CATE and LATE in Stata: Conditional Average Treatment Effects and Local Average Treatment Effects

In this workshop, we present two of the latest features for causal inference added into Stata 19: Conditional
average treatment-effects (CATE) and local average treatment-effects (LATE) estimation. We cover the assumptions of both models, discuss when to use which, and demonstrate how to easily implement them with many practical examples using the software.

We begin with the estimation of heterogeneous treatment effects using the cate command. We discuss the different types of heterogeneity that can be estimated and the available estimators for each. We also show how to use postestimation tools to visualize and test treatment-effect heterogeneity, study how treatment effects change with
covariates, and evaluate alternative treatment-assignment policies.

We then move to estimating treatment effects when some individuals do not comply with the treatment to
which they were assigned. Because the decision to undertake treatment might be affected by unobserved characteristics, treatment-effect estimates from traditional methods may be biased. With the lateffects command, we show how to consistently estimate the LATE or average treatment of compliers. We finalize the workshop by showing how to use the postestimation tools to verify the assumptions of the LATE model.