smj-tables-figures

Tables & Figures (smj-tables-figures)

When to trigger

• You have many tables and are unsure which earn their place in the main text
• Tables are not self-contained (a reader cannot read them without the prose)
• You report coefficients but have no figure showing the effect or mechanism
• Descriptive statistics, correlations, and regressions are disorganized or inconsistent

Standard SMJ exhibit set

A typical empirical SMJ paper carries a compact, predictable set of exhibits:

1. Descriptive statistics + correlation matrix (one table). Flag high correlations and multicollinearity concerns here.
2. Main results table. Hierarchical/nested models: controls-only column, then add focal X, then moderators/mechanism. Each column builds.
3. Identification / endogeneity table. First stage + IV, DID estimates, matching balance, or selection model — the evidence that defeats the threat.
4. Robustness table(s). Alternative DVs, samples, estimators.
5. Figure(s): marginal-effects / interaction plot, event-study plot for DID, or a mechanism/path diagram.

Move secondary tables to an online appendix/supplement rather than crowding the main text. SMJ guides authors toward a tight page budget — full articles around 40 pages (including tables, figures, and references) and short research articles around 20 pages (verify the current limit) — so exhibits compete for space: keep the main paper lean and use APA-style table formatting with clear notes.

Table craft

• Self-contained: title states what the table shows; notes define every variable, the estimator, the sample, the unit, what stars mean, and what is in parentheses (SE vs. t). A reader should not need the prose.
• Report standard errors (in parentheses) and state the clustering level in the note; do not report bare stars.
• Interaction interpretation: never interpret an interaction from the coefficient alone — accompany it with a marginal-effects plot or computed effects at meaningful values of the moderator.
• Consistency: identical variable names, decimal places, and ordering across tables. Sample N should reconcile across tables (explain any drop).
• Three-line / clean style: no vertical rules, minimal horizontal rules; align decimals.

Figure craft

• A marginal-effects plot communicates a moderation hypothesis far better than a triple-interaction coefficient.
• For DID, an event-study figure (pre-trends flat, effect emerging post-treatment) is often the single most persuasive exhibit.
• A mechanism/path figure can summarize the theorized model; keep it conceptual, not decorative.
• Greyscale-legible (SMJ is widely printed/read in B&W); label axes and units; no chartjunk.

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. SMJ is strategy — firm-level panels where strategic choices are endogenous; foreground IV / DiD identification and the endogeneity-of-strategy objection.

• Tables: etable (multi-model columns) 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 effect size in interpretable units.

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

Checklist

• Each main-text table is self-contained (title + complete notes)
• Standard errors shown; clustering level stated in the note
• Main results presented as nested/hierarchical columns that build
• An identification/endogeneity exhibit is present and prominent
• Interactions are visualized (marginal-effects plot), not read off coefficients
• Variable names, decimals, ordering consistent across all exhibits; N reconciles
• Secondary tables moved to the appendix/supplement (main paper within the ~40-page / ~20-page budget)
• Coefficients interpretable as economic magnitudes, not bare stars
• Figures are greyscale-legible with labeled axes

Anti-patterns

• A table that cannot be read without the surrounding text
• Reporting only significance stars with no standard errors or clustering note
• Interpreting an interaction term's sign without a marginal-effects plot
• Eight near-identical robustness tables in the main text
• Inconsistent N or variable naming across tables
• A color-only figure that becomes unreadable in print
• DID with no event-study/pre-trends figure

Output format

【Main-text exhibits】[descriptives, main results, identification, robustness, figure(s)]
【Self-contained?】yes / fix: [...]
【SE + clustering noted】yes / fix
【Interaction visualized】yes / needs marginal-effects plot
【Event-study figure (if DID)】present / missing
【Appendix moves】[tables relocated]
【Within page budget (~40 / ~20)】yes / trim
【Next step】smj-writing-style

Templates & resources

• ../../resources/external_tools.md — table/figure tooling (estout/esttab, modelsummary, coefplot, marginsplot, ggplot2)
• ../../resources/official-source-map.md — SMJ formatting, APA reference style, and page-limit guidance