Jeg Tables Figures

Tables & Figures (jeg-tables-figures)

When to trigger

• Main results exist but exhibits do not communicate growth mechanisms
• Calibration or simulation output needs a readable structure
• Empirical tables lack sample, fixed-effect, clustering, or unit notes

Exhibit architecture

• Descriptive growth facts: long-run trends, cross-country dispersion, transition paths, or cohort patterns.
• Main estimates or model results: one table/figure per claim, not per estimator experiment.
• Mechanism exhibits: human capital, fertility, technology, institutions, financial development, migration, or political economy channels.
• Robustness/sensitivity: alternative samples, periods, specifications, calibration targets, and parameter values.

JEG-specific polish

• Use economically meaningful units: annual percentage points of growth, log GDP per worker, schooling years, fertility rates, TFP, steady-state ratios.
• Label whether an exhibit is empirical, calibrated, simulated, or theoretical.
• For transition paths, show initial condition, time scale, and steady state.
• For cross-country panels, disclose country coverage and period.

Exhibit sequence

Most JEG papers need one of these sequences:

• Empirical growth: descriptive growth fact -> identification diagnostic -> main estimate -> mechanism or heterogeneity -> robustness.
• Theory/calibration: model mechanism diagram or proposition summary -> calibration table -> transition path -> sensitivity -> welfare or growth decomposition.
• Mixed paper: empirical fact motivating the mechanism -> model result -> calibration/estimation -> data validation -> counterfactual.

The sequence should make the growth mechanism visible before the robustness appendix expands.

Growth exhibit contract

Each exhibit should state whether it shows a fact, identifies a mechanism, validates a model, or tests sensitivity. Do not let all tables look like robustness tables. A JEG reader should be able to follow the growth story from exhibits alone:

Fact -> mechanism -> model/estimate -> magnitude -> sensitivity -> implication

For calibration exhibits, report the target moment, model moment, parameter value, and source. For empirical exhibits, report country/region coverage, period, units, fixed effects, clustering, and whether the estimate is meant to be causal or descriptive.

Map and spatial-exhibit standards

Persistence and comparative-development submissions live and die by their spatial exhibits:

• Every map states the projection, the unit (grid cell, district, ethnic homeland), the period of the plotted variable, and the source layer.
• Show the treatment variation on a map before the regression tables appear; growth referees want to see where identification comes from geographically.
• Table notes for geocoded outcomes state the spatial-SE method and cutoffs (e.g., "Conley SEs, 250 km, in brackets") next to the clustered SEs — in the main table, not buried in an appendix.
• Avoid choropleth color scales that flatten the variation actually identifying the model; bin by the estimation sample's own distribution.

Worked vignette — rebuilding a persistence main table

Illustrative redesign of a weak Table 2 (outcome: log GDP per capita 2015 by district; regressor: early printing-press adoption):

• Column 1: OLS with country FE — 0.19 (clustered SE 0.04).
• Column 2: adds geography controls (ruggedness, latitude, coast distance) — 0.16 (0.04).
• Column 3: same specification reporting Conley 250 km SEs — 0.16 (0.06); an inference column, not a new estimate.
• Column 4: IV using distance to an early adoption hub — 0.24 (0.09), with first-stage F = 21 in the notes.
• One row added: the outcome mean, so magnitudes read directly as percent effects.

The old version's six estimator-variation columns move to the appendix, and the schooling-channel table is promoted into the main text — mechanism before robustness.

Exhibit-count anchor (hedged)

Accepted articles at this journal commonly carry 6-10 main exhibits backed by a deep online appendix; treat that as a prior from recent issues rather than a rule, and check figure-format specifics against the journal's current author guidelines.

Execution bridge (StatsPAI / Stata MCP)

Generate exhibits from the fitted result, not by retyping numbers. Full map: execution-with-mcp. JEG (growth) uses cross-country and long-run panels with deep endogeneity; foreground identification and robustness to alternatives.

• 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.

Output format

[Exhibit] table / figure / appendix
[Claim supported] ...
[Units and sample] ...
[Model/data status] empirical / calibrated / simulated
[Missing note fields] ...
[Next step] jeg-writing-style

Source: brycewang-stanford/Awesome-Journal-Skills → Journal-of-Economic-Growth-Skills/skills/jeg-tables-figures/SKILL.md