry Implicit Price Deflator: Derived as the ratio
of GNP by Industry, billions of current dollars
(NIPA Table 6.1) to GNP by Industry, billions of
1982 dollars (NIPA Table 6.2), by two-digit SIC. This
variable was supplied by Shapiro as extracted from
the BEA NIPA tape.
Implicit Price Deflator for Fixed Nonresidential Investments
in Structure: Derived as the ratio of Fixed
Investment by Type, Structures, billions of current
dollars (NIPA Table 5.12) to Fixed Investment by
Type, Structures, billions of 1982 dollars (NIPA
Table 5.13); extracted from CITIBASE as GDIS and
converted to annual average.
Miscellaneous Variables
Tax Rate: U.S. Federal marginal corporate tax rate
(.46 for 1980 to 1986, .40 for 1987, and .34
thereafter).
Derived Variables Used in the Compensation
and Performance Analysis
The estimate of the replacement cost of total assets
was developed using the methods of Brainard et al.
(1988) and Hall et al. (1988). The flow of the
calculation is described here. Programming is available
from the author. Abowd and Tracy (1989)
contains a detailed discussion of the methodology.
The result of the calculation is an estimate of the cost
of reproducing a company's current total assets
(primarily property, plant, equipment, and inventory)
without purchasing the company outright (by
purchasing all of its outstanding stocks and bonds).
For each company, a complete history was assembled
from annual industrial data, research data, back
data, and research back data files supplied by
COMPUSTAT. Some company histories begin in
1950, others in 1960, and still others in 1968 (the
start date of the September 1988 annual industrial
file for most companies). The company histories
were used to impute a series of structure, equipment,
inventory, and other investments that were converted
from historical cost to current replacement
cost, depreciated according to economic life, and
summed over the economic life of each asset to
produce an estimate of the replacement cost of the
asset.
Plant investments were assumed to have an
economic life of 26 years; equipment investments
were assumed to have an economic life of 14 years.
The first five years of data for each company were
used to estimate the growth rate of the asset Gross
This content downloaded from 147.251.55.59 on Fri, 06 Apr 2018 11:01:31 UTC
All use subject to http://about.jstor.org/terms
PERFORMANCE-BASED COMPENSATION 71-S
Property, Plant and Equipment (subject to a minimum
of zero and a maximum of 10% per annum).
This growth rate was used to impute a history of
gross structure and equipment investments prior to
the initial data year the sum of which was exactly
equal to the earliest value of Gross Property, Plant
and Equipment. For each subsequent year, gross
investment in property, plant, and equipment was set
equal to Gross Investment less Dispositions. Current
and (imputed) historical cost gross investments were
then adjusted (by imputing a writeoff or addition, as
appropriate) so that the historical cost investment
series always summed to the current Gross Property,
Plant and Equipment asset.
The vintage history of gross investments was
divided into structure and equipment using the
Structure as a Proportion of Industry Investment
and Equipment as a Proportion of Industry Investment
series for the appropriate years. The historical
cost plant investments were then converted to
current replacement cost by multiplying by the ratio
of the current value of the Industry Implicit Price
Deflator for Structure Investments to the appropriate
historical value of the same series. Historical cost
equipment investments were similarly converted to
current replacement cost using the Industry Implicit
Price Deflator for Equipment Investment. Current
replacement cost investments were depreciated using
straight line depreciation over the economic lives
assumed for plant and equipment, respectively. The
sum of the current replacement cost structure
investments for the current and 25 preceding years
plus the sum of current replacement cost equipment
investments for the current and 13 preceding years is
Gross Property, Plant, and Equipment, Replacement Cost.
Net Property, Plant, and Equipment, Replacement Cost
was calculated according to the same formula as
Gross Property, Plant, and Equipment, Replacement
Cost using the depreciated replacement cost estimates
for structure and equipment investments.
