Difference-in-Differences Models for Causal Inference in Stata

Music City Center 

Room: CC-107B 

Difference in differences (DID) is one of the most popular tools for causal inference in the social sciences. In this workshop, we introduce the different treatment-effect estimators based on the DID methodologies that are available in Stata. We cover the conceptual and theoretical foundations of DID models and demonstrate how to easily implement them with many practical examples using the software.

We start with the classic DID models implemented by commands -didregress- and -xtdidregress-, which estimate a treatment effect that is constant and homogeneous across all treatment cohorts in the study. We also introduce the postestimation tools available to test the validity of these models.

We then move to more complex DID models, in which treatment effects may be heterogeneous across treatment cohorts or may change with time. These models can be used to study, for example, the effect of job training programs on earnings or the effectiveness of COVID vaccines. To capture this heterogeneity, Stata has two commands that estimate treatment effects specific to each cohort and time period: -hdidregress- and -xthdidregress-. Both commands let you aggregate treatment effects by cohort and exposure to treatment and visualize these effects graphically. Tests of pretreatment parallel trends are also available.

Stata