brycewang-stanford/aer-introduction

AER Introduction

Overview

The introduction is the only part of the paper most editors read in full. Top-5 desk rejection decisions are typically made on pages 1-3. This skill produces an introduction that survives that filter and an abstract that fits AER's 100-word constraint.

Two non-negotiable AER formatting facts:

1. No "Introduction" heading. The introductory section is unlabeled and begins immediately after the title and abstract.
2. Abstract ≤ 100 words. Roughly 4-5 sentences. Manuscripts exceeding the limit are returned without review.

When to Use

• Drafting an introduction from scratch
• Rewriting an introduction that drew a desk rejection
• Compressing a 250-word working-paper abstract to AER's 100-word limit
• The introduction is over 3 typeset pages and needs surgery
• The user has results but cannot explain why they matter in one paragraph

The Five-Paragraph Formula (Keith Head)

Every AER-style introduction has exactly five components, in this order:

Paragraph 1 — The Hook

Open with one of:

• Y matters. Welfare consequences, magnitudes, policy stakes.
• Y is puzzling. A stylized fact existing theory cannot explain.
• Y is controversial. Two camps disagree; new evidence resolves the question.
• Y is big. A first-order phenomenon (the service sector, urban inequality, the trade balance).

Two to three sentences. Cite one number that anchors the magnitude. Do not yet name the paper's contribution.

Paragraph 2 — The Question

State exactly what this paper does:

This paper [estimates / documents / characterizes] [the causal effect of X on Y /
the response of Y to shock S / the distribution of Y in setting D].

One paragraph. Define the unit of observation, the outcome, and the variation that identifies the answer. Avoid the word "we" if possible; use "this paper."

Paragraph 3 — The Identification (or the Model)

Empirical papers: name the identification strategy in one sentence, then explain in 1-2 paragraphs what variation drives identification and why the parallel trends / exclusion / smoothness assumption is credible in this setting.

Theory papers: name the modeling discipline — what's tractable, what's general, what the comparative statics give you.

This paragraph is where desk rejection happens. Editors check whether the method matches the claim. If you write "we examine the relationship between X and Y" while using OLS with controls, the paper is desk-rejected for methodology mismatch.

Paragraph 4 — The Antecedents and Value-Added

Often the single most important paragraph for surviving referee review. Two halves:

Antecedents (1-2 paragraphs). Position the paper relative to its 3-6 closest published predecessors. Be specific: cite by author-year, identify what each did, and what each missed.

Value-added (1 paragraph or 3 bullet points). State approximately three contributions relative to the antecedents. These are the sentences the referee will quote in their report. Each contribution should make sense only in light of the prior work — otherwise it belongs in the Question paragraph.

Avoid:

• "To the best of our knowledge, this is the first paper to ..." (unverifiable, often false)
• "We contribute to the literature on X" (vacuous unless you say how)
• Padding the contribution list past four items

Paragraph 5 — The Roadmap

One short paragraph. "Section 2 describes the data. Section 3 presents the empirical strategy. Section 4 reports results. Section 5 explores mechanisms. Section 6 concludes."

Some AER authors omit the roadmap entirely. Acceptable for short papers (AER: Insights). Required for full-length AER.

The 100-Word Abstract

AER abstracts are 100 words maximum, including all numbers. The high-impact pattern allocates word budget as:

| Function | Sentences | Words | |-----------------------------------|-----------|-------| | Question or setting | 1 | 15-20 | | Method / data / identification | 1 | 15-20 | | Main quantitative result | 1-2 | 30-40 | | Implication | 1 | 15-20 |

Allocate the most words to results. Resist motivation-heavy abstracts — that is what the introduction's first paragraph is for. High-citation AER abstracts dedicate three of four sentences to findings.

Abstract Template

[Setting and question — 1 sentence].
[Data and identification — 1 sentence].
[Main result with magnitude and sign — 1-2 sentences].
[Implication — 1 sentence].

Word-Count Discipline

If the draft is over 100 words:

1. Delete every adjective that does not change the result
2. Replace clauses with semicolons
3. Drop the implication sentence — the introduction handles it
4. Replace "we find that ... " with active "X raises Y by Z%"
5. Numbers count as one word each; don't burn budget on "approximately"

AER-Specific Formatting

• No "Introduction" heading. Begin the introductory material immediately after the abstract.
• No vertical space markup. Use \section and \subsection; do not insert \vspace or \bigskip.
• Footnotes, not endnotes. Inline citations use \cite{} (author-year), not numbered references.
• Style emphasis sparingly. No \textbf for emphasis in body text — italics only, rarely.
• No "Section 1" label. The first numbered section after the introduction is \section{Data} or whatever the title is. AER convention treats the intro as section 0.

Common Failure Modes

• Three-page introduction that never names the identification strategy → desk reject
• Five contributions in the value-added paragraph, one of them weak → referee picks the weak one
• Abstract with 130 words after a "minor tweak" → editor bounces submission
• Hook paragraph that pitches a methods contribution when the paper is empirical (or vice versa)
• Burying the magnitude of the main result until section 4
• Lit review that runs for 2 pages and does not say what this paper adds

Repository Resources

When working from the AER-skills repository or plugin bundle, read examples/intro-example.md only when the user asks for a model introduction, a concrete before/after rewrite, or abstract compression.

Handoff

ABSTRACT WORD COUNT: <n>/100
INTRODUCTION PARAGRAPHS: Hook | Question | Identification | Antecedents+Value | Roadmap
CONTRIBUTIONS LISTED: <n> (target: 3, max 4)
KILL SWITCHES: <list of remaining red flags, or "none">
NEXT SKILL: <aer-tables-figures | aer-submission>

Reference Pattern

A canonical AER-style intro architecture (paragraph-by-paragraph):

1. Hook. "The richest 1% of US households hold X% of wealth. This share has risen by Y percentage points since Z."
2. Question. "This paper estimates the causal effect of [policy] on top-wealth concentration using [variation]."
3. Identification. "We exploit [quasi-experiment]. The identifying assumption is [...]. We validate this by [pre-trends test / placebo / institutional argument]."
4. Antecedents + value-added. "Three prior papers (A 2019, B 2021, C 2023) study related questions. A used [method] but [limitation]. B documented [fact] but did not establish causation. C addressed causation but in [different setting]. This paper makes three contributions: first, [...]; second, [...]; third, [...]."
5. Roadmap. "Section 2 describes [...]. Section 3 [...]."