Regional Evolutions

Causal Question / Estimand

How does a U.S. state’s labor market adjust to a shock to labor demand? The object is not a single ATE but a set of dynamic responses (impulse responses from a VAR) of employment, unemployment, relative wages, and migration to an innovation in labor demand.

Identification Strategy

The empirical problem is separating movements driven by labor demand from those driven by labor supply. Two identification devices do the work. (1) Sign/timing restrictions in the VAR: a positive demand shock raises employment, lowers unemployment, and raises wages together, whereas a supply shock moves them differently. (2) An external instrument for the demand innovation — the industry-mix (proto-Bartik) instrument: the employment growth a state would have experienced given its fixed two-digit sectoral composition had each sector grown at its national rate (∑_k share_ik · nationalgrowth_k), used in deviation from the aggregate national rate. This is a Shift-Share-Instrument. They also use military-contract (defense-spending) shocks as an alternative demand instrument.

Key Assumptions

• Shift-Share-Instrument — the industry-mix predicted-growth term isolates demand-driven variation.
• Exclusion-Restriction — national industry growth rates are exogenous to the individual state (a small state does not drive national sectoral growth), so the predicted term affects state outcomes only as a demand shifter.
• Relevance / first stage — the instrument is informative only because states differ enough in sectoral employment composition.

Threats to Validity

National industry growth rates may embed aggregate shocks correlated with a state’s own supply conditions; large states can influence national aggregates; fixed base-year shares may proxy for slow-moving local trends. These are the same concerns later formalized by the shift-share methods literature.

Setting / Data

U.S. states over roughly 40 postwar years (~1950–1990): state employment, unemployment, hourly manufacturing wages, migration, and personal income; two-digit industry employment shares; defense procurement data.

Key Claims

• Employment shocks have permanent effects on the level of state employment (states return to a common growth rate on a permanently shifted path), while their effects on relative unemployment and wages are transitory (dissipating within roughly a decade).
• The dominant margin of long-run adjustment is worker out-migration, not local job creation or wage adjustment.
• Introduces the industry-mix labor-demand instrument that (with Bartik 1991) became the canonical shift-share design.

Connections

• Origin (with Bartik 1991) of the Shift-Share-Instrument; both are named as the founding applications in GoldsmithPinkhamEtAl2020-BartikInstruments and BorusyakEtAl2022-QuasiExperimentalShiftShare.
• Same instrument family, applied to immigration, in Card2001-ImmigrantInflows and to trade in AutorDornHanson2013-ChinaSyndrome.
• See also IV. (Bartik 1991 is a book — flagged for a dedicated pass, not yet noted.)

Citation

Blanchard, O. J., & Katz, L. F. (1992). Regional Evolutions. Brookings Papers on Economic Activity, 1992(1), 1–75.