Inventory was only adjusted to the extent that
LIFO accounting was used. Up to three different
inventory valuation methods were allowed. The
proportion of Inventory valued using LIFO was
estimated from the COMPUSTAT Inventory Method
variable. For Non-LIFO Inventory the replacement
cost and historical cost are equal. LIFO Inventory
was converted to replacement cost by multiplying the
LIFO proportion of last year's Inventory, valued at
replacement cost, by the ratio of the current value of
the Industry Implicit Price Deflator to last year's
Industry Implicit Price Deflator and adding the
change in historical cost LIFO Inventory between the
current and previous fiscal years. The sum of LIFO
and non-LIFO Inventory, valued at replacement
cost, is the series Inventory, Replacement Cost. The
value of replacement cost and historical cost Inventory
are equal for the first year a company appears in
the COMPUSTAT data regardless of Inventory
Method.
Other Assets were converted to replacement cost
by multiplying last year's Other Assets, valued at
replacement cost, by the ratio of the current value of
the Implicit Price Deflator for Fixed Nonresidential
Investments in Structure to last year's Implicit Price
Deflator for Fixed Nonresidential Investments in
Structure and adding the change in historical cost
Other Assets between the current and previous fiscal
years. The resulting series is Other Assets, Replacement
Cost. The value of replacement cost and historical
cost Other Assets are equal for the first year a
company appears in the COMPUSTAT data.
Replacement Cost of Total Assets: Net Property, Plant,
and Equipment, Replacement Cost + Inventory,
Replacement Cost + Other Assets, Replacement
Cost + (Current Assets - Inventory) - (Current
Liabilities - Current Debt).
After-Tax Return on Assets: Ratio defined as 100 (Net
Income + Interest Expense (1 -Tax Rate))/((Total
Assets + Total Assets previous fiscal year end)/2).
After-Tax Return on Equity: Ratio defined as
100 . (Net Income - Minority Interest Income Preferred
Dividends Paid)/((Common Equity +
Common Equity previous fiscal year end)/2).
After-Tax Gross Economic Return: Ratio defined as
100 (Operating Income - Income Taxes)/Replacement
Cost of Total Assets previous fiscal year end.
Total Shareholder Return: Ratio defined as
100 . ((Dividends per Common Share/Cumulative
Adjustment Factor) + (Common Stock Price/
Cumulative Adjustment Factor) - (Common Stock
Price previous calendar year end/Cumulative Adjustment
Factor previous calendar year end))/((Common
Stock Price previous calendar year end/Cumulative
Adjustment Factor previous calendar year end)).
A thorough discussion of accounting measures of
profitability can be found in Bernstein (1983,
Chapter 19). Some of the pitfalls are discussed in
Foster (1986, Chapter 3). The definitions of after-tax
return on assets and after-tax return on equity are
from Bernstein and appear to conform to current
accounting practice.
REFERENCES
Abowd, John M., George T. Milkovich, and John M.
Hannon. 1990. "The Effects of Human Resource
Management Decisions on Shareholder Value."
Industrial and Labor Relations Review, this issue.
Abowd, John M., and Joseph S. Tracy. 1989.
"Unions and Market Power." Unpublished working
paper, Cornell University.
Antle, Rick, and Abbie Smith. 1986. "An Empirical
Investigation of the Relative Performance Evaluation
of Corporate Executives." Journal of Accounting
Research, Vol. 24, No. 1 (Spring), pp. 1-39.
Baker, George P., Michael C. Jensen, and Kevin J.
Murphy. 1988. "Compensation and Incentives:
Practice vs. Theory." Journal of Finance, Vol. 43,
No. 3 (July), pp. 593-616.
Becker, Gary S., and George J. Stigler. 1974. "Law
This content downloaded from 147.251.55.59 on Fri, 06 Apr 2018 11:01:31 UTC
All use subject to http://about.jstor.org/terms
72-S JOHN ABOWD
Enforcement, Malfeasance, and Compensation of
Enforcers." Journal of Legal Studies, Vol. 3 (January),
pp. 1-18.
Bernstein, Leopold A. 1983. Financial Statement
Analysis: Theory, Application, and Interpretation. Homewood,
Ill.: Richard D. Irwin.
Bhagat, Sanjai, James Brickley, and Ronald C. Lease.
1985. "Incentive Effects of Stock Purchase Plans."
