Use this when the method has to prove both statistical behavior and empirical usefulness.
compactly in the main text; use the supplement for details.
work.
and convergence or failure diagnostics.
decision, or policy-relevant estimate.
Before drafting results, create a one-page map with these rows:
sensitivity, or coverage breakdowns.
The main text should report only the rows that teach the reader why the method works and when it fails.
Full grids belong in the supplement or replication package.
Track every reported number in a ledger:
| Manuscript item | Script | Seed/config | Output path | Runtime |
|---|---|---|---|---|
| Table/Figure X | ... | ... | ... | ... |
Use the ledger to decide what must be in the main replication path and what can remain optional.
EctJ referees read Monte Carlo sections as tests of the theory, and RES guidance caps the
main-text simulation summary near one page, so every theoretical claim needs exactly one matching
finite-sample exhibit:
| Theoretical claim | Required Monte Carlo evidence | Typical display |
|---|---|---|
| Asymptotic normality | Coverage of nominal 95% intervals across n | Coverage row per sample size |
| Size control of a test | Null rejection rates near 5% at the relevant boundary | Size table with nominal level in header |
| Local power gain | Power against the incumbent under drifting alternatives | One power figure |
| Rate or bias reduction | Bias and RMSE relative to the strongest competitor | Compact bias/RMSE panel |
| Tuning robustness | Behavior across bandwidth or penalty choices used in practice | Supplement grid, one-line main-text summary |
A theorem with no matching row invites the classic EctJ objection that the asymptotics carry no
finite-sample evidence; a simulation with no matching theorem is decoration to cut.
The other classic objection is a simulation design detached from the empirical illustration. Fix
it by calibration (illustrative numbers): if the application is a firm panel with N=180, T=12, and
residual serial correlation around 0.6, the core DGP should be N=200, T=12 with AR(1) errors at
rho in {0, 0.3, 0.6}, not an i.i.d. cross-section with n=10,000. State in the simulation preamble
which DGP parameters were estimated from the application data and which probe theoretical
boundaries. One calibrated design plus one boundary design beats six arbitrary grids at this
venue, and the pairing lets the empirical section reuse the simulation's vocabulary when it
explains why the new procedure changes the applied conclusion.
more when coverage is pushed to the third digit).
the EctJ replication policy makes these numbers checkable after conditional acceptance.
drops change rejection rates, and referees at this venue know it.
Run the battery, don't just enumerate it. Full map:execution-with-mcp. The Econometrics Journal is a methods venue — estimator validity + simulation; pair estimates with diagnostics.
romano_wolf (step-down FWER) or benjamini_hochberg.oster_delta / sensemakr.wild_cluster_bootstrap (few clusters), twoway_cluster / conley.audit_result(result_id) lists missing checks + the exactsuggest_function for each.
etable / did_summary_to_latex from the handle — no retyped numbers.Decisive checks in the body, exhaustive battery in the appendix.
JF execution walkthrough.
[Evidence readiness] strong / adequate / weak
[Monte Carlo role] <theory validation or stress test>
[Empirical application role] <applied-value demonstration>
[Missing baseline or diagnostic] <item>
[Next analysis] <single run or table>