Panel-data analysis and difference-in-differences modeling: An applied Stata workshop

Session Submission Type: Workshop

Description

Panel-data methods, often referred to as longitudinal methods, are widely used to analyze repeated observations on individuals, households, firms, countries, or other units over time. By exploiting both cross-sectional and time-series variation, panel data allow researchers to account for unobserved heterogeneity and evaluate changes that cannot be identified using cross-sectional data alone.

This workshop provides an applied introduction to panel-data analysis and difference-in-differences (DID) methods using Stata. We begin by reviewing core panel-data concepts, including fixed-effects, random-effects, and correlated random-effects models for addressing unobserved individual-level heterogeneity. The session covers both linear and nonlinear panel-data models, with an emphasis on practical implementation, model comparison, and interpretation, including the use of specification tests to guide modeling choices. Throughout the workshop, empirical applications are illustrated using hands-on examples in Stata.

After establishing an understanding of panel-data models, the workshop turns to causal inference using DID designs, a widely used approach for estimating treatment effects in applied social science research, including policy evaluation and program analysis. We demonstrate how to implement DID models in Stata using the -didregress- and -xtdidregress- commands for repeated cross-sectional and panel data. The session emphasizes the key identification assumption underlying DID, the parallel trends assumption, and provides hands-on examples showing how to estimate and interpret the average treatment effect on the treated (ATET).

Recent advances in the DID literature have emphasized treatment-effect heterogeneity, recognizing that the timing and impact of interventions may vary across groups and over time. To reflect this development, the workshop also introduces Stata's -hdidregress- and -xthdidregress- commands for estimating heterogeneous ATETs. Participants will explore the intuition behind these estimators and learn to use postestimation tools to aggregate, visualize, and assess the robustness of heterogeneous treatment effects.

Sub Unit

• Workshops / Teaching Workshop

Presenter

• Chris Cheng, StataCorp

Session Organizer

• Brooke Johnson, StataCorp