Jmcb Identification

Identification Strategy (jmcb-identification)

When to trigger

• A monetary-policy "shock" is just a residual from a regression, with no defense against the information effect
• An SVAR's ordering / sign restrictions are unstated or unmotivated by theory
• A bank/loan-level causal claim rests on OLS + controls, or TWFE on staggered regulation
• A banking-model parameter is "calibrated" with no statement of what in the data pins it
• You are unsure the identification clears JMCB's policy-relevant, mechanism-explicit bar

The JMCB identification bar

JMCB referees are monetary/banking economists who judge identification by whether the data-to-object mapping is explicit, theory-disciplined, and robust to the field's known critiques. Because the journal spans macro and micro, "identification" means different things by engine — pick the engine and make the argument transparent, with timing and institutions on the page. The dominant failure mode JMCB punishes is a "shock" or "treatment" that is not exogenous to the thing it is supposed to move.

Branch paths

Branch A: Macro shock identification (SVAR / narrative / external instruments)

• High-frequency monetary surprises: identify off tight windows around FOMC/ECB announcements (Kuttner; Gürkaynak–Sack–Swanson factors; Gertler–Karadi external instrument). Defend against the information effect (Nakamura–Steinsson; Jarociński–Karadi sign decomposition or poor-man's sign restrictions) — a surprise that moves rates and raises growth forecasts is not a clean policy shock.
• SVAR ordering / sign restrictions: motivate every restriction from theory or institutions; report sensitivity to the ordering and to the restriction set; for set-identified models report the full identified set, not a single draw.
• Narrative shocks (Romer–Romer style): document the construction, the information set, and orthogonalization to forecasts.
• IRF inference: report bands that match the method (e.g., bias-corrected/bootstrap for VARs; lag-robust / HAR for local projections); state the local-projection vs. VAR choice and why.

Branch B: Structural monetary / banking model identification

• Name what identifies each parameter — tie it to a moment or institutional fact, not "the estimator converged." For a banking-friction model, say which data feature (loan spreads, deposit betas, capital ratios) disciplines the friction.
• Targeted vs. untargeted moments: report fit to targeted moments and validate on untargeted ones.
• Policy-invariance: argue the parameters survive the counterfactual you run (Lucas critique) — central to JMCB's policy use.
• Report estimation regularity (objective, multi-start, Monte Carlo recovery) when estimated rather than calibrated.

Branch C: Micro-banking causal design (bank / loan / firm panel)

• Staggered policy/regulation: move beyond TWFE (Callaway–Sant'Anna, Sun–Abraham, de Chaisemartin–D'Haultfœuille); show clean event-study leads; Goodman-Bacon decomposition.
• Khwaja–Mian style: absorb credit demand with firm×time (or industry×region×time) fixed effects to isolate bank-driven credit supply — the workhorse banking identification; state explicitly that you net out demand.
• IV / Bartik credit-supply shifters: strong first stage (effective F); defend exclusion; for shift-share defend share or shock exogeneity.
• Inference: cluster at the treatment-assignment level (often bank and time, two-way); address few-cluster issues (wild-cluster bootstrap); Driscoll–Kraay if cross-sectional dependence is a concern.

Referee pushback mapped to the identification fix

• "Your monetary-policy shock is contaminated by the information effect." → Decompose the surprise into pure-policy and information components (Jarociński–Karadi sign restrictions, or orthogonalize to Greenbook/Blue-Chip forecast revisions); show the IRF is robust to the cleaned shock.
• "The SVAR result is an artifact of your recursive ordering." → Report the headline IRF under alternative orderings and a theory-motivated sign-restriction scheme; for set-identified models show the full identified set.
• "You measure credit demand, not supply." → Absorb demand with firm×time (or industry×region×time) fixed effects, à la Khwaja–Mian; report the within-firm cross-bank coefficient.
• "Staggered TWFE is biased here." → Re-estimate with Callaway–Sant'Anna or Sun–Abraham; show flat event-study leads and a Goodman-Bacon decomposition.
• "Your calibrated parameters won't survive the counterfactual." → Argue policy-invariance (Lucas critique); show the parameters are not functions of the policy you change.
• "Standard errors are understated." → Two-way cluster (bank and time) or wild-cluster bootstrap with few clusters; Driscoll–Kraay for cross-sectional dependence.

Matching identification to the JMCB policy use

A JMCB identification argument is not complete until it connects to the policy question. A clean shock that identifies an IRF still needs the reader to see which policy lever the response speaks to; a clean bank-supply coefficient still needs the reader to see what it implies for transmission or regulation. State the estimand in policy-relevant units (a transmission elasticity, a pass-through, a welfare number) so the identification serves the contribution rather than standing alone as a methodological exercise.