Causal inference Part II is a 4-day workshop in design based causal inference series. It will cover three contemporary research designs in causal inference -- difference-in-differences, synthetic control and matching/weighting methods -- as well as introduce participants to causal graphs developed by Judea Pearl and others. Each day is 8 hours with 15 minute breaks on the hour plus an hour for lunch. We will review the theory behind each design, go into detail on the intuition of the estimation strategies and identification itself, as well as explore code in R and Stata and applications using these methods. The goal as always is that participants leave the workshop with competency and confidence. This class will be a sequel to the 4-day workshop on Causal Inference Part I.
Basic Difference-in-Differences
Slides
Introducing the fundamentals of DiD
Code
Google Spreadsheet for simple DiD Calculations
Readings
Difference-in-Differences Estimation with and without Covariates
Slides
Introducing OLS and various estmators with covariate adjustments
Code
Readings
Outcome regression (Heckman, Ichimura and Todd 1997),
Inverse probability weight estimator (Abadie 2005),
Doubly robust (Sant'Anna and Zhao 2020)
Twoway Fixed Effects and Bacon-Decomposition
Slides
Code
Fixed effects and Pooled OLS example,
Shiny App for Bacon Decomposition
Readings
Callaway and Sant'Anna
Slides
Code
Readings
Sun and Abraham
Slides
Code
Readings
Imputation Estimators
Slides
Code
Two-stage DID
Robust efficient imputation estimator
Readings