Economics
Saved & recent

Difference-in-differences

Difference-in-differences compares outcome changes in treated and comparison groups. A causal interpretation requires parallel untreated trends and absence of confounding differential shocks.

Read first
Causal identification
Related
Natural experiments, External validity and policy transfer
Further reading
World Bank: Impact Evaluation in Practice (2011), chapters 5–6
-1 0 1 15 20 25 30 35 40 45 50 55 period outcome / currency per person treatment begins treated 50 counterfactual 40 control 30 (50 − 35) − (30 − 25) = 10
Hypothetical group means. Causal interpretation requires parallel untreated trends, no differential coincident shock, and suitable stable composition. Similar pretreatment trends are a diagnostic, not proof of the missing post-treatment counterfactual.

View in the full Economics collection