Tables & Figures — Exhibits that Carry the Policy Message (aejpol-tables-figures)

When to trigger

• Tables are dense, over-decorated with significance stars, or hard to read
• The paper lacks one exhibit a reader could take away as the policy result
• Figures show coefficients but not the policy-relevant magnitude or its uncertainty
• You are preparing the final exhibit set for an AEA submission

AEA / AEJ: Policy exhibit norms

• Report standard errors (or confidence intervals), not significance asterisks/boldface. Put SEs in parentheses below estimates; the reader judges significance from the SE/CI. This is the house convention to follow.
• Self-contained. Title and notes let an exhibit be read without the text: sample, units, estimator, clustering level, what is controlled, and what the number means in policy terms.
• Self-contained, not anonymized. Review is single-blind, so exhibits need not hide authorship; keep notes clean and neutral (avoid stray local file paths) for readability, not for blinding.
• Figures are the workhorse for policy communication: event-study plots with CIs, RDD plots with binned means and the fitted discontinuity, dose-response or cost-benefit curves with uncertainty bands. Vector output; ≥300 dpi raster only if unavoidable; readable greyscale.

The headline exhibit (AEJ: Policy-specific)

Every AEJ: Policy paper should have one exhibit a policymaker could screenshot: the policy effect in interpretable units with its welfare/cost-benefit reading where possible. Examples (illustrative formats):

• An event-study figure of the outcome around the reform, with the long-run effect annotated in policy units.
• A cost-benefit / MVPF figure: net cost per unit of outcome across policy variants, with bands.
• An incidence figure: who gains and who pays, by income/region group.

Table craft

• Three-line tables (esttab/booktabs), no vertical rules; align decimals; consistent digits.
• Lead column = the policy-relevant specification, not a kitchen-sink spec.
• Put the policy-relevant magnitude (elasticity, cost-per-X, MVPF) in the paper's units, not only a raw coefficient; add a row translating the coefficient into the policy number where natural.
• Sample size, mean of the dependent variable, and clustering level on every table.

Execution bridge (StatsPAI / Stata MCP)

Generate exhibits from the fitted result, not by retyping numbers (the usual source of body-vs-appendix drift). Full map: execution-with-mcp.

• Tables: etable (multi-model columns) or did_summary_to_latex straight from the

result_id — one variable definition, one set of numbers, body and appendix in sync.

• 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 (from the result's diagnostics) and

states the magnitude in interpretable units.

See a full fitted-result → exhibit chain in the JF execution walkthrough.

Checklist

• ] No significance asterisks or boldface-for-significance anywhere; SEs/CIs shown
• ] One self-contained headline exhibit carrying the policy magnitude (and welfare reading if available)
• ] Every table notes: sample, units, estimator, controls, clustering level, N, dep-var mean
• ] Figures show effects with uncertainty bands (event study / RDD / dose-response)
• ] Policy-relevant magnitude in interpretable units, not only raw coefficients
• ] Vector figures, readable in greyscale, no chartjunk

Anti-patterns

• Significance stars and bold "" — disallowed by AEA house style here
• A coefficient table with no translation into the policy magnitude
• A 12-column kitchen-sink table where the policy spec is buried
• Figures with point estimates but no confidence bands
• Notes too thin to read the exhibit without the text
• An exhibit whose notes are too thin to read without the body text

Exhibit-by-design-type quick guide

| Design | Workhorse figure | What the notes must state | |---|---|---| | DID / event study | Event-study coefficients with CI bands; flat pre-period leads visible | estimator (CS/SA), comparison group, clustering level | | RDD | Binned means + fitted discontinuity + bandwidth | running variable, bandwidth, density-test result | | Bunching | Empirical vs. counterfactual density at the kink | counterfactual construction, excluded region | | RCT | Treatment-control means / dose-response with CIs | randomization unit, take-up, ITT vs. ToT | | Welfare | MVPF / cost-per-outcome across variants, with bands | which estimates feed the ledger, assumptions |

Worked vignette (illustrative)

A tax-credit paper's main table reports a coefficient of 0.08 (s.e. 0.02) on the credit. Reworked for AEJ: Policy: the lead exhibit becomes a figure of employment around the credit's introduction with a CI band, the long-run effect annotated as "+3.1 pp employment (90% CI 1.9, 4.3])," and a companion row translating it into cost per additional job with its band — the number a policymaker takes away. No asterisks; SEs in parentheses throughout.

Output format

【Headline exhibit】figure/table + the policy magnitude it carries
【Significance reporting】SEs/CIs, no asterisks? [Y/N]
【Self-contained notes】sample/units/estimator/clustering/N/dep-mean present? [Y/N]
【Policy-units translation】coefficient → cost-per-X / MVPF / incidence shown? [Y/N]
【Next step】aejpol-writing-style