Identification Strategy (jme-identification-strategy)

When to trigger

• The "monetary shock" is a raw policy-rate change with no exogeneity argument
• A SVAR uses a recursive ordering that referees will dispute
• High-frequency surprises are not separated from the Fed "information effect"
• A DSGE mechanism is asserted but not identified from the data/moments
• You are unsure the design clears the JME bar for a credible macro claim

The JME identification bar

JME is monetary economics and macroeconomics, so "identification" means isolating an exogenous shock or pinning down a structural mechanism, not a micro treatment effect. With single anonymized review and at least two specialist referees, the identifying assumption must be legible and defended in the registers macroeconomists expect. The credibility ladder the field implicitly applies (strong → weaker):

1. High-frequency identification — monetary surprises in a narrow window around FOMC announcements, with the pure policy shock separated from the information/forward-guidance component (Nakamura–Steinsson, Miranda-Agrippino–Ricco, Jarociński–Karadi-style decompositions).
2. Narrative shocks — Romer–Romer-style records of policy intent, orthogonalized to Greenbook/Tealbook forecasts; narrative fiscal shocks (military news, tax changes).
3. Proxy / external-instrument SVAR — the surprise series as an instrument for the policy rate inside a VAR; report relevance and the implied IRFs.
4. Sign / long-run restrictions — theory-consistent restrictions when timing or IV are unavailable; report robustness to the restriction set.
5. Model-based (quantitative) identification — discipline structural parameters with micro moments and matched second moments; report identification diagnostics (e.g., Iskrev) and prior sensitivity.

Branch paths

Branch A: High-frequency / narrative monetary shocks

• Define the event window precisely; show the surprise is news, not anticipated.
• Strip the information effect: a contractionary surprise that raises expected output signals the central bank's private information — decompose it.
• Aggregate to the desired frequency carefully; discuss attenuation and measurement error.

Branch B: SVAR / proxy-SVAR

• State the identification scheme (recursive, sign, long-run, IV/proxy) and defend it.
• Report instrument relevance (for proxy-SVAR), IRF confidence bands, lag selection, and stability.
• Show robustness to ordering / restriction choices; report FEVDs.

Branch C: Local projections (Jordà)

• Compare LP and VAR IRFs; LP is more robust to misspecification but noisier.
• Use lag-augmentation; LP-IV when using an external instrument; HAC/Newey–West or clustered SEs.

Branch D: DSGE / quantitative mechanism

• Identify the new mechanism from data: which moments move when the friction is on vs. off?
• Report Bayesian estimation diagnostics (convergence, identification, posterior predictive) or calibration targets.
• Show the policy conclusion changes because of the mechanism, not an unrelated assumption.

Execution bridge (StatsPAI / Stata MCP)

Estimate and audit the design, don't only describe it. Full map: execution-with-mcp. JME is monetary macro — SVAR, local projections, high-frequency identification; local_projections/irf are in StatsPAI, DSGE/calibration is outside this toolchain.

• 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

• The shock/mechanism named in one sentence and defended as exogenous/identified
• Monetary surprises separated from the information effect (if high-frequency)
• SVAR identification scheme stated and robustness shown
• LP-vs-VAR comparison reported where relevant
• DSGE identification diagnostics / calibration targets reported
• The claim never exceeds what the identification supports

Output format

【Approach】high-frequency / narrative / proxy-SVAR / LP / sign-restriction / model-based
【Shock or mechanism】one sentence
【Exogeneity / identification argument】one line
【Information-effect handling】Y/N (if HF)
【Robustness done】[ordering, restriction set, LP-vs-VAR, identification diagnostics, ...]
【Next step】jme-data-analysis