brycewang-stanford/aer-paper-body

AER Paper Body

Overview

The introduction decides whether the editor sends the paper out; the body sections decide what the referees write. Referees live in Data, Empirical Strategy, and Results, checking that the estimand is defined, the assumption stated, the magnitudes interpreted, and the prose matched to the tables. Draft the body before the introduction — the introduction summarizes a paper that already exists, and writing it first produces promises the body fails to keep.

When to Use

• The empirics are stable (after aer-identification and aer-robustness) and the manuscript needs full section drafts
• A results section reads like a table walk-through ("Column 1 shows...") and needs narration surgery
• A referee called the paper "hard to follow," "under-interpreted," or "reads like a report"
• Coefficients are reported but never converted into economic magnitudes
• The conclusion restates the abstract and needs to do real work

Canonical Section Architecture

A full-length empirical AER paper, after the unlabeled introduction:

I.   Background  (or: Institutional Setting; Policy Context)
II.  Data        (sources, sample construction, measurement, summary stats)
III. Empirical Strategy   (estimand, equation, identifying assumption, inference)
IV.  Results     (main estimates, dynamics, robustness pointers)
V.   Mechanisms  (or: Heterogeneity and Mechanisms; Interpretation)
VI.  Conclusion

Variants: a conceptual framework goes between Background and Data (or replaces Background for theory-led papers); AER: Insights compresses to Data and Design → Results → Discussion; structural papers add Model and Estimation, where the rules below bind with more force, not less. One rule binds everywhere: every term, dataset, and design feature is defined before first use — referees read linearly on the first pass.

Section I — Background / Institutional Setting

Give exactly the institutional detail needed to (a) locate the identifying variation and (b) believe the identifying assumption — nothing else.

• Lead with the institution or policy, not the literature. History belongs here only if it explains the variation.
• State who is treated, when, by what rule, and who decided. If assignment is discretionary, say so and explain how the design handles it.
• Length 1.5-3 typeset pages; surplus is almost always literature review in disguise — cut it or move it to the intro's antecedents paragraph.
• Cite a primary source (statute, agency document, administrative report) for every institutional claim, not a secondhand economics paper.

Optional Section — Conceptual Framework

Include one only if it generates a testable prediction the empirics then test, defines the estimand's welfare interpretation (a reduced-form coefficient as a sufficient statistic), or disciplines magnitudes (what effect size theory permits).

• Keep it stylized — two-period, two-type, linear where possible. Push derivations to the appendix; state propositions and intuition in the text.
• End with an explicit bridge naming the predictions taken to the data; each must reappear, by name, in Results or Mechanisms.
• Do not bolt on a model that re-derives "demand slopes down." Referees read decorative theory as padding and say so.

Section II — Data

Referees check this section against the replication package line by line.

• Sources. For each dataset: name, producer, access mode, years, unit of observation, and a citation. If access is restricted, say how it was obtained (this feeds aer-replication).
• Sample construction — the most under-written block in rejected papers. Report the funnel explicitly with counts (raw file → each drop with its count and rationale → analysis sample). Counts must match the replication package exactly (aer-consistency audits this); flag consequential restrictions the robustness section relaxes.
• Variables and measurement. Define outcome and treatment precisely, with units, in prose (the full variable table lives in the appendix). Discuss measurement error where a referee would: self-reports, imputation, top-coding, deflators (name the index and base year). State the level of aggregation and why it matches the design.
• Summary statistics. Narrate Table 1, do not re-read it aloud. Two to four facts, each tied to a design decision: is the sample representative, are the groups comparable pre-treatment, and what features (skewness, mass points, attrition) shape specification choices.

Section III — Empirical Strategy

The section referees read most carefully. Four mandatory components, in order: estimand → equation → identifying assumption → inference.

Estimand first, one sentence before any equation: "Our object of interest is the average effect of [treatment] on [outcome] among [population], [horizon]." If the design recovers a local effect (LATE, effect at the cutoff, ATT for switchers), say so here, not in the conclusion's limitations paragraph.

Y_{ict} = \beta\, D_{ct} + \alpha_c + \gamma_t + X_{ict}'\delta + \varepsilon_{ict}

• Define every symbol in the sentence after the equation, every subscript included (what each index ranges over; what $D_{ct}$ codes). Keep subscript order consistent across equations and tables. Number displayed equations; refer to "equation (1)."
• State which coefficient is the object of interest and what variation identifies it once controls and fixed effects absorb the rest.
• For modern DiD, write the ATT(g,t) estimand and aggregation explicitly rather than pretending the design is equation-(1) TWFE — the estimator choice comes from aer-identification.
• Identifying assumption: state it formally and in words, then preview the evidence. The unit of writing is "assumption — threat — evidence"; name the two or three most plausible violations and point to where each is met.
• Inference: one paragraph — clustering level and why it matches the design's variation, number of clusters, and any small-cluster correction (wild cluster bootstrap) or design-specific inference (AR confidence sets, randomization inference). Never leave inference to the table notes alone.

