A Principal-Agent Model of Elections with Novice Incumbents
| DOI | 10.1177/095162902774006813 |
| Published date | 01 October 2002 |
| Date | 01 October 2002 |
| Author | Sugato Dasgupta,Kenneth C. Williams |
| Subject Matter | Articles |
A PRINCIPAL±AGENT MODEL OF ELECTIONS
WITH NOVICE INCUMBENTS
SOME EXPERIMENTAL RESULTS
Sugato Dasgupta and Kenneth C. Williams
ABSTRACT
The ef®cient aggregation of information is a prerequisite for the operationof a
democratic system. In order for political of®ce-holders to act responsively so
that policy outcomes re¯ect public sentiments, suf®cient levels of information
about their actions must be aggregated within the electorate. This is funda-
mental since very few citizens will be aware of the of®ce-holder's actions.
One way in which information is aggregated is through public opinion polls.
This mechanism extracts information from informed citizens and transmits
it in a manner so that collectively the electorate behaves as if it were fully
informed. We examine this aggregation process within a principal±agent
model of elections. We further assume novice incumbents whose choices are
more uncertain than experienced public of®cials. A rational expectations equi-
librium is speci®ed in which uninformed citizens condition their beliefs on poll
results. Laboratory experiments are conducted to test the predictions of the
model. We ®nd that, compared to competing models, a rational expectations
hypothesis best explains the aggregation process.
KEY WORDS .asymmetric information elections .experimental methods .
information aggregation .principal±agent .rational expectations
1. Introduction
In a principal±agent theory of elections, voters (the principal) elect a political
of®ce-holder (an agent) to address their policy concerns. However, once
elected the responsiveness of the incumbent to the voters is determined
largely by the amount of information that is transmitted between the parties.
In the absence of information transmission, incumbents are not accountable
to the voters; their actions are private information. By contrast, when infor-
mation is transmitted fully, incumbents are held accountable since voters
Journal of Theoretical Politics 14(4): 409±438 Copyright &2002 Sage Publications
0951±6928[2002/10]14:4; 409±438; 027703 London, Thousand Oaks, CA and New Delhi
The authors would like to thank James Kuklinski, Richard McKelvey, Rebecca Morton,
Arthur Schram, Frans van Winden, Rick Wilson, and the co-editor Nicholas Miller for their
helpful advice and suggestions. The authors are responsible for any remaining errors.
have complete information about their actions. Many have alleged that this
latter state is unrealistic. Since information acquisition is costly and the prob-
ability of casting a decisive vote in an election is small, voters have little
`rational' incentive to be informed about relevant political events and
issues (Downs, 1957).
1
Consequently, while some citizens are cognizant of
the of®ce-holder's actions, most citizens remain unaware.
2
However, a poorly informed electorate does not preclude the possibility of
popular control. So long as a fraction of citizens gather pertinent informa-
tion, electoral mechanisms manage to channel the acquired information
from the informed to the uninformed.
3
The important question therefore
becomes: what types of mechanisms allow for the ef®cient aggregation and
transmission of the electorate's collective information? Within a principal±
agent model, and examining both adverse selection and moral hazard
problems, we consider theoretically and experimentally the effectiveness of
polling mechanisms. We posit that rational expectations, wherein un-
informed voters condition their beliefs on poll results, automatically aggre-
gates and transmits the electorate's collective information.
Our theoretical model draws together two different strands of modeling in
the political science literature. First, our emphasis on opinion polls is inspired
by the seminal work of McKelvey and Ordeshook (1984, 1985a, b, 1986).
They considered a spatial model of elections with two of®ce-seeking candi-
dates. Although only a subset of voters observed the policy locations of
the two candidates, opinion polls fully transmitted location-speci®c data
from the informed to the uninformed voters. Consequently, results from the
standard spatial model setup (e.g. convergence to the median) were robust
to the violation of the `complete information' assumption.
4,5
We adapt the
McKelvey±Ordeshook construction to a different electoral environment
410 JOURNAL OF THEORETICAL POLITICS 14(4)
1. We should note that within a rational choice framework, Ledyard (1984) and Palfrey and
Rosenthal (1985) demonstrate that a single vote may be more relevant than is generally believed.
2. A large literature has documented the many ways by which voters can short circuit the high
cost of information. Information extracted from partisan labels, ideology or endorsements from
interest groups allows initially uninformed voters to make reasoned choice (see, e.g. Collier et al.
(1987), Ferejohn and Kuklinski (1990), Grofman and Norander (1990), Grofman (1993), Hinich
and Munger (1992), Lupia (1993, 1994) and Popkins (1991)).
3. We conjecture that this fraction can be `quite' small, but the actual size is subject to empiri-
cal veri®cation.
4. McKelvey and Ordeshook's idea is based on rational expectations models of the stock
market. In a rational expectations equilibrium, stock prices convey all information that traders
privately possess (see, e.g. Grossman (1981), Lucas (1972), Muth (1961) and Radner (1979)).
5. A substantial literature in experimental economics has focused on markets and the
ability of market variables such as prices to transmit information from informedto uninformed
traders. Studies such as Forsythe and Lundholm (1990), Forsythe et al. (1982), Plott and Sunder
(1982, 1988) and Sunder (1992) have found support for rational expectations behavior in the
laboratory.
To continue reading
Request your trial