Identifying the impact of deterrence on crime: internal versus external instruments

We consider estimation of the economic model of crime exploiting instrumental variables (IV) techniques for panel data. We extend the empirical analysis of Cornwell and Trumbull (1994) and show that their instruments are very weak. We propose an alternative identification strategy based on the sequential moment conditions of Arellano and Bond (1991). The resulting GMM estimates of deterrence effects on crime are considerably more precise.