Fragmented Users of Crime Predictions

As Dean (and Judge) Guido Calabresi says, it is the duty of academics to offer half-baked ideas for others to perfect. Henderson, Wolfers, and Zitzewitz, with their proposal to establish open markets for the prediction of crime, have created a gloriously half-baked idea. My contribution in the kitchen must remain small in this brief comment. I explore a few implications stemming from one fact: the prediction market concept—an effort to coordinate decentralized sources of information—would operate in an exceptionally decentralized world of users, a world where the institutional users of crime predictions are fragmented among many different locations and levels of government.

I begin with a sketch of the changing data landscape in criminal justice and note how these changes have contributed to long-term centralization in the response to crime at the highest levels of government. I show how, despite this long-term trend, the social response to crime in the United States today still remains quite decentralized. Fragmented local institutions, especially police departments and prosecutors’ offices, would find it difficult to use crime prediction markets. Instead, institutions at higher levels of government, such as state-level correctional authorities, would be best positioned to rely on prediction markets. They would use markets to overcome the greatest information challenge at the higher level: coordinating input from many incompatible sources. Finally, I close by noting that a few actors at the local level—in particular, sentencing judges—combine features of fragmented and centralized users. The ability of local actors to use far-flung crime data poses the question of whether markets at the case level would be either appealing or possible.