|
Abstract:
|
A central goal in traffic safety research is to evaluate the effectiveness of safety countermeasures. Observational before-after design is common in these studies, where safety outcomes are recorded for each roadway segment both before and after the countermeasures are implemented. Despite its causal nature from a statistical perspective, such an evaluation has rarely utilized causal inference methods. In this paper, motivated from a real application of evaluating the effects of rumble strips on reducing vehicle crashes, we consider a difference-in-differences (DID) framework for causal inference in traffic safety before-after studies. Within this framework, we connect the DID outcome regression method and the standard Empirical Bayes (EB) approach in traffic safety research, giving the latter a causal interpretation. We further propose a new double-robust DID estimator that hybridizes regression and propensity score weighting. We assess the fundamental parallel trend assumption in DID indirectly through analyzing the crash outcomes in the pre-treatment periods. We conduct a simulation study to demonstrate the advantage of the double-robust method over alternative methods. Our empi
|
