Tables & Figures (jole-tables-figures)

When to trigger

• The main table has many columns and the paper is getting long
• You are unsure whether a result should be a table, a figure, or appendix-only
• Exhibits are pushing the manuscript toward or past the word limit
• Notes under tables/figures are not self-contained

JOLE's exhibit economy (this is the distinctive constraint)

JOLE does not normally publish articles exceeding 20,000 words, and it counts a full-page table or figure as 500 words toward that total. That rule changes how you design exhibits: every full page of tables/figures costs you 500 words of text. A paper with ten full-page exhibits has spent 5,000 words before a single sentence. So the discipline is:

• Promote only the headline exhibits to the body. The 1–3 results that carry the paper become main tables/figures; everything else goes to the online appendix.
• Consolidate. Combine related specifications into one well-organized table rather than several; a tight panel beats five thin ones.
• Prefer a figure when it carries the result more efficiently (event-study plots, RDD discontinuity plots, binscatters, wage-profile and decomposition charts) — but remember a full-page figure still costs 500 words, so it must replace, not duplicate, text.
• Self-contained notes. Each exhibit's note states the sample, the estimator, the SE clustering, and the units, so a referee can read it without hunting through the text.

Designing the main exhibits

1. Lead exhibit = the result. The single table or figure a labor referee should remember (e.g., the main treatment effect with its event-study, or the decomposition).
2. Units and magnitudes visible. Report effects in interpretable labor units (percent, percentage points, log points, dollars) with CIs.
3. No chartjunk. No 3D, minimal color, confidence bands shown; vector output (PDF/EPS) for print.
4. Numbering and call-outs in order. Every exhibit is referenced in the text in sequence.
5. Budget the pages. Tally how many full pages of exhibits you have and convert at 500 words/page; trim against the ~20,000-word cap.

Execution bridge (StatsPAI / Stata MCP)

Generate exhibits from the fitted result, not by retyping numbers. Full map: execution-with-mcp. JOLE is labor economics — the home of clean identification; DiD/IV/RDD and selection corrections are the binding constraint.

• 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

• Only 1–3 headline exhibits in the body; the rest in the online appendix
• Full-page exhibits counted at 500 words each against the ~20,000-word cap
• Related specifications consolidated, not scattered across many small tables
• Figures used where they carry the result more efficiently than prose
• Effects in interpretable labor units with confidence intervals
• Self-contained notes: sample, estimator, SE clustering, units
• Vector output; no chartjunk; exhibits numbered and called out in order

Anti-patterns

• A 12-column kitchen-sink main table that eats the word budget
• Many thin tables where one consolidated panel would do
• A full-page figure that merely restates a paragraph (it still costs 500 words)
• Notes that force the referee back into the text to learn the sample or SEs
• Forgetting that exhibits count toward the length cap and overrunning it
• Effects reported as bare coefficients with no interpretable units

Output format

【Headline exhibits】[1–3] in body, rest to appendix? [Y/N]
【Page budget】full pages of exhibits × 500 words = ___ ; within cap? [Y/N]
【Consolidation】related specs combined? [Y/N]
【Figure vs. table】each result in the more efficient form? [Y/N]
【Notes】self-contained (sample/estimator/SE/units)? [Y/N]
【Next step】jole-writing-style