PARTISANSHIP AS A SOCIAL CONVENTION

Existing research models partisanship as a function of either running tallies of party performance evaluations or emotional identifications with parties. However, these models are arguably insufficient to account for the variation in the propensity to act on behalf of a party. This article develops and tests a model of partisanship as a social convention. The decision to act on behalf of a party is modeled as an asymmetric n-person coordination game with multiple equilibria, where the payoffs from being coordinated on acts of partisanship are higher than the payoffs from coordination on abstention if the costs of those actions are sufficiently low. Given that such a coordination game will be easier to solve where acts of partisanship are more public, we should see a greater incidence of partisanship in states with laws providing for publicly available party registration, relative to states without such laws. The model is tested using data from the 1984-96 American National Election Studies merged with data on states' party registration laws. Several measures of partisanship are shown to be responsive to the presence of laws providing for party registration, controlling for other factors known to affect the propensity to be partisan.