Rational Expectations Voting in Agent-Based Models: An Application to Tax Ceilings

This paper introduces rational expectations voting into an agentbased model of collective choice. Our model is unique because it generates sophisticated forecasts of endogenous policy outcomes by computationally sampling the space of exogenous random variables. Together these forecasts generate a common prior, a joint distribution of all random variables as a function of the set of policy choices, which agents use to select the policy that maximizes their expected utility. We apply our simulated rational expectations methodology by using administrative data on property taxes from two U.S. cities to investigate how observed levels of (plausibly exogenous) tax-payment uncertainty affect collective choice. Specifically, we show that, for sophisticated risk-averse or loss-averse voters, higher levels of tax-payment uncertainty generate majority support for a binding constraint on collective choice.