Equilibria in multi-party competition under uncertainty (Q809849)

The mathematical theory of party competition deals with a continuum of alternatives (possible platforms) and a continuum of voters. A voter \(x\in [0,1]\) is identified with its ideal point. Voters are uniformly distributed over [0,1]. Each party \(i=1,...,n\) is characterized by its platform \(x_ i\in [0,1]\). Each voter votes for the party whose platform is closest to its ideal point and each party selects the platform maximizing its vote share. The outcome of the game is a platform configuration \((x_ 1,...,x_ i,...,x_ n)\) and the structure of Nash equilibria is now well-known. The authors consider an alternative approach of the game when the parties recognize that for some extraneous reason the voters may not vote for the nearest party. Let the utility of voter \(x\in [0,1]\) voting for party \(i=1,...,n\) be given by \(U_ i(x)=f(| x-x_ i|)+\epsilon_{ix}\) where f is strictly decreasing and \(\epsilon_{ix}\) is a continuous random variable with zero mean such that \(P_ i(\epsilon_{ix}=\epsilon_{jx})=0\), \(j\neq i\). The probability that voter \(x\in [0,1]\) votes for party \(i=1,...,n\) is \(P_ i(x)=P_ i\{U_ 1(x)=\max_{1\leq j\leq n}U_ j(x)\}\). Now the game on \((x_ 1,...,x_ n)\) is in terms of the expected vote shares \[ \bar S_ i(x_ 1,...,x_ n)=\int^{1}_{0}P_ i(x)dx. \] The authors actually consider a special case where \(P_ i\) is given by a ``multinominal logit''. The main feature of the model is the multiplicity of Nash equilibria. Under ``sufficien'' uncertainty about the voters' behaviour it is an equilibrium that the parties select a common platform corresponding to the median voter position (Minimum Differentiation). But there are other equilibria. The examples studied by the authors indicate that, when the number of parties is large enough, political competition may lead to ``unstable'' situations and when the number of parties is not too large neither too small there always is an equilbrium in which peripherical parties are paired (Local Clustering).