Trust and Incentives in Principal-Agent Negotiations
| DOI | 10.1177/095169280201400204 |
| Author | Andrew B. Whitford,Gary J. Miller |
| Date | 01 April 2002 |
| Published date | 01 April 2002 |
| Subject Matter | Articles |
TRUST AND INCENTIVES IN PRINCIPAL±AGENT
NEGOTIATIONS
THE `INSURANCE/INCENTIVE TRADE-OFF'
Gary J. Miller and Andrew B. Whitford
ABSTRACT
The canonical principal±agent problem involves a risk-neutral principal who
must use incentives to motivate a risk-averse agent to take a costly, unobserv-
able action that improves the principal's payoff. The standard solution
requires an inef®cient shifting of risk to the agent. This article, however, sum-
marizes experimental research that throws doubt on the validity of this conclu-
sion. Experimental subjects were routinely able to achieve ef®ciency in agent
effort levels without inef®cient risk-sharing. These experimental outcomes,
while anomalous from the standpoint of principal±agency theory, are quite
consistent with other experimental data testing notions of trust-based implicit
contracting. Such contracting within a hierarchy may allow an outcome
preferred, by both principal and agent, to that deemed possible by principal±
agency theory. If this is true, then the lessons to be learned from principal±
agency theory are all the wrong ones. Concentrating on incentives can
crowd out the very qualities in a relationship that make social ef®ciency
possible.
KEY WORDS .agency .contracts .incentives .risk .trust
More than half a century ago, Berle and Means (1932) argued that many
diffuse shareholders could not hope to constrain corporate managers to
act in the shareholders' interest. Corporate managers were, therefore, free
to pursue their own interests. Since the interests of managers were quite dis-
tinct from those of shareholders, managers were presumably free to use the
resources of the ®rm in their own interests. The legal rights of the stock-
holders were not matched, they argued, by effective control of the ®rm.
Journal of Theoretical Politics 14(2): 231±267 Copyright &2002 Sage Publications
0951±6928[2002/04]14:2; 231±267; 021887 London, Thousand Oaks, CA and New Delhi
We thank Dan Carpenter for insightful comments. This article was presented at the Scienti®c
Study of Bureaucracy conference in College Station, Texas, and the annual meetings of the
Midwestern Political Science Association and the International Society for New Institutional
Economics. All remaining errors are our own.
Not coincidentally, Berle and Means wrote at the dawn of the New Deal, a
period in which the evolving power of unelected federal bureaucrats raised
similar concerns about managerial accountability in the public sphere. The
problem of public bureaucratic accountability was, in some ways, analogous
to that of private managerial accountability. The legal rights of citizens in a
democracy are held by a large diffuse constituency; the necessity of energetic
and effective management of large organizations seemed at times to create a
real mismatch between citizenship (or ownership) and control of the organi-
zational infrastructure of the state (or ®rm).
In the 1970s, concern with the problem of corporate accountability gener-
ated a body of research known as principal±agency theory (Fama, 1980;
Mirrlees, 1976; Ross, 1973; Shavell, 1979). This literature recognizes an
information asymmetry. Shareholders (the principal) cannot monitor corpo-
rate managers directly; therefore, they need to induce an agent to take costly
actions in the interests of the principal, without monitoring. The early litera-
ture found that, despite the information asymmetry, principals could ®nd a
(constrained) optimal solution that induces the greatest possible agent
effort even though the principal can never directly observe the effort.
As popularly interpreted, the message of principal±agency theory has been
that Berle and Means underestimated the ability of principals to shape
agents' behavior; even in the presence of an information asymmetry, share-
holders may use incentives to control the behavior of managers. In
academics, this has caused an explosion of research on incentive contracts
for corporate managers, sales staff and all other positions in the ®rm
(Milgrom and Roberts, 1992). It has also fueled an expansion in the use of
such incentives in the world of corporate America.
Principal±agency theory has had enormous in¯uence in the study of public
agencies as well as private ®rms. Part of the attraction has been the use of
principal±agency theory to study accountability in hierarchical relationships
where monitoring costs make direct control infeasible. For example, the
literature on congressional oversight of bureaucracy had traditionally
lamented the scarcity of direct monitoring by congressional committees;
and Fiorina (1981) had lamented a `Mismatch of Incentives and Capabilities'
in congressional control of the bureaucracy. However, Weingast (1984) used
principal±agency theory to argue that Congress's lack of direct oversight
could be a manifestation of effective use of incentives, which allow Congress
to control agencies without expensive oversight procedures.
The purpose of this article is to re-examine canonical principal±agency
theory and its implications for public bureaucracy. We argue that the pri-
mary message of the theory is really a negative one ± ef®ciency is not to be
obtained easily. We further argue that in the case of public bureaucracy
the costs of using incentives are likely to be high. We also argue that empiri-
cal research on the canonical principal±agency problem casts doubt on the
232 JOURNAL OF THEORETICAL POLITICS 14(2)
predictive power of the theory, and that those doubts are again especially
applicable in the case of public bureaucracy. Furthermore, if the predictions
of principal±agent theory are falsi®ed, the focus on incentives brought about
by the interest in principal±agency theory may lead to a set of managerial
prescriptions that are counter-productive.
Because of these issues, the concentration on incentive plans as a solution
to the problems facing private and public bureaucracies is misplaced.
A greater emphasis should be placed on traditional concerns of the public
administration literature: internal motivation, selection, organization and
bureaucratic politics.
In the next section, we review the tradeoff between insurance and incen-
tives implied by principal±agency theory. The subsequent section addresses
an experiment meant to test the principal±agency solution to this tradeoff.
We then turn to reciprocity as an explanation for the anomalies observed
in principal±agent exchanges, and then to the crowding out alternative moti-
vations by a focus on incentives. Last we offer a set of illustrations of the
power of reciprocity in work environments, with special emphasis on
public management.
Insurance versus Incentives: The Lessons of Principal±Agency
To induce the agent to work hard, we will have to give up some of the ef®ciency that is
obtained by putting all the risk on the principal. The question is, How can we do this
as ef®ciently as possible? (Kreps, 1990: 584)
The basic principal±agency problem is very simple. There is one principal
and one agent. The principal cannot observe the agent's efforts, but can
clearly measure outcomes that affect her/his well-being ± in fact there is
only one outcome that matters. In Dixit and Nalebuff 's highly accessible
illustration, that outcome is the success or failure of the principal's computer
game (1991).1
If the outcome in question were dependent only on the agent's actions, then
the principal could work backwards from the observed outcome to deduce
the agent's actions. For example, if the computer program's success were
MILLER AND WHITFORD: TRUST AND INCENTIVES IN NEGOTIATIONS 233
1. In a number of ways, the language of principal±agency theory has been borrowed by poli-
tical science without coming to grips with its real meaning. In essence, principal±agency theory
has come to be simply a metaphorical synonym for `hierarchical'. This article achieves two prox-
imate goals: to evaluate the principal±agency's empirical power in its canonical form (hence the
Dixit and Nalebuff example), and to extend the theoretical environment of principal±agency to
reincorporate both older and more recent strains of literature with perhaps greater importance
for the study of public bureaucracies.
To continue reading
Request your trial