Third parties in equilibrium

Under standard assumptions equilibria with three parties normally do not exist in spatial models of electoral competition. In this paper I show that such equilibria are possible if it is assumed that voters are uncertain about the exact policies parties will adopt once elected. Substantive predictions can be derived from the model, explaining some features of three-party competition. First, the least risk party will always take the most moderate position. Second, this position is also winning. Third, the two riskier parties are always on opposite sides of the median voter and also of the moderate party.