Two aspects of the electoral process are modelled formally: rational voting when information is costly and imperfect; and candidate strategies when election outcomes cannot be predicted with certainty, under various assumptions about candidates' goals. In both cases, the problem is to rigorously define and characterize rational behavior when uncertainty is taken explicitly into account. Voters are seen as engaging in a process of optimal sequential sampling, gathering information and updating prior beliefs about candidate positions while deciding at each stage whether to continue gathering information or to stop and choose a candidate. Under quite general conditions, this process is shown to yield a well defined rule for rational decision making, characterized by a functional equation. The following properties hold for this process: (1) if: the cost of information increases, the set of prior belief states at which further sampling occurs is made smaller; (2) certain kinds of increased uncertainty about the desirability of candidates will increase the value of sampling; (3) under special conditions, the candidate chosen will be the one seen as preferred in the final observation, but this relation may fail under more general assumptions. When voters are assumed to observe only the utility level of candidate platforms, and not the platforms themselves, the conditions of (3) can be generalized. Also, a model of voting and the development of party identification can be defined, bearing a close resemblance to certain non-rational-choice models, which parsimoniously predicts many observed properties of voting. Candidates are modelled as either seeking to win office or as seeking to implement preferred policies; and as being either certain or uncertain about the outcome of the election given both candidates' strategy choices. The "median voter" or convergent platforms result of spatial modelling holds whenever candidates can predict the outcome with certainty, or when uncertain candidates seek only to win office. But policy-oriented candidates under uncertainty will never adopt identical platforms in equilibrium.
Randall L. Calvert (Wed,) studied this question.