AN INSTITUTIONAL THEORY OF DIVIDED GOVERNMENT AND PARTY POLARIZATION

This paper develops a simple model of institutionally driven party competition which may explain the persistence of divided government. In the basic model, two parties that are solely motivated to win elections compete for two offices, the legislature and the executive. Parties offer positions in a single dimension to a single-district sophisticated electorate that understands the institutionally defined relationship between electoral and policy outcomes. When parties are risk-averse, competition for the two offices leads to a prisoner's dilemma: the parties can jointly improve on the one-shot Nash equilibrium in which they adopt convergent positions. Several extensive form election games are shown to allow the parties to solve their prisoner's dilemma in a subgame-perfect Nash equilibrium. In such an equilibrium, parties adopt widely separated positions, and divided government and split-ticket voting occur with high probability. Unlike typical voting models, in the basic model there is no tendency for electoral equilibrium to produce the (expected) median voter's ideal policy outcome. These results are robust to a variety of extensions, including significant sincere voting, multiple policy dimensions, and defections by incumbent candidates.