NBER WOR~G PAPER SERIES
EXECUTIVE COMPENSATION, STRATEGIC
COMPETITION, AND RELATIVE
PERFORMANCE EVALUATION: THEORY
AND EVIDENCE
RajeshAggarwal
Andrew A. Samwick
Working Paper5648
NATIONAL BUREAU OF ECONOMIC RESEARCH
1050MassachusettsAvenue
Cambridge,MA 02138
Ju]y 1996
We thank Sheri Aggarwal, Gary Engelhardt,Wayne Gray, Benjamin Hermalin, Dennis Logue,
Jonathan Skinner, Matthew Slaughter,Sang-Seung Yi, and seminar participantsat Dartmouth
College for theirhelpful comments.
We are gratefulto John Fitzgerald for assistancewith the
ExecuComp dataset;AndrewHaitfor assistancewiththeCensusof Manufactures,andErvinTu for
researchassistance.All errorsareourown. Any opinionsexpressedarethoseof theauthorsandnot
those of theNationalBureauof Economic Research.
@ 1996by RajeshAggarwalandAndrew A. Samwick. All rightsreserved. Shortsections of text,
notto exceed two paragraphs,maybe quoted withoutexplicit permissionprovided thatfull credit,
including O notice, is given to thesource.
NBER Working Paper5648
July 1996
EXECUTIVE COMPENSATION, STRATEGIC
COMPETITION, AND RELATIVE
PERFORMANCE EVALUATION: THEORY
AND EVIDENCE
ABSTRACT
We arguethatstrategicinteractionsbetween firms in anoligopoly can explain thepuzzling
lackof high-poweredincentivesinexecutivecompensationcontractswrittenby shareholderswhose
objective is to maximize thevalue of theirshares. We derive theoptimal compensation contracts
for managersanddemonstratethattheuse of high-powered incentiveswill be limited by theneed
to softenproductmarketcompetition. Inparticular,when managerscan be compensated based on
theirown andtheirrivals’ performance,we show thattherewill be aninverserelationshipbetween
the magnitude of high-powered incentives and the degree of competition in the industry. More
competitiveindustriesarecharacterizedby weakerpay-performanceincentives.Empirically,we find
strongevidence of thisinverserelationshipin thecompensation of executives in theUnitedStates.
Oureconometric resultsarenot consistentwith alternativetheoriesof theeffect of competition on
executive compensation. We conclude thatstrategicconsiderationscan preclude theuse of high-
powered incentives, in contrastto thepredictionsof thestandardprincipal-agentmodel.
Rajesh Aggarwal
Amos Tuck School
DartmouthCollege
Hanover,NH 03755
Andrew A. Samwick
Departmentof Economics
DartmouthCollege
Hanover,NH 03755
andNBER
1. Introduction
The separation of ownership and control of large corporations is a central feature of modern
economies. The largely unobserved choice of actions by a firm’s managers can have a major
impact on the wealth of its shareholders. The literature on principal-agent theory suggests
that the primary means for the shareholders to ensure that managers take optimal actions
is to tie managers’ pay to the performance of their firms; that is, to provide high-powered
intentives for managers to maximize the returns to shareholders. However, the empirical
literature has not found evidence of high-powered incentives. Instead, the typical finding is
a low pay-performance sensitivity, which seems to imply that shareholders are not providing
managers with incentives to maximize the returns to shareholders.
In this paper, we argue that strategic interactions between firms in an oligopoly can ex-
plain the puzzling lack of high-powered incentives in executive compensation contracts writ-
ten by shareholders whose objective is to maximize the value of their shares. We derive the
optimal compensation contracts for managers and demonstrate that the use of high-powered
incentives will be limited by the need to soften product market competition. In particu-
lar, when managers can be compensated based on their own and their rivals’ performance,
we show that there will be an inverse relationship between the magnitude of high-powered
incentives and the degree of competition in the industry. More competitive industries are
characterized by weaker pay-performance incentives. Empirically, we find strong evidence of
this inverse relationship in the compensation of executives in the United States.
