Difference-in-Differences (DiD) Checklist

A comprehensive checklist for conducting Difference-in-Differences analysis in econometrics and causal inference.

Purpose: To systematically guide researchers through the process of conducting a Difference-in-Differences analysis, ensuring key steps and considerations are addressed.

Background: This checklist is inspired by a tweet from @pedrohcgs, which outlined essential steps in DiD analysis. I’ve expanded on this to create a more comprehensive, actionable checklist for researchers and analysts.

Download: You can download this checklist as a Markdown file to use offline or integrate into your own research workflow.

Checklist

1. Prepare Data and Visualize Treatment
   • Plot treatment rollout (use panelView R package)
   • Record number of units treated per cohort
   • Plot average outcomes over time by cohort
2. Select Comparison Groups
   • Assess Parallel Trends assumption
   • Consider who decides treatment and what they know
   • Determine allowable types of selection
3. Conduct Initial Analysis
   • Perform event study (no covariates) to assess Parallel Trends
4. Refine Analysis with Covariates
   • If needed, add covariates affecting untreated outcome growth
   • Check for overlap in covariates
   • Use regression adjustment DiD if extrapolating
   • Rerun event study with covariates to reassess Parallel Trends
5. Sensitivity Analysis
   • Perform sensitivity analysis for Parallel Trends violations
   • Consider using honestDiD R package
6. Alternative Methods
   • Explore other methods if Parallel Trends remains implausible

Additional Resources

• Difference-in-Differences (NBER)
• A Practical Guide to Difference-in-Differences (World Bank)

Action: Use this checklist to ensure a thorough and methodologically sound Difference-in-Differences analysis. Document your decisions and findings at each step to support your research conclusions.