ectj-identification-strategy

EctJ Identification Strategy

Use this for theory and methods integrity. EctJ readers will tolerate compactness, but not

hidden assumptions or vague asymptotic claims.

Audit

• State the population object, identifying restrictions, estimator or test statistic, and

target parameter before derivations.

• Label each regularity condition by role: existence, identification, consistency,

asymptotic normality, bootstrap validity, finite-sample approximation, or computation.

• Show why the leading case is not a toy example; connect assumptions to the empirical

application.

• Keep proofs in the main text or printed appendix when current RES guidance requires that;

do not park mathematical proofs only in the online appendix.

• Separate theorem statements from implementation advice and simulation claims.
• Flag any assumption that is convenient but empirically fragile.

Referee attack surface

EctJ referees usually attack the bridge between compact theory and practical use. Pre-answer these points:

• Object drift: the target parameter in the theorem is not the object estimated in the application.
• Assumption opacity: a regularity condition is stated but never tied to a data feature or estimator step.
• Leading-case weakness: the theorem solves a toy case whose constraints make the empirical example

irrelevant.

• Proof placement risk: critical derivations are hidden in unreviewed online material or an untraceable

appendix.

• Simulation mismatch: the Monte Carlo design does not probe the assumption most likely to fail.

For each attack, write the exact theorem, assumption, table, or paragraph that will answer it.

Assumption ledger

Create a compact ledger before rewriting the theory section:

Condition | Role | Where used | Empirical/simulation check | If weakened

Use the ledger to remove decorative assumptions and expose missing ones. If a condition is used only for

proof convenience, say whether it can be relaxed, whether it is standard in the closest EctJ-adjacent

literature, and whether the simulation explores failure near that boundary. If a condition is essential

but empirically unverifiable, the paper needs an interpretation paragraph that tells applied readers what

kind of data-generating process would make it plausible.

Do not let notation hide the identification argument. A reader should be able to trace, in order, the

target object, restrictions, estimator or statistic, asymptotic claim, and finite-sample diagnostic.

Worked trace: a debiased panel treatment-effect estimator

A hypothetical EctJ vignette (illustrative throughout): the paper proposes an orthogonalized

estimator for an average treatment effect in a panel where nuisance functions are fit by machine

learning. The traceable chain referees expect:

• Target object: the ATE under unconfoundedness conditional on high-dimensional firm controls.
• Restrictions: overlap bounded away from zero; nuisance estimators converging faster than

n^{-1/4}; cross-fitting with K=5 folds.

• Estimator: the Neyman-orthogonal score averaged over folds.
• Asymptotic claim: root-n normality with a variance estimator valid under cross-fitting.
• Finite-sample diagnostic: coverage simulated at n in {250, 1000}; the rate condition is

stressed by deliberately slowing one nuisance learner and showing where coverage degrades.

If any link is missing, that link is what the report will quote back. A rate condition of the

n^{-1/4} kind is exactly the assumption that must be tied to a data feature: say which learner

plausibly meets it in the application and what the simulation shows when it fails.

Proof-economy rules for the compact format

• Every theorem keeps its full proof in the printed paper or printed appendix; the online appendix

carries only secondary lemmas, and only when current RES guidance permits it — confirm against

the journal's current author guidelines before moving any derivation out of print.

• A leading-case theorem may delegate generality to a remark, but the remark must say what breaks

in the general case, not just that extensions are straightforward.

• Each asymptotic statement should name the exhibit where its finite-sample counterpart appears;

EctJ referees treat unpaired asymptotics as an unfinished result.

Execution bridge (StatsPAI / Stata MCP)

Estimate and audit the design, don't only describe it. Full map:

[execution-with-mcp](../../../shared-resources/empirical-methods/execution-with-mcp.md). The Econometrics Journal is a methods venue — estimator validity + simulation; pair estimates with diagnostics.

• 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](../../../Journal-of-Finance-Skills/resources/worked-examples/02-execution-walkthrough.md).

Output format

[Identification status] defensible / needs repair / not ready
[Target object] <parameter, estimator, test, or procedure>
[Critical assumptions] <condition -> role>
[Proof gaps] <missing lemma, rate, regularity, or edge case>
[Applied connection] <how the application validates the setup>