Crime and punishment: evidence from dynamic panel data model for North Carolina (2003–2012)

Abstract

This paper exploits a dynamic panel data model to analyze the endogeneity of certain deterrent covariates, measurement errors in crime and inertial features of crime, and to check for unobserved heterogeneity in order to extract more precise and reliable estimates of the deterrent effect of law enforcement in North Carolina. It uses panel data on 90 counties of North Carolina over the period 2003-2012. After checking various socioeconomic covariates, among the deterrent variables, likelihood of arrest and conviction always showed deterrent effects on criminal behavior for different kinds of violent and property crimes. Since property crimes have relatively higher inertia than violent crimes, their associated long-run elasticities with respect to significant deterrence variables are greater, which implies that allocating more resources to increasing the likelihood of arrest for property crimes can effectively decrease them in the long run.

Does higher educational attainment imply less crime? Evidence from the Indian states

“…Therefore, we consider police efficiency as weakly exogenous. In addition to this, both the problem of endogeneity of some regressors and measurement error in crime can be well addressed by these difference-in-difference and system- GMM by applying suitable instruments available within the data (Ghasemi, 2017).…”

“…persistence over time) and feasibility of dynamic panel model (which observes small T and large N ), we employ an instrumental variable approach for panel data using Difference-GMM and system-GMM-system estimator (Arellano and Bond, 1998). The GMM – system estimator, which allows for the use of lagged regressors as instruments, has the advantage of controlling estimated parameters for endogeneity, which is fairly well documented in the empirical literature of economics of crime (Sah, 1991; Glaeser et al , 1996; Ghasemi, 2017).…”