Journal of Financial Economics, Vol. 14, No. 2
(June), pp. 195-215.
Brainard, William C., Matthew D. Shapiro, and John
B. Shoven. 1988. "Fundamental Value and Market
Value." Yale University working paper, May.
Brickley, James, Sanjai Bhagat, and Ronald C. Lease.
1985. "The Impact of Long-Range Managerial
Compensation Plans on Shareholder Wealth."
Journal of Accounting and Economics, Vol. 7, No. 1-3
(April), pp. 115-29.
Citicorp Database Services. 1987. CITIBASE: Citibank
Economic Database (Machine-Readable Magnetic Data
File, 1946-Present). New York: Citibank.
Coughlan, Anne T., and Ronald M. Schmidt. 1985.
"Executive Compensation, Management Turnover,
and Firm Performance: An Empirical Analysis."
Journal of Accounting and Economics, Vol. 7, No.
1-3 (April), pp. 43-66.
Ehrenberg, Ronald G., and George T. Milkovich.
1987. "Compensation and Firm Performance." In
Morris Kleiner et al., eds., Human Resources and the
Performance of the Firm. Madison, Wis.: Industrial
Relations Research Association, pp. 87-122.
Fama, Eugene F., and Michael C. Jensen. 1983a.
"Separation of Ownership and Control." Journal of
Law and Economics, Vol. 26 (June), pp. 301-25.
Fama, Eugene F., and Michael C. Jensen. 1983b.
"Agency Problems and Residual Claims." Journal of
Law and Economics, Vol. 26 (June), pp. 327-49.
Foster, George. 1986. Financial Statement Analysis, 2d
Edition. Englewood Cliffs, N.J.: Prentice-Hall.
Gibbons, Robert, and Kevin J. Murphy. 1990.
"Relative Performance Evaluation for Chief Executive
Officers." Industrial and Labor Relations Review,
this issue.
Gomez-Mejia, Luis R., Henry Tosi, and Timothy
Hinkin. 1987. "Managerial Control, Performance,
and Executive Compensation." Academy of ManagementJournal,
Vol. 30, No. 1, pp. 51-70.
Grossman, Sanford J., and Oliver D. Hart. 1983. "An
Analysis of the Principal-Agent Problem." Econometrica,
Vol. 51, No. 1 (January), pp. 7-46.
Hall, Bronwyn H., Clint Cummins, Elizabeth S.
Laderman, and Joy Mundy. 1988. "The R&D
Master File Documentation." NBER Technical
Working Paper No. 72, December.
Hart, Oliver, and Bengt Holstrbm. 1987. "The
Theory of Contracts." In Truman F. Bewley, ed.,
Advances in Economic Theory, Fifth World Congress.
Cambridge, England: Cambridge University Press.
Healy, Paul M. 1985. "The Effect of Bonus Schemes
on Accounting Decisions." Journal of Accounting and
Economics, Vol. 7, No. 1-3 (April), pp. 85-107.
Holmstr6m, Bengt, "Moral Hazard and Observability."
Bell Journal of Economics, Vol. 10, No. 1
(Spring), pp. 74-91.
Jensen, Michael C., and William H. Meckling. 1976.
"Theory of the Firm: Managerial Behavior, Agency
Costs, and Ownership Structure." Journal of Financial
Economics, Vol. 3 (October), pp. 305-60.
Jensen, Michael C., and KevinJ. Murphy. 1987. "Are
Executive Compensation Contracts Structured Properly?"
University of Rochester unpublished working
paper.
Jensen, Michael C., and Kevin J. Murphy. 1988.
"Performance Pay and Top Management Incentives."
University of Rochester unpublished working
paper.
Lambert, Richard A., and David F. Larker. 1985.
"Golden Parachutes, Executive Decision-Making,
and Shareholder Wealth." Journal of Accounting and
Economics, Vol. 7, No. 1-3 (April), pp. 179-203.
Lambert, Richard A., and David F. Larker. 1987.
"An Analysis of the Use of Accounting and Market
Measures of Performance in Executive Compensation
Contracts." Journal of Accounting Research, Vol.
