We examine the effect of unionization on police productivity in large U.S. metropolitan areas. We define police output in the context of a production function model that draws also on the crime literature. We estimate the resulting model using a data set that includes published and unpublished government statistics as well as our own survey of police departments. Results suggest that the effect of unions on police productivity varies according to categories of police performance. In particular, if performance is stratified according to the severity of crimes, unions seem to have an insignificant effect on police productivity with respect to serious crimes. For minor crimes, unionization alters the parameters of police production function, leading to diminished productivity.