Tables & Figures (jfe-tables-figures)
When to trigger
- A main table has so many columns no reader can parse it
- Table notes do not let a reader understand the table without the text
- Standard errors / t-stats are reported inconsistently across tables
- Figures are decorative rather than carrying part of the argument
- You are deciding what stays in the main text vs. the Internet Appendix
JFE exhibit standards
JFE tables are dense but disciplined. Each main exhibit should make one clear point and be self-contained: a reader should grasp it from the title, column headers, and notes alone. Because JFE papers are long and robustness-heavy, ruthless triage between main-text and Internet-Appendix exhibits is part of the craft. The overflow goes to the online appendix that JFE asks you to append to the end of the main manuscript file (see jfe-internet-appendix). Manuscripts are submitted double-spaced, 12-pt+, with >= 1-inch margins, so dense tables must still read at that scale. Verify current formatting on jfinec.com.
Table craft
Structure
- One exhibit, one message. If a table answers two questions, split it.
- Order columns from baseline to richest specification, left to right.
- Group related rows; keep the variable of interest visually prominent (top rows).
Reporting conventions
- Report coefficients with standard errors (or t-statistics) in parentheses — pick one convention and use it everywhere.
- State the significance-star convention once and apply consistently.
- Always report N, R-squared (or relevant fit), fixed effects included, and the cluster level — in the table, not only the text.
- Report economic magnitudes alongside coefficients where readers need them.
Notes
- The note states: sample and period, variable definitions (or a pointer to the definition table), the standard-error treatment, and what each star means.
- Notes should make the table replicable in spirit without flipping back to the text.
Figure craft
- Event-study/coefficient plots should show confidence bands and the zero line.
- Label axes with units; no unexplained acronyms.
- Use figures where a pattern (pre-trends, a discontinuity, a return path) is clearer than a table — not as decoration.
- Keep them legible in grayscale; do not rely on color alone to distinguish series.
Execution bridge (StatsPAI / Stata MCP)
Generate exhibits from the fitted result, not by retyping numbers. Full map:
execution-with-mcp. JFE is finance top-3 (with JF, RFS) — corporate-causal chain for corporate papers, factor-zoo haircut for asset pricing; attribute canon to the correct top-3 journal.
-
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
Anti-patterns
- A 12-column main table that no referee will read
- Parentheses that mean SEs in one table and t-stats in another
- Omitting the cluster level or fixed-effects rows from the table
- Three-decimal coefficients with no sense of economic size
- Figures that repeat a table without adding clarity
- Color-only series distinctions that vanish in print
Output format
【Main exhibits】list, each with its one message
【Reporting convention】SEs|t-stats ; star scheme
【Per-table footer fields】N / fit / FE / cluster present? yes/no
【Figures】[purpose of each]
【Moved to appendix】[exhibits relocated]
【Next】jfe-internet-appendix