25 (Supplement), pp. 85-125.
Larker, David F. 1983. "The Association Between
Performance Plan Adoption and Corporate Capital
Investment." Journal of Accounting and Economics,
Vol. 5 (April), pp. 3-30.
Larker, David F. 1986. "Choosing a Performance
Measure: An Analysis of the Alternatives." Topics
in Total Compensation, Vol. 1, No. 1 (Fall), pp.
57-71.
Lazear, Edward P. 1979. "Why Is There Mandatory
Retirement?" Journal of Political Economy, Vol. 87,
No. 6 (December), pp. 1261-84.
Lazear, Edward P. 1986. "Salaries and Piece Rates."
Journal of Business, Vol. 59, No. 3 (July), pp.
405-31.
Leonard, Jonathan S. 1990. "Executive Pay and
Corporate Performance." Industrial and Labor Relations
Review, this issue.
Lewellen, Wilbur G. 1968. Executive Compensation in
Large Industrial Corporations. New York: National
Bureau of Economic Research.
Lewellen, Wilbur G., and Blaine Huntsman. 1970.
"Managerial Pay and Corporate Performance."
American Economic Review, Vol. 60, No. 4 (September),
pp. 710-20.
Lewellen, Wilbur G., Claudio Loderer, and Kenneth
Martin. 1987. "Executive Compensation and Executive
Incentive Problems: An Empirical Analysis."
Journal of Accounting and Economics, Vol. 9, No. 3
(December), pp. 287-310.
Masson, Robert T. 1971. "Executive Motivations,
Earnings, and Consequent Equity Performance."
Journal of Political Economy, Vol. 79, No. 6
(November), pp. 1278-92.
Miller, Merton H., and Myron S. Scholes. 1982.
"Executive Compensation, Taxes and Incentives."
In William F. Sharpe and Cathryn M. Cootner,
eds., Financial Economics: Essays in Honor of Paul
Cootner. Englewood Cliffs, N.J.: Prentice-Hall, pp.
179-201.
Murphy, Kevin J. 1985. "Corporate Performance
and Managerial Remuneration: An Empirical
Analysis." Journal of Accounting and Economics, Vol.
7 (April), pp. 11-42.
This content downloaded from 147.251.55.59 on Fri, 06 Apr 2018 11:01:31 UTC
All use subject to http://about.jstor.org/terms
PERFORMANCE-BASED COMPENSATION 73-S
Murphy, Kevin J. 1985. "Top Executives Are Worth
Every Nickel They Get." Harvard Business Review,
Vol. 64 (March-April), pp. 125-32.
Musgrave, John C. 1986. "Fixed Reproducible
Tangible Wealth in the United States, 1982-85."
Survey of Current Business, Vol. 66, No. 8 (August),
pp. 36-39.
Rabin, Bonnie R. 1986. "Executive Compensation
and Firm Performance: The Case of Employment
Agreements." M.S. thesis, School of Industrial and
Labor Relations, Cornell University.
Rabin, Bonnie R. 1987. "Executive Compensation
and Firm Performance: An Empirical Analysis."
Ph.D. thesis, School of Industrial and Labor
Relations, Cornell University.
Ross, Stephen A. 1973. "The Economic Theory of
Agency: The Principal's Problem." American Economic
Review (Papers and Proceedings), Vol. 63,
No. 2 (May), pp. 135-39.
Standard and Poor. 1988. COMPUSTAT. New York:
Standard and Poor Corporation.
Tehranian, Hassan, and James F. Waegelein. 1985.
"Market Reaction to Short Term Executive Compensation
Plan Adoption." Journal of Accounting
and Economics, Vol. 7, No. 1-3 (April), pp. 131-44.
U.S. Department of Commerce. Bureau of Economic
Analysis. Survey of Current Business, monthly.
U.S. Department of Commerce. Bureau of Economic
Analysis. 1987. Fixed Reproducible Tangible Wealth in
the United States, 1925-85. Washington, D.C.: GPO.
This content downloaded from 147.251.55.59 on Fri, 06 Apr 2018 11:01:31 UTC
All use subject to http://about.jstor.org/terms