Economics
Saved & recent

Randomized experiments

Random assignment creates comparable treatment groups in expectation. Differences in outcomes can estimate assignment effects, while attrition, noncompliance and spillovers need separate consideration.

Read first
Counterfactual outcomes
Related
Causal identification, Sampling uncertainty and confidence intervals, External validity and policy transfer
Further reading
World Bank: Research design for impact evaluation
100 volunteers random draw 50 assigned offer 45 observed 45 · missing 5 50 assigned control 48 observed 48 · missing 2 comparison by original assignment
Hypothetical allocation. Randomization creates baseline comparability in expectation, not identical realized groups or a representative population. No effect is calculated from these incomplete outcomes; selective attrition, noncompliance and spillovers require separate checks.

View in the full Economics collection