---
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

1. **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).
2. **One concept per table** where possible — do not mix size, power, and point-estimation
   quality in one undifferentiated block.
3. **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.
4. **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.
5. **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).
6. **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).

## Checklist

- [ ] Caption + notes make each exhibit fully self-contained
- [ ] DGP/design, n, replications, nominal level, and per-column meaning all stated
- [ ] Decimal precision deliberate and consistent
- [ ] Competitor methods included; the key comparison reads at a glance
- [ ] Monte Carlo precision conveyed where size/coverage carries the argument
- [ ] Three-line / booktabs style; no vertical rules or shading
- [ ] Figures fully labeled; vector format; readable in grayscale
- [ ] Every exhibit traces to a named script + seed; text numbers match tables

## 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
```