Comment on a “Modest Proposal” for a Crime Prediction Market

Henderson, Wolfers, and Zitzewitz conclude their paper, *Predicting Crime*, with what they term a “modest proposal”: an entity, such as the National Institute of Justice (NIJ), should establish a crime prediction market to develop and test the authors’ approach for forecasting crime rates and informing crime policy. They argue that prediction markets, which have been used to predict election outcomes and are used by businesses to forecast sales, costs, and other outcomes, can also be used to forecast crime rates. They argue that prediction markets can be used to provide crime policy decisionmakers with transparent, actionable information that can guide decision-making as well as evaluate the effects of crime policy decisions. The authors state that the market mechanism would provide a centralized location for aggregating information about crime forecasts, and that policymakers can then evaluate the information about future crime rate predictions and thereby make a more informed decision. The authors propose that a federal entity, such as NIJ, should establish the prediction market to overcome potential political battles that might occur at the local level and that a federal role is appropriate because of overlapping jurisdictions.

The proposed initial experiment is straightforward: it would focus on predicting various levels of Uniform Crime Reports (UCR) crimes over the next three years, relying on a group of about 200 invited crime experts who would be endowed with $200 to purchase prediction contracts in the new markets. Stakes would be modest, and traders could cash in only after accumulating $500. The top trader each year would receive a non-financial award. The objectives of this experiment would be to learn more about the organization and operation of the market, such as how to write contracts, how much participation to allow, and what stakes to set.

Our comment principally focuses on practical considerations for implementing the authors’ modest proposal. We begin by evaluating the proper scope of an experimental crime prediction market. In that part we question whether a national prediction market would be useful for a first experiment. We examine the alternative—a local crime market—as a way of articulating the challenges that would attend any experimental crime prediction market. The scope of the prediction market is linked to several other key considerations that we address in turn: who the right traders are; what the payoff is and how market incentives would work; and how success is to be measured. Finally, we conclude with a discussion of whether prediction markets about crime—national or local—will have a chance to inform crime policy and decision-making.