Shift-Share Instrument

A shift-share (or “Bartik”) instrument averages a common set of shocks (“shifts”, e.g. national industry growth rates) using unit-specific exposure weights (“shares”, e.g. local industry employment shares): the regional instrument is ∑_k share_ik · shock_k. It supplies as-good-as-random variation in a treatment (local labor demand, import exposure, immigrant supply) that differs across units only because they were differentially exposed to the same aggregate shocks. Identification can be argued to come from exogenous shares (a pooled-exposure design) or from exogenous shocks (quasi-random shifters with endogenous shares) — a distinction that is itself the main methodological debate. Validity rests on an Exclusion-Restriction-type argument at the level (share or shock) taken as exogenous; relevance requires that units differ enough in exposure.

Relied on by

IV designs in labor, trade, public, macro, development, and finance economics wherever a regional/group treatment is instrumented by exposure-weighted aggregate shocks. Origin: Bartik (1991) and BlanchardKatz1992-RegionalEvolutions.

Referenced by

New-papers pass (2026-07-20): BlanchardKatz1992-RegionalEvolutions (industry-mix instrument for labor demand), Card2001-ImmigrantInflows (ethnic-enclave supply-push instrument), AutorDornHanson2013-ChinaSyndrome (China import-exposure instrument), GoldsmithPinkhamEtAl2020-BartikInstruments (share-exogeneity design + Rotemberg weights), BorusyakEtAl2022-QuasiExperimentalShiftShare (shock-exogeneity design), AdaoEtAl2019-ShiftShareDesigns (inference under share-induced cross-regional correlation).