Recent Developments in the Econometrics of Program Evaluation

Causal Question / Estimand

A comprehensive survey of estimating treatment effects (ATE, ATT, LATE, quantile effects) across the major identification strategies, in the Potential-Outcomes framework.

Identification Strategy

Synthesizes the field under two broad regimes. (1) Unconfoundedness / selection on observables (Ignorability + Overlap): regression, Propensity-Score methods, matching, weighting, and doubly-robust estimators, with extensive treatment of overlap diagnostics and efficiency. (2) Designs relaxing unconfoundedness: instrumental variables / LATE, regression discontinuity (Continuity-at-Cutoff), and difference-in-differences (Parallel-Trends). Emphasizes assessing assumptions and matching the estimator to the design.

Key Assumptions

Ignorability, Overlap, Parallel-Trends, Continuity-at-Cutoff, Exclusion-Restriction, Monotonicity — on a Potential-Outcomes/SUTVA foundation.

Threats to Validity

Limited overlap and misspecification (selection on observables); the untestability of unconfoundedness; design-specific failures for IV/RD/DiD. Stresses overlap checks and robustness across estimators.

Setting / Data

n/a — methodological survey (the canonical JEL reference for program evaluation).

Key Claims

  • A unified potential-outcomes account links matching, weighting, regression, IV, RD, and DiD as answers to different assignment mechanisms.
  • Under unconfoundedness, many estimators target the same parameter; overlap is the key practical constraint and doubly-robust methods add protection.
  • The credibility of any estimate rests on the plausibility of its (often untestable) identifying assumption.

Connections

Citation

Imbens, G. W., & Wooldridge, J. M. (2009). Recent Developments in the Econometrics of Program Evaluation. Journal of Economic Literature, 47(1), 5–86.