Ashenfelter Dip
The empirical regularity that participants in training (and similar) programs experience a transitory decline in earnings just before enrollment. Because the same shock that depresses pre-period outcomes also drives selection into treatment, it violates Parallel-Trends: a pre-program baseline overstates the counterfactual and biases before/after and DiD estimates. The canonical example of why the choice of comparison period and selection model matters in quasi-experimental evaluation.
Relied on by
DiD threat analysis; motivates Conditional-Parallel-Trends and randomized designs.
Referenced by
- AshenfelterCard1985-LongitudinalEarnings
- Meyer1995-NaturalAndQuasiExperiments
- Abadie2005-SemiparametricDiD
New-papers pass (2026-07-04): FreyaldenhovenEtAl2019-PreEventTrendsPanelEventStudy (a general fix for endogeneity-induced pre-trends of exactly this kind); KahnLangLang2020-PromiseAndPitfallsOfDiD (level differences signal mechanisms that plausibly also affect trends).