imfer-tables-figures
GitHub为IMF经济评论稿件构建透明、易读的经济图表。要求明确国家覆盖范围，使用标准误或置信区间替代显著性星号，并将系数转化为政策相关单位（如GDP占比），确保同时满足学术审查与政策读者的需求。
Trigger Scenarios
Install
npx skills add brycewang-stanford/Awesome-Journal-Skills --skill imfer-tables-figures -g -y
SKILL.md
Frontmatter
{
"name": "imfer-tables-figures",
"description": "Use when an IMF Economic Review (IMFER) manuscript's exhibits must communicate an international-macro result to a dual academic\/policy audience — transparent country coverage, readable cross-country panels, no significance asterisks. Builds the exhibits; it does not run the robustness behind them (imfer-robustness) or write the prose (imfer-writing-style)."
}
Tables and Figures (imfer-tables-figures)
When to trigger
- A cross-country regression table is dense and the reader cannot find the headline coefficient
- Country/sample coverage is buried — a referee cannot tell which economies and years are in
- Impulse responses, event-study leads, or spillover paths need a figure that a policymaker can read at a glance
- Significance asterisks are still on the tables, or SEs/CIs are missing
- A map or panel-by-country exhibit would communicate the international margin better than a table
- A summary-statistics table omits the cross-country dispersion a referee needs to judge the sample
What IMFER exhibits must do
IMFER exhibits serve two readers at once: a referee checking the econometrics and a policy reader who wants the magnitude and its country scope. So the country coverage must be transparent (which economies, which years, balanced or not), the headline number must be findable, and the uncertainty must be visible as standard errors and confidence intervals — not asterisks. The best IMFER figure is one that could appear in a WEO/GFSR chart pack and survive referee scrutiny: clear axis units, sample noted, mechanism legible.
| Exhibit job | The right form | Watch for |
|---|---|---|
| Show the headline cross-country effect | a compact main table, preferred spec first | not a 12-column kitchen sink |
| Show country/sample coverage | a coverage table or a map; note balanced/unbalanced | hidden sample; undated coverage |
| Show dynamics (spillover, IRF) | a line figure with CI bands | bands omitted; ambiguous units |
| Show a DID/event study | event-time plot with leads and lags | no pre-period leads shown |
| Show heterogeneity (AE vs EM) | split panels or a coefficient plot | a wall of subgroup rows |
| Show a policy magnitude | units a policymaker uses (% of GDP, bps) | raw coefficients with no economic scale |
Exhibit craft
- The summary-statistics table earns its keep. For a cross-country paper this is not boilerplate: report the dispersion across countries (not just the pooled mean), the coverage by income group, and the key conventions (currency, gross/net) so a referee can judge whether the sample supports the claim before reading a single regression.
- One headline exhibit. Decide the single table or figure that is the paper; make it self-contained — a reader who sees only it should grasp the result, the sample, and the uncertainty.
- Make country coverage explicit. State the economies and years in the note; if the panel is unbalanced, say so; a coverage table or map pre-empts the "which countries?" referee question.
- No asterisks. Report standard errors (or CIs); for the headline, prefer a confidence interval so the policy reader sees the range, not a star.
- Put magnitudes in policy units. Translate coefficients into % of GDP, basis points, or a number a mission economist would quote.
- Show dynamics with bands. Spillover paths and IRFs need confidence bands and labeled horizons; event studies need visible pre-trend leads.
- Self-contained notes. Estimator, sample, clustering, and units in the note — the exhibit should not require the text to be intelligible.
Country panels: the recurring exhibit problems
Cross-country exhibits fail in patterned ways. Heterogeneous scale: a few large economies dominate any pooled scatter — log, normalize by GDP, or facet by income group. Unbalanced coverage masked: a table reporting "N = 1,200" hides that half the cells are one decade; show coverage by country-year. Currency and vintage drift: a figure mixing USD and local-currency series, or two data vintages, misleads; state the convention in the note. Map theatrics: a choropleth that colors countries but encodes no number the argument uses is decoration — prefer it only when the geographic pattern is the finding (contagion, regional clustering). The fix is always the same: make the cross-country comparison the reader needs explicit and quantitative.
Execution bridge (StatsPAI / Stata MCP)
Generate exhibits from the fitted result, not by retyping numbers. Full map:
execution-with-mcp. IMFER is international macro/finance; cross-country panels with confounded policy — emphasize identification and clustering.
- Tables:
etable(multi-model) ordid_summary_to_latexstraight from theresult_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
- One self-contained headline exhibit; preferred specification first
- Summary-statistics table shows cross-country dispersion and coverage by income group
- Country and year coverage stated; balanced/unbalanced noted (coverage table or map if helpful)
- No significance asterisks; SEs or CIs reported, CI preferred for the headline
- Magnitudes in policy-readable units (% of GDP, bps), not bare coefficients
- Dynamics (IRF / spillover / event study) shown with CI bands and labeled horizons
- Event studies show pre-period leads; heterogeneity shown as panels/coefficient plots not row walls
- Currency / gross-net / deflator convention stated in every relevant note
- Large economies do not dominate any pooled scatter (log scale / GDP normalization used)
- Notes carry estimator, sample, clustering, units — exhibit readable on its own
Anti-patterns
- A 12-column regression table where the headline coefficient is impossible to locate
- Asterisks/boldface for significance instead of SEs/CIs
- Country coverage hidden — the reader cannot tell which economies or years are in
- Coefficients reported with no economic scale a policymaker could use
- IRFs / spillover paths with no confidence bands
- An event-study figure with no pre-period leads (parallel-trends evidence missing)
- A choropleth map that looks impressive but encodes no quantitative comparison the text uses
- A pooled scatter dominated by a few large economies, with no log scale or GDP normalization
- A summary table reporting only pooled means, hiding the cross-country dispersion
- Exhibit notes that omit the estimator, clustering, or units, forcing the reader back to the text
Worked vignette (illustrative)
A spillover paper's first draft has a 10-column table of US-shock coefficients across outcomes, asterisked, with the sample described only as "47 countries." The IMFER revision keeps the preferred column as Table 1 with the elasticity translated to "−0.8% of GDP per 100bp" and a 95% CI in the cell; a coverage table lists the 47 economies, the 1996–2023 window, and the unbalanced cells; and the dynamics move to a figure of the impulse response with bands over 12 quarters, AE and EM panels side by side. A referee can now verify the econometrics and a mission economist can quote the magnitude — both from the exhibits alone.
Referee pushback mapped to the exhibit fix
- "Which countries and years are in this panel?" → Add a coverage table or map and note balanced/unbalanced in every exhibit footnote.
- "What does this coefficient mean economically?" → Report the magnitude in % of GDP or basis points alongside the raw coefficient.
- "Drop the asterisks." → Replace stars with standard errors and, for the headline, a confidence interval.
Output format
【Journal】IMF Economic Review
【Skill】imfer-tables-figures
【Headline exhibit】table/figure that IS the paper: ___
【Country coverage】economies + years + balanced?: ___
【Uncertainty】SEs / CIs (no asterisks)? [Y/N]
【Policy units】magnitude in % of GDP / bps: ___
【Dynamics】IRF/spillover/event-study with bands? [Y/N]
【Next skill】imfer-writing-style
Version History
- 1839142 Current 2026-07-05 13:21