Section IV — Results

The narration discipline

One paragraph per claim, not per table. Each results paragraph:

1. Finding first. The opening sentence states the economic finding with its magnitude, not the table's existence: "The reform raises earnings by 4.2 percent (Table 3, column 4)," never "Table 3 presents the results."
2. Evidence. Where the number comes from and how it moves across specifications (added controls, finer fixed effects, alternative samples).
3. Interpretation. Convert the coefficient into economic units and benchmark it.

Column-by-column narration without a finding-first sentence is the most reliable marker of a weak results section.

Magnitude interpretation

Every headline coefficient gets three conversions: (1) native units — log-outcome coefficients are log points; use the exact 100·(e^β − 1) whenever |β| > 0.10, and never confuse percent with percentage points; (2) relative to the sample — against the dependent variable's mean or SD; (3) relative to the literature or a policy lever. Report the baseline rate next to every binary marginal effect. If the implied magnitude is implausible, say so and investigate before a referee does. Worked conversions: examples/results-section-example.md; the percent/percentage-point table lives in aer-consistency.

Precision and back-of-envelope

• Quote point estimates and standard errors exactly as the table shows them. "Significant at the 5 percent level," never "almost significant." For nulls, report the CI and what it rules out, distinguishing a precise zero from an uninformative one. Line-level rules: docs/style-guide.md.
• Close a high-impact results section with one disciplined aggregate calculation: every input and its source stated, multiplied through in one visible chain, rounded honestly — not a cascade of unsourced products. If it needs more than a paragraph, it becomes its own short section. Worked example: examples/results-section-example.md.

Robustness pointers

The results section cites robustness, it does not contain it: one paragraph summarizing the aer-robustness battery ("stable across clustering, sample, and specification variants; Appendix B reports the full set") plus the single most important check shown inline.

Section V — Mechanisms

Organize by candidate explanation, not by table:

1. Name the two or three channels consistent with the main result, including the ones you will rule out.
2. For the favored channel, present the theory-predicted heterogeneity and auxiliary outcomes ("if the channel is skill-biased adoption, effects should concentrate in tradeable services — they do, Table 5").
3. For each alternative, state the sharpest version of the referee's objection, then the evidence against it. Steelman first, then answer.
4. Close with calibrated language ("most consistent with [channel]," not "we prove the mechanism"). Mechanism evidence is consistency evidence; only the main effect carries design-based identification (see aer-identification).

Section VI — Conclusion

Short (half a page to one page), doing four things:

1. Restate the question and the headline magnitude in fresh words — no abstract sentences copied verbatim (aer-consistency flags duplication).
2. State the external-validity boundary: population, period, policy margin, and the complier/locality caveat if the design implies one.
3. Draw the one policy or theory implication the evidence supports — one step beyond the results, never three.
4. At most two sentences on specific open questions this paper makes answerable, not "more research is needed."

No new results, no new citations, no new caveats that belong in Results.

Cross-Section Consistency Rules

• The introduction's promises (contributions, magnitudes, evidence) map one-to-one onto body sections; nothing promised is undelivered and nothing major appears unannounced.
• Numbers, sample sizes, and specification labels match the tables exactly everywhere — aer-consistency audits this before submission.
• Tense: present for what the paper does and finds; past for data collection and historical events.
• Prose follows docs/style-guide.md — finding-first sentences, no filler transitions, no AI-pattern tics.

Repository Resources

Bundled with the installed skill, no repository checkout needed --- read it before the repo resources below:

• references/section-skeletons.md --- per-section skeletons and conclusion-first results narration rules

Load only the relevant resource:

• Worked results-section narration (same fictional paper as the intro example): examples/results-section-example.md
• Sentence- and paragraph-level prose rules: docs/style-guide.md
• Estimator, diagnostic, and citation defaults for Section III: docs/methods-reference.md
• Model introduction the body must stay consistent with: examples/intro-example.md
• Code that should produce every number quoted in Results: templates/stata/, templates/r/, or templates/python/

Handoff

SECTIONS DRAFTED: <Background | Framework | Data | Strategy | Results | Mechanisms | Conclusion>
ESTIMAND STATED: <yes / no — sentence>
SAMPLE FUNNEL REPORTED: <yes / no>
HEADLINE MAGNITUDE CONVERSIONS: <log-points / percent / vs-mean / vs-literature>
BACK-OF-ENVELOPE CALCULATION: <present / not applicable>
MECHANISM CHANNELS: <favored + ruled-out list>
NEXT SKILL: <aer-introduction | aer-tables-figures | aer-consistency>

Anti-Patterns

• Writing the introduction first and forcing the body to keep its promises
• Narrating tables instead of findings ("Table 4 shows the results. Column 1 reports...")
• Reporting a 0.31 log-point effect as "31 percent" (it is 36 percent), or never benchmarking the magnitude at all
• The estimand defined nowhere — "the effect" means three populations in three sections
• The sample funnel absent, so the referee reconstructs (and mistrusts) the Ns
• A conceptual framework added after the results to decorate them
• Mechanisms organized as "more regressions," or claimed with the confidence of the identified main effect
• A two-page conclusion introducing limitations Results hid, or paragraphs recycled verbatim into the abstract
• Institutional background with no source for any factual claim