Jegeo Identification

Identification Strategy (jegeo-identification)

When to trigger

• A spatial regression rests on OLS + region fixed effects, or TWFE on staggered place-based policy
• A quantitative-spatial / NEG model is estimated but it is unclear what in the spatial data identifies the key elasticities
• Treatment in one region plausibly spills over to "control" regions (SUTVA across space is violated)
• A qualitative/comparative-case paper makes a causal-sounding claim with no explicit logic of inference
• You are unsure the strategy reads as credible to BOTH an economist and a geographer

The JEG identification bar

Because JEG bridges geographical economics and human geography, "identification" means different things by branch — but in all of them the spatial structure of the data is part of the identification problem, not a nuisance. Two threats are nearly universal at JEG and referees expect them confronted head-on: spatial autocorrelation in errors (inference) and spatial spillovers / general-equilibrium leakage across units (SUTVA). Pick the branch and make the data-to-claim mapping explicit.

Branch A: Spatial causal design (place-based policy, regional treatment)

• Spatial DID / event study: with staggered adoption move beyond TWFE (Callaway–Sant'Anna, Sun–Abraham, de Chaisemartin–D'Haultfœuille); show clean event-study leads; report a Goodman-Bacon decomposition.
• Spillovers / SUTVA across space: the control region is often the treated region's neighbor. Use donut/ring specifications, model spatial spillovers explicitly, or argue why leakage is bounded — do not assume independence across adjacent units.
• Spatial RDD / border designs: geographic discontinuities (administrative borders) are powerful but demand a continuity argument across the border and attention to what else changes at it.
• IV with a spatial instrument: Bartik/shift-share and geography-based instruments are common; defend exogeneity of the shares (Goldsmith-Pinkham et al.) or of the shocks, not just first-stage strength.
• Inference: cluster at the spatial-treatment level AND address spatial correlation across clusters with Conley spatial-HAC standard errors; report how the cutoff distance was chosen.

Branch B: Quantitative-spatial / NEG model identification

• Name what identifies each structural elasticity (trade elasticity, agglomeration elasticity, migration elasticity) — tie it to specific spatial variation or moments, not "the estimator converged."
• Calibration vs. estimation: if elasticities are borrowed, say from where and show the counterfactual is not driven by an indefensible borrowed value; report sensitivity.
• General-equilibrium counterfactuals: the headline welfare/relocation number depends on the model's spatial linkages — show which parameters and which spatial structure move it.

Branch C: Case-based / qualitative geographic inference

• Make the logic of inference explicit: comparative cases, process tracing, or theory-building from a critical case — and state what would have falsified the claim.
• Justify case selection on substantive spatial grounds; address generalizability rather than claiming it.

Shift-share / Bartik instruments in a spatial setting

Shift-share instruments are pervasive in economic geography (regional exposure to national shocks via local industry mix), and JEG referees scrutinize them closely. Two defenses, two literatures:

• Exogenous shares (Goldsmith-Pinkham–Sorkin–Swift): identification rests on the pre-period industry shares being as-good-as-random; defend the shares' exogeneity and report the Rotemberg weights that show which industries drive the estimate.
• Exogenous shocks (Borusyak–Hull–Jaravel): identification rests on many quasi-random national shocks; defend the shocks and the equivalent shock-level regression.

State which justification you rely on — "we use a Bartik instrument" without naming the identifying assumption is exactly the move a JEG referee flags.

Execution bridge (StatsPAI / Stata MCP)

Estimate and audit the design, don't only describe it. Full map: execution-with-mcp. JEG is spatial economics — spatial dependence and sorting; emphasize identification and Conley/spatial-robust inference.

• detect_design → recommend → fit with as_handle=true → audit_result.
• Observational causal claims: staggered DiD (callaway_santanna / sun_abraham + bacon_decomposition + honest_did_from_result); IV (effective_f_test + anderson_rubin_ci); RDD (rdrobust + mccrary_test).
• Experiments: randomization-based inference + romano_wolf for many-outcome control.
• Sensitivity: oster_delta / sensemakr for observational claims.

Report the magnitude in interpretable units; route the full battery to the appendix. A run end-to-end (synthetic data, real returns) is in the JF execution walkthrough.

Checklist

• Branch chosen; the spatial-data-to-claim mapping stated in one sentence
• Spatial autocorrelation addressed in inference (Conley SEs / appropriate clustering; cutoff justified)
• Cross-unit spillovers / SUTVA across space confronted, not assumed away
• Staggered designs use a modern estimator; pre-trends/leads shown
• Structural: each key elasticity tied to identifying spatial variation; counterfactual sensitivity shown
• Qualitative: explicit inference logic + falsification condition + case-selection justification
• The claim never exceeds what the spatial design supports