Econometric Methods for Program Evaluation

Causal Question / Estimand

A survey of the methods that identify the ATT (and related parameters) of programs and policies — organizing the credibility-revolution toolkit by the identifying assumption each method invokes.

Identification Strategy

Reviews the main designs and what licenses each: selection-on-observables (Ignorability + Overlap, via regression, matching, and Propensity-Score weighting); regression discontinuity; difference-in- differences and Parallel-Trends; synthetic control; and instrumental variables. The organizing theme is that credible program evaluation comes from a transparent, defensible identification strategy, not from estimation machinery.

Key Assumptions

Covers the field’s assumptions: Ignorability, Overlap, Parallel-Trends, Continuity-at-Cutoff, Convex-Hull-Restriction, Exclusion-Restriction, all on a Potential-Outcomes/SUTVA base.

Threats to Validity

Method-specific: unobserved confounding (selection on observables), manipulation (RD), non-parallel trends (DiD), poor fit / extrapolation (synthetic control), invalid or weak instruments (IV).

Setting / Data

n/a — methodological review (econometrics).

Key Claims

  • Modern program evaluation is organized around identification strategies, each with an explicit, often untestable, key assumption.
  • No method dominates; choice depends on the assignment mechanism and available data.
  • Estimation is secondary to a credible identification argument.

Connections

Citation

Abadie, A., & Cattaneo, M. D. (2018). Econometric Methods for Program Evaluation. Annual Review of Economics, 10, 465–503.