Ier Tables Figures

Tables and Figures (ier-tables-figures)

When to trigger

• A reader cannot tell from a table alone what claim it supports
• A structural paper's parameter table lists estimates but not what identifies them or their precision
• A theory paper has no figure making the mechanism or comparative static visible
• Regression tables are wall-to-wall coefficients with significance stars and no economic magnitudes
• Exhibits must fit the ≤50-page IER ceiling and are crowding out the argument

What IER exhibits must do

IER exhibits serve a rigor-leaning, broad audience, so each one must answer a single question and make the economics — not just the statistics — legible. For a theory/structural-tilted journal, the most valuable exhibits are often the ones that make a mechanism or a comparative static visible, and the parameter table that shows the model is disciplined by data. Sort your exhibits by their job:

House-style discipline

• Report standard errors and confidence/coverage sets, not significance asterisks. A rigor journal wants the precision shown, not thresholded — strip ***/**/* and report SEs in parentheses and economic magnitudes in the text.
• Self-contained notes. Every table/figure note states the sample, the estimator, the unit of observation, the inference method, and what each column varies — a referee should not need the text to read the exhibit.
• One question per exhibit. If a table answers two questions, split it. Dense omnibus tables are where IER referees lose the thread.
• Magnitudes in the text. The table holds the numbers; the prose says what they mean ("a one-SD increase raises X by 8%, roughly a third of the cross-country gap").
• Mind the page budget. With a ≤50-page ceiling (检索于 2026-06；以官网为准), move secondary exhibits to the online/proof appendix (ier-replication-package) and keep the main text exhibits that carry the contribution.

The structural paper's signature exhibit

For the structural/quantitative papers that are IER's bread and butter, the single most important exhibit is usually the parameter table read together with the sensitivity/identification panel. A referee scanning a new structural paper goes straight to it to answer one question: are these parameters pinned by the data, or assumed? Make that table do the work — estimate, standard error, the identifying moment, and (ideally) the sensitivity entry showing which moment moves it. A parameter table that is just a list of numbers tells the referee the model was calibrated, not estimated, and that perception is hard to reverse.

Worked example (illustrative)

A draft presents a 12-column regression table with three sets of controls and stars on everything. It answers no single question and the reader cannot find the result. The fix splits it: one main-text table reporting the headline coefficient as an economic magnitude with its confidence interval, columns ordered by threat retired (baseline → drop influential subsample → alternative inference), and the control-set permutations moved to an online appendix. The note states sample, estimator, unit, and clustering. Now the table answers exactly one question — "is the effect there and how big" — and the reader sees it in five seconds.

Why no asterisks at a rigor journal

The case against significance stars is not stylistic at IER — it is methodological. A rigor-leaning readership wants the precision of an estimate (its standard error or confidence set) shown directly, so the reader judges economic and statistical significance themselves, rather than having a coarse three-level threshold imposed by the author. Stars also encourage the very specification search the robustness section is meant to rule out. Report the SE in parentheses, state the economic magnitude in the text, and let the confidence interval do the work the stars used to do. This aligns with the broader econometric-society house norm and is the safer default even where the exact style guideline is 待核实.