We conclude that strategic considerations can preclude the use of high-powered incen-
tives, in contrast to the predictions of the standard principal-agent model. That model
posits an economic trade-off between inducing the correct amount of unobservable effort by
the agent and minimizing the amount of risk she is required to bear. In practice, that model
1
predicts that managerial compensation will be correlated with the total return to sharehold-
ers, typically through ownership of shares of the firm’s stock or grants of options on the
fire’s stock. In a seminal article, Jensen and Murphy (1990) test and reject this theoretical
prediction, finding that the compensation of chief executive officers increased by ody $3.25
per $1,000 increase in shareholder wealth.1
They hypothesize that political forces (e.g.,
societal notions of fairness and equity) work to reduce the pay-performance sensitivity.2
This contrasts with our explanation that the pay-performance sensitivity is low because a
high pay-performance sensitivity would induce overly aggressive behavior in the product
markets.
Another prediction of the principal-agent model of executive compensation is that pay
should be based not only on the returns to the firm’s shareholders but on every variable that
provides unique information about the actual action taken by the manager.3 A prominent
example of such a variable is the profits of other firms in the same product market, leading
to a compensation contract based on the relative performance of the manager compared to
her rivals. The prediction from the relative performance evaluation model is that, other
things equal, an executive will receive lower compensation if executives of rival fires deliver
higher returns to their shareholders. In our model, the validity of this prediction depends on
the nature of strategic competition. When the returns to attenuating competition are high,
compensation contracts will not exhibit relative performance evaluation. When the returns
to achieving product market leadership are high, contracts will have incentives that resemble
1 Later work by Haubrich (1994) has demonstrated that some parameterizations of the principal-agent model
do allow for pay-performance sensitivities as low as the 0.003 found by Jensen and Murphy.
2 Finkelstein and Hambrick (1996) discuss the political forces that shape executive compensation in great
detail.
q This is the sticient statistic restit from Holmstrom’s (1979) work on moral hazard.
2
relative performance evaluation. The empirical literature on whether relative performance
evaluation is an important source of managerial incentives finds mixed results.4 We do not
find evidence in support of relative performance evaluation in our empirical work.
The lack of empirical confirmation is a consequence of the absence of strategic con-
siderations from the standard principal-agent model, with or without relative performance
evaluation. With few exceptions, the typical principal-agent model fails to recognize that
the interaction between shareholders and managers occurs in an environment of strategic
interactions between firms in imperfectly competitive markets.5
A model that considers
compensation divorced from competition is incomplete. We show that when the optimal con-
tract allows compensation to vary with own and industry performance, strategic interactions
support relative performance pay—a negative relationship between executive compensation
and industry performanc=nly when firm outputs are strategic substitutes, as in the stan-
dard Cournot model.
When outputs are strategic complements, as in the differentiated Bertrand model of
price competition, relative performance pay lowers the shareholders’ returns by encouraging
more aggressive price setting by managers and is therefore not observed in equilibrium. In
this case, the level of compensation under the optimal contract incremes with both own
and industry performance, softening competition and raising the returns to shareholders.
d Jensen and Murphy find that relative performance is not an important source of managerial incentives.
Gibbons and Murphy (1990) test more directly for relative performance pay and find that, holding constant
the rate of return on a firm’s common stock, a higher value-weighted industry rate of retmn lowers the
growth of CEO pay.
5 The earliest papers that do recognize this interaction are Vickers (1987), Fershtman and Judd (1987),
and Sklivas (1987). Vickers (1987) demonstrate es that precommit ment to managerial incentive cent racts can
facilitate collusion between rival firms. Fershtman and Judd (1987) and Sklivaa (1987) derive similar results
by allowing the compensation contract to include sales in addition to profits. Reitman (1993) demonstrates
that the nonlinearity in profits introduced by a stock option improves the outcome for the shareholders
relative to a contract that is linear in profits. See also Katz (1991), Hermalin (1992), and Fumas (1992).
3