Empirical Strategies in Labor Economics
Causal Question / Estimand
A survey of how labor economists estimate causal relationships (returns to schooling, union wage effects, immigration, military service, class size). It coins/popularizes the term “empirical strategy” — a research plan spanning identification, data collection, and measurement — and targets well-defined treatment effects (Causal-Estimand), with attention to for whom an estimate is valid (LATE).
Identification Strategy
Organizes applied work around an “experimentalist” perspective: sharply distinguish causal variables from control variables and outcomes, then find a research design that makes the causal variable as-good-as-randomly assigned. Reviews four identification strategies — control for confounders (Ignorability), fixed-effects / difference-in-differences, instrumental variables, and regression discontinuity — illustrating each with a labor example.
Key Assumptions
Potential-Outcomes, Ignorability (selection on observables / control strategies), Randomization (the benchmark and the as-good-as-random ideal), Parallel-Trends (fixed-effects/DiD), Exclusion-Restriction and Monotonicity with the LATE interpretation (IV), and Continuity-at-Cutoff (RDD).
Threats to Validity
Omitted-variables bias in control strategies; weak and invalid instruments; measurement error (a major focus — reliability of schooling, earnings, and union status data); and the gap between internal and external validity.
Setting / Data
n/a — methodological survey. Recurring empirical illustrations: schooling/returns to education, unions, immigration (Mariel boatlift), Vietnam-draft military service, and class size (Project STAR). Also surveys secondary, primary, and administrative data sources and survey weighting.
Key Claims
- “Empirical strategy” = identification + data + measurement; identification is shorthand for research design.
- The experimentalist program: name the causal variable, find exogenous variation, defend the design’s assumptions explicitly.
- Measurement and data quality are first-class concerns, not afterthoughts.
- Different designs answer subtly different questions (e.g. IV recovers a complier-weighted effect, LATE).
Connections
- See also: Angrist2022-EmpiricalStrategiesIlluminatingPath (Angrist’s Nobel lecture extends the IV/LATE argument), AngristPischke2010-CredibilityRevolution, Meyer1995-NaturalAndQuasiExperiments, Foundations
- Builds on: Potential-Outcomes framework (Holland1986-StatisticsAndCausalInference)
- Foreshadows: the AshenfelterCard1985-LongitudinalEarnings fixed-effects/DiD lineage
Citation
Angrist, J. D., & Krueger, A. B. (1999). Empirical Strategies in Labor Economics. In O. Ashenfelter & D. Card (Eds.), Handbook of Labor Economics (Vol. 3A, pp. 1277–1366). Elsevier.