Two-Stage Bayesian Model Averaging (2SBMA): an alternative for quantitative criminal studies

This study applies Two-Stage Bayesian Model Averaging (2SBMA) to a crime deterrence framework to address endogeneity, model uncertainty, and weak instruments. Using a simulated crime-like dataset, we redefine the offending rate as a per capita measure and compare 2SBMA(1), 2SBMA(2), and 2SLS. The results show that the just identified 2SBMA(1) provides the most stable and credible deterrence estimates, with findings robust to alternative priors. Although simulation yields upper bound performance, the comparative evidence highlights 2SBMA as a promising tool for criminological inference.