jpart-tables-figures

Tables & Figures (jpart-tables-figures)

Exhibits are where an expert public-management reviewer checks whether the result — and the mechanism — are real. At JPART they also cost words: the manuscript limit (~12,000 words) includes tables and references, so every exhibit must earn its space. Numbers must match the public data-and-code package.

When to trigger

• Designing the main results table/figure or a key descriptive exhibit
• Deciding what belongs in the article vs. the online appendix/supplement
• A reviewer found an exhibit unclear, mislabeled, or not self-contained
• Trimming to the word cap by moving exhibits to the appendix

Principles

1. Self-contained. A reader should understand each exhibit from its title, labels, and note alone — state units, sample, N, the estimand, and the inference (CI, clustering level). No hunting in prose.
2. Show effects, not stars. Coefficient/forest plots, marginal-effects and predicted-probability plots, AMCE plots for conjoints, event-study/RD plots — these communicate magnitude and uncertainty better than a wall of coefficients. Always show intervals.
3. Mechanism exhibits. A first-stage/mediation panel, a heterogeneity plot keyed to a scope condition, or a balance/manipulation-check table — make the theory visible, not just the top line.
4. Accessible. Colorblind-safe palettes; legible in grayscale; no chartjunk or 3D.
5. Reproducible. Each exhibit is generated by the master script; values match the deposited package exactly (the data and code are public — see jpart-transparency-and-data).

PA-specific exhibits

• Multilevel results: report ICCs and variance components, not just fixed-effect coefficients.
• Survey-experiment treatment-effect plots; conjoint AMCE/marginal-means plots.
• Representative-bureaucracy: passive→active representation shown as an interaction/marginal-effects plot.
• Descriptive table of the public-employee/citizen sample with attention/attrition flags.

Theory-to-exhibit trace

Before keeping an exhibit in the article, write a one-line trace:

public-management construct -> observable measure -> exhibit -> theoretical implication

Examples:

• Representative bureaucracy: passive representation -> agency/client subgroup match -> marginal-effects plot -> when passive representation translates into active representation.
• Red tape: administrative burden -> process-step count or perceived burden scale -> mechanism figure with subgroup heterogeneity -> why rules constrain behavior rather than merely correlate with outcomes.
• Performance information: information cue -> treatment arm or information-use measure -> predicted response plot -> whether citizens/managers use performance data as accountability or learning signals.

If the trace ends at "interesting pattern" rather than a public-management mechanism, move the exhibit to the supplement or cut it. JPART reviewers should be able to infer the theory move from the exhibit title, axis labels, and note before reading the surrounding prose.

Main-exhibit budget

Use a tight main-text budget:

Full model grids, alternative coding schemes, balance details, and exploratory heterogeneity belong in the online supplement unless they are necessary for the theory contribution.

Fit the word cap (limit includes tables + references)

• Move balance tables, full specifications, and extended robustness grids to the online supplement.
• Keep the few exhibits that carry the argument and the mechanism in the article.
• Prefer one decisive marginal-effects figure to three redundant coefficient tables.

Execution bridge (StatsPAI / Stata MCP)

Generate exhibits from the fitted result, not by retyping numbers. Full map: execution-with-mcp. JPART is public management — observational and experimental designs on public organizations; identification + clustered/multilevel inference.

• 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.

Anti-patterns

• Tables that require the prose to be intelligible (not self-contained)
• Significance stars with no effect size or interval
• Cramming every robustness check into the main text (blows the budget that counts tables/refs)
• Color-only encoding that fails in grayscale or for colorblind readers
• Exhibit numbers/values that don't match the deposited public code output

Output format

【Main exhibit】what it shows + why a figure/table
【Mechanism visible?】scope condition / mediation / first stage shown? [Y/N]
【Self-contained?】title + labels + note + N/units + inference present? [Y/N]
【Theory-to-exhibit trace】construct -> measure -> exhibit -> implication
【Main-text slot】descriptive / main effect / mechanism-scope / necessary robustness
【Accessible?】grayscale-legible + colorblind-safe? [Y/N]
【Article vs supplement】split decided, word-budget impact noted
【Reproducible?】generated by master script, matches package? [Y/N]
【Next】jpart-writing-style

Supplementary resources

• ../../resources/external_tools.md — plotting/marginal-effects/conjoint packages
• ../../resources/official-source-map.md — word-limit rule (includes tables/references)