Difference‑in‑Differences (DID) estimates the causal effect of an intervention by comparing changes over time in a treated group to changes over time in a comparable control group.
Core Assumption: Parallel Trends
In the absence of treatment, treated and control groups would have followed similar trends. We test this using pre‑trend diagnostics and event‑study plots.
Designs
- Single intervention date: classic two‑period DID or multi‑period DID
- Staggered adoption: groups adopt at different times; use modern estimators
Estimators
- Two‑way fixed effects (TWFE) for simple settings
- Sun–Abraham or Callaway–Sant’Anna for staggered adoption
Diagnostics
- Event‑study coefficients (pre‑treatment should be near zero)
- Placebo periods where applicable
- Robust SEs clustered at the unit or group level
Interpretation
Report the average treatment effect and show the time path of effects. State the population and period the estimate applies to and discuss sensitivity to specification.