SDS Seminar Series – Rafael Campello de Alcantara, University of Texas at Austin

Oct

2025

2:00 pm – 3:00 pm

In Person (view details)

Featured Speaker(s): Rafael Campello de Alcantara

Cost: Free

Searching for Parallel Trends: A Decision Tree Algorithm for Discovering Conditional Diff-in-Diff Estimators

Description

The Fall 2025 SDS Seminar Series continues on October 3rd from 2:00 p.m. to 3:00 p.m. with Dr. Rafael Campello de Alcantara (Postdoctoral Fellow, Department of Statistics and Data Sciences, University of Texas at Austin). This event is in-person in the Avaya Room (POB 2.302).

Title: Searching for Parallel Trends: A Decision Tree Algorithm for Discovering Conditional Diff-in-Diff Estimators

Abstract: Difference-in-differences (DiD) designs represent one of the most common empirical strategies for learning treatment effects in the social sciences. The most basic DiD design consists of two periods and two groups. In the first period, both groups are not treated, while only one of the groups receive treatment in the second period. Identification in that setting relies crucially on the parallel trends assumptio: on average, the outcomes of the treated and untreated groups evolve following the same trend. Typically, two versions of PTA are considered in the literature. On one extreme, PTA is invoked unconditionally. On the other extreme, much of the literature invokes PTA for every possible covariate value. We discuss identification in the intermediate ( and potentially more realistic) case where PTA holds only for some subsets of the covariate space. Importantly, in that scenario, it is likely that none of the commonly targeted causal estimands are identified. Then, we draw on placebo tests for DiD to design an algorithm that uses auxiliary data to search for regions of the covariate space where PTA can be assumed to hold. The regions flagged by our algorithm can then be used to learn about conditional effects on the data set we are interested in.