| name | ecta-tables-figures |
| description | Use when finalizing the simulation tables and any empirical / illustrative exhibits in an Econometrica manuscript so each is self-contained, precise, and reproducible. Formats and audits exhibits; it does not design the Monte Carlo (use ecta-robustness) or run the estimation. |
Tables and Figures (ecta-tables-figures)
When to trigger
- A Monte Carlo or empirical table is overloaded, ambiguous, or not self-contained
- Notes do not say what is being reported (size? coverage? bias?) at what sample size and nominal level
- Figures (power curves, coverage plots, value-function illustrations) lack axis labels or a legend
- Numbers in the table cannot be traced to a specific script / seed
Econometrica exhibits are read closely by referees who will recompute or sanity-check
entries — and at conditional acceptance the ES Data Editor re-runs the code behind every
table, figure, and number that comes from data or simulation. Every table and figure must
stand alone and trace to a seeded script in the replication package. Two Econometrica
specifics that differ from applied siblings: (1) the 45-page main-text limit (incl.
references and appendices) forces ruthless triage — keep the headline size/power/coverage
tables in the body and push secondary designs to the Supplemental Material; (2)
reproducibility is enforced by the Econometric Society Data Editor with a Zenodo
deposit, not the AEA/openICPSR pipeline the AER/AEJ packs assume.
Simulation-table standards
- Self-contained. The caption + notes must let a reader interpret every number without
the body text: estimator(s), DGP / design, sample size(s), number of replications, nominal
level, and what each column reports (bias, RMSE, empirical size, coverage, length).
- One concept per table where possible — do not mix size, power, and point-estimation
quality in one undifferentiated block.
- Consistent precision. Choose decimal places deliberately (e.g., sizes to 3 decimals so
0.052 vs. 0.048 is visible) and keep them consistent down a column.
- Highlight the comparison. The reader should see at a glance that the proposed method
controls size / has lower RMSE than the competitor — order columns/rows so the contrast reads.
- Monte Carlo error. Where a size or coverage near the nominal level carries the argument,
make the simulation precision clear (replication count in notes, or a reported MC standard error).
- Three-line / booktabs style. Clean rules (top, mid, bottom), no vertical lines, no
shading — the standard for formal economics journals.
Figures
- Label everything: axes with units, every series, the nominal-level reference line on a
size/coverage plot, the null on a power curve.
- Power curves: plot rejection probability against the departure parameter; mark the null
and the nominal level; use size-adjusted curves for fair comparison.
- Coverage plots: show coverage across designs / sample sizes against the nominal target.
- Theory illustrations: equilibrium / value function / comparative statics across the
parameter range, with the relevant region annotated.
- Vector format (PDF/EPS) so curves stay crisp; legible at print size; colorblind-safe and
readable in grayscale.
Traceability
- Each table/figure should be regenerated by a named script with a fixed seed (tie this to
ecta-replication-package). Keep a map: Table N ↔ script ↔ seed ↔ output file.
- Numbers in the text must match the tables exactly (no rounding drift between prose and table).
Execution bridge (StatsPAI / Stata MCP)
Generate exhibits from the fitted result, not by retyping numbers. Full map:
execution-with-mcp. Econometrica publishes econometric theory and applied micro; the chain below serves its applied/empirical papers (weak-IV-robust and modern-DiD reporting expected) — pure theory uses its own apparatus.
- Tables:
etable (multi-model) or did_summary_to_latex straight from the result_id.
- Figures:
plot_from_result / enhanced_event_study_plot / event_study_table —
axis units and the SE/clustering note baked in.
- Every note names the estimator + clustering and states the magnitude in interpretable units.
See a full fitted-result → exhibit chain in the JF execution walkthrough.
Checklist
Anti-patterns
- A size table with no nominal level stated, so 0.07 is uninterpretable
- Mixing size, power, and bias in one block with no clear separator
- Reporting power without indicating size, making the method look artificially good
- Inconsistent decimals down a column, hiding the comparison
- A figure with unlabeled axes or an unexplained second series
- Numbers in the text that disagree with the table by a digit
- A table no script can regenerate, so replication review cannot verify it
Output format
【Exhibit】Table/Figure N — [what it shows]
【Self-contained?】yes / no — missing: [...]
【Reports】bias / RMSE / size / power / coverage / length / illustration
【Design+n+reps+level stated】yes/no
【Style】booktabs three-line / needs fixing
【Traceable to script+seed】yes/no
【Next step】ecta-writing-style
