Tables & Figures (eer-tables-figures)
When to trigger
- Tables are dense, over-decorated with significance stars, or missing SEs
- Figures are cluttered, lack uncertainty bands, or are not self-explanatory
- A referee said an exhibit is "unreadable" or "buries the result"
- Preparing final editable source files for Elsevier production
The EER exhibits bar
EER follows Elsevier economics house style: a paper is read through its main tables and a few clean figures, each self-contained (legible without the text). The discipline norm is to report standard errors (and, increasingly, confidence intervals) and to let the magnitude carry the result rather than a wall of asterisks. Figures and tables can be embedded for review and are supplied as editable source files at final stage. Aim for fewer, sharper exhibits: the headline result should be findable in one table and one figure.
Table design rules
-
One headline table carries the main result; supporting tables are secondary.
-
Report SEs (in parentheses) under every coefficient; report N, R²/pseudo-R², and the FE/cluster structure in the notes.
-
Magnitudes legible: scale coefficients to interpretable units; add a sentence-ready effect size.
-
Significance: report SEs/CIs; if stars are used, do not let them replace the SE or the magnitude — EER norms emphasize the standard error, not the star.
-
Notes are complete: sample, period, clustering level, estimator, and what each column varies.
-
No redundant digits: 2–3 significant figures; align decimals.
Figure design rules
-
Event-study plots with leads/lags and confidence bands are the EER-standard causal exhibit.
-
Model-fit overlays (data vs. model moments) for quantitative/structural work.
-
Counterfactual / policy curves with uncertainty bands.
-
Self-contained captions: a reader should grasp the figure from the caption alone.
-
Vector output (
.eps/.pdf) or high-resolution .tif/.jpg; legible fonts; colorblind-safe palettes; avoid chartjunk and 3-D effects.
Execution bridge (StatsPAI / Stata MCP)
Generate exhibits from the fitted result, not by retyping numbers. Full map:
execution-with-mcp. EER is a general economics field journal; the DiD/IV/RDD chain serves its applied lane.
-
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 table whose only signal is the number of asterisks (no SEs, no magnitude)
- Twelve nearly-identical regression columns where two would do
- Figures with no uncertainty bands or no axis labels
- Captions that require the body text to interpret the exhibit
- Raster screenshots of tables instead of editable source
- Reporting six decimal places of false precision
- Color encodings that fail in grayscale or for colorblind readers
- An appendix figure carrying the headline result the main text never shows
Worked vignette (illustrative)
A labor paper's first draft has a 10-column table dense with stars and no SEs. The EER revision: one headline table (4 columns: baseline, +controls, modern DiD, IV) with SEs in parentheses, clustering noted, and effect sizes spelled out ("a 1 SD increase raises wages 2.3%, s.e. 0.6"); plus one event-study figure with flat leads and a shaded 95% band. The remaining specifications move to the Supplementary material. The result is now findable in ten seconds.
Output format
【Headline table】columns + what each varies; SEs present? [Y/N]
【Reporting】SEs/CIs under estimates; stars not load-bearing? [Y/N]
【Key figure】event-study / fit-overlay / counterfactual + uncertainty bands? [Y/N]
【Self-contained】captions+notes legible alone? [Y/N]
【Final files】editable/vector source ready? [Y/N]
【Next step】eer-writing-style (weave exhibits into the narrative)
