Presidential betting markets predict election outcomes more
accurately than polls because of their ability to effectively aggregate information. Empirical research and theory indicates that the result extends to other contexts. Betting markets, more formally called information markets, provide accurate predictions about future product sales, box office receipts, and other future events. Moreover, market predictions generally outperform other prediction mechanisms. This paper argues that empirical research and theory indicates that we should use information markets’ predictive power to make administrative decisions. In addition, it presents a model information market designed to help policy makers evaluate policies prior to their implementation by providing policy makers information about the policies’ effects in the form of market predictions. To design such a market, it is necessary to determine how the market should pay off bettors when the agency does not implement a policy because the market predicts it will have an adverse effect. The problem is that bets pay off based on the outcome of an event, but when the policy makers decide not to implement a policy, the policy has no effect and thus it is unclear how to compensate bettors. This paper shows that through clever market design it is possible to return the market price of a bet, prior to an agency’s decision not to implement the policy on which the bet depends, without fear of market manipulation. Consequently, even in cases where using market predictions to make administrative decisions appears problematic, it is possible.