Biased learning from elections

Political economy models of elections study how electoral incentives encourage parties to be responsive to voters. Election results can be informative about how platforms diverge from voter preferences, but the complexity of the environment and cognitive biases may inhibit learning. We develop a repeated model of elections with motivated reasoning and uncertainty about the fairness of the electoral system. A party who wants to believe the electorate is close to their preferences picks more extreme platforms, which in turn causes the other party to be more extreme as well. Motivated beliefs also allow parties to infer from poor results that elections are unfair rather than that their platforms are unpopular. Repeated elections exacerbate this problem: disagreement about the fairness of the electoral system increases over time, even if platform divergence decreases.