Online Program

Data Analysis Using Standard vs. Bracketed Difference in Difference Estimation (305314)

Keywords: Bracketing, Causal inference, Difference-in-Difference, Violent Crime, Policy Evaluation, Economics

The difference-in-differences (DiD) model is a popular statistical model in a variety of disciplines that compares the change in an outcome variable of interest in a treatment group that receives an intervention to a control group that does not receive the intervention. It is often difficult for researchers to assess the validity of the assumptions of the DiD model when using observational data. One violation that might commonly occur is the influence of unobservable confounding variables on the outcome variable of interest. Hasegawa et al. introduce a bracketed DiD approach that provides an unbiased interval estimate of the treatment effect in some such cases when the influence of the confounder in certain situations. We then implement both DiD approaches to estimate the effect of Medical Marijuana Legislation (MML) on violent crime rates using a sample of fifteen states. Here we show how to implement the bracketing approach of Hasegawa et al to the two-way fixed effects DiD estimator. Recent estimates of the effect of MML on violent crime rates vary significantly with several papers finding a negative effect of MML on violent crime but others finding a positive effect.