Are U . S . Cities Underpoliced ? :

The socially optimal number of police hinges on the extent to which police reduce the most costly crimes, which are also the most difficult to model econometrically because they are rare. In the hope of minimizing simultaneity bias, papers in the recent literature have focused on quasi-experimental approaches that disregard most of the variation in police staffing levels, compounding the modeling difficulty. We argue that the central empirical challenge in this literature is not simultaneity bias, as has been supposed, but measurement error bias. Using a new panel data set on crime in medium to large U.S. cities over 1960-2010, we obtain measurement error corrected estimates of the police elasticity, with much greater parameter certainty for the most costly crimes. Our analysis suggests that U.S. cities are in fact underpoliced. JEL Classification: K42, H76, J18