Institute for Mathematical Behavioral Sciences

Models of elections tend to predict that parties will maximize votes by converging to an electoral center. There is no empirical support for this prediction. In order to account for the phenomenon of political divergence, this paper o¤ers a stochastic electoral model where party leaders or candidates are di¤erentiated by di¤ering valences-the electoral perception of the quality of the party leader. If valence is simply intrinsic, then it can be shown that there is a “convergence coe¢ cient”, de…ned in terms of the empirical parameters, that must be bounded above by the dimension of the space, in order for the electoral mean to be a Nash equilibrium. This model is applied to elections in Turkey in 1999 and 2002. The idea of valence is then extended to include the possibility that activist groups contribute resources to their favored parties in reponse to policy concessions from the parties. The equilibrium result is that parties, in order to maximize vote share, must balance a centripetal electoral force against a centrifugal activist e¤ect. This model is used to explain the location of the major parties in Turkey away from their inferred equilibrium positions according to the stochastic electoral model with intrinsic valence. Finally this model is applied to the case of a non-democratic regime, where an autocrat must bargain with economic and military elites in order to stay in power.