Social Science Paper Writing — Econ & PolSci Tracks

Econ writing norms:

• Introduction must include: (a) motivating empirical fact, (b) research question, (c) identification strategy, (d) main findings with magnitudes, (e) contribution to literature, (f) paper roadmap
• Results stated numerically with coefficients: "A one-standard-deviation increase in X is associated with a Y-unit increase in Z (p<0.01)"
• Avoid normative language; let data speak
• Short sentences, direct prose. AER style is terse.

Political Science Track (APSR/AJPS/JOP)

1. Introduction          ← 800-1200 words; puzzle → argument → contribution
2. Theory & Hypotheses   ← Dedicated section; formal or informal theory; explicit H1, H2...
3. Research Design       ← Case selection + identification strategy
4. Data & Measurement    ← Variable operationalization + descriptive stats
5. Empirical Analysis    ← Main results + heterogeneity
6. Robustness Checks     ← Alternative measures, samples, specs
7. Discussion            ← Scope conditions + mechanisms + alternative explanations
8. Conclusion

PolSci writing norms:

• Theory section must derive testable hypotheses from logical argument
• Literature review integrated into theory — establish "what we know" before "what this paper adds"
• Research design section: justify case selection, explain identification, acknowledge threats
• More prose-heavy than economics; less equation-forward
• APSR: ~12,000 words + references; JOP: ~10,000 words

Section-by-Section Writing Guide

1. Introduction

Economics Introduction Template (5-paragraph NBER style):

¶1: Hook — motivating empirical fact or puzzle (1-3 sentences + figure/table reference)
    "Does X cause Y? [Motivating statistic.] Despite [size/importance], we know remarkably little about [mechanism/heterogeneity/context]."

¶2: This paper — what we do and how
    "This paper provides [causal/descriptive] evidence on [question]. We exploit [identification strategy] to isolate [variation]. Our setting is [context] because [institutional detail that makes identification credible]."

¶3: Main findings — numerically specific
    "We find that [main result with coefficient and significance]. This effect is [large/small] relative to [benchmark]. We also find [secondary finding]. [Mechanism evidence if available]."

¶4: Contribution to literature — 3 literatures, 2-3 sentences each
    "Our paper contributes to three strands of literature. First, [Literature A]: [Our contribution]. Second, [Literature B]: [Our contribution]. Third, [Literature C]: [Our contribution]."

¶5: Roadmap — 1 sentence per section
    "Section 2 describes the institutional context. Section 3 presents our empirical strategy..."

Political Science Introduction Template:

¶1: Open with the empirical puzzle or policy stakes
    "[Striking fact about the phenomenon]. This is puzzling because [existing theory predicts X but we observe Y]."

¶2: Existing explanations and their limits
    "Scholars have argued [Explanation A] and [Explanation B]. However, [limitation of existing work — empirical gap, theoretical underspecification, scope condition ignored]."

¶3: This paper's argument (the "I argue that...")
    "I argue that [Core Argument]. The mechanism is [causal chain]. This argument has implications for [broader stakes]."

¶4: Evidence — data, design, main findings
    "I test this argument using [Data]. The identification strategy exploits [variation]. I find [main result], consistent with [theoretical prediction]."

¶5: Contribution — theory + empirics + policy
    "This paper contributes to [Literature A] by [Contribution]. For [Literature B], the findings suggest [Implication]. The analysis also speaks to [Policy Domain]."

¶6: Roadmap

2. Theory Section (PolSci Track)

Structure:

2.1  Prior literature — what we know and the gap
2.2  Theoretical framework — assumptions, actors, mechanism
2.3  Hypotheses — derived from the theory, numbered, testable

Format for hypotheses:
H1: [Direction] + [Mechanism trigger] → [Outcome direction]
Example: "Hypothesis 1 (Compliance Mechanism): Provincial party secretaries with more extensive central government experience will be more likely to implement central directives, even at the cost of local economic performance."

PolSci theory section norms:

• One hypothesis per distinct theoretical prediction; don't bundle
• Hypotheses should be falsifiable — what would we observe if H1 is wrong?
• For game theoretic models: present intuition in prose first, formal model in appendix
• For mechanism papers: distinguish necessary vs. sufficient conditions

3. Empirical Strategy / Identification Section

This is the most important section in any causal empirical paper.

Economics Style (Empirical Strategy)

## 3. Empirical Strategy

### 3.1 Baseline Specification

Our baseline specification is:

Y_{it} = α + β·X_{it} + γ·Z_{it} + δ_i + λ_t + ε_{it}

where Y_{it} is [outcome] for unit i in year t, X_{it} is [treatment variable],
Z_{it} is a vector of controls including [list], δ_i are unit fixed effects,
and λ_t are time fixed effects. Standard errors are clustered at the [level].

### 3.2 Identification

The key identifying assumption is that, conditional on fixed effects and controls,
variation in X_{it} is uncorrelated with unobserved determinants of Y_{it}.

**Threats to identification:**
1. [Threat 1] — Addressed by [approach]
2. [Threat 2] — Addressed by [approach]

**Supporting evidence for identification:**
- Pre-trend test: [Result — Table X / Figure X]
- Placebo: [Result — Appendix Table X]
- [Additional test if applicable]

PolSci Style (Research Design)

## 3. Research Design

### 3.1 Case Selection
I focus on [context] for [N] reasons. First, [substantive reason — why this case is theoretically relevant].
Second, [data/measurement reason — why measurement is possible here].
Third, [identification reason — why causal inference is more credible here than elsewhere].

### 3.2 Identification Strategy
To identify the effect of [X] on [Y], I exploit [source of variation].
The key assumption is [parallel trends / exclusion restriction / unconfoundedness].

**Validity of this assumption:**
- [Test 1]: [Evidence]
- [Test 2]: [Evidence]

### 3.3 Estimation
I estimate [equation or description of estimator]. I cluster standard errors at the [unit]
level because [justification].

4. Results Section

Norms for both tracks:

• Lead with the headline result ("Table 2, Column 3 — our preferred specification")
• Report coefficient, standard error, significance, AND magnitude in words
• Never say "X is significant" alone — always report sign, magnitude, and interpretation
• For interaction terms: present marginal effects plot (Brambor-Clark-Golder 2006)

Economics results prose template:

Table 2 reports our baseline estimates. Column (1) includes only province and year
fixed effects; Column (3) adds the full set of controls and is our preferred specification.
The coefficient on [X] is [β] ([SE], p [<0.05/0.01]), indicating that a one-[unit]
increase in [X] is associated with a [β × unit]-[outcome unit] increase in [Y].
This represents [X%] of the sample mean / [X] standard deviations.

PolSci results prose template:

Model 3 — our preferred specification with the full set of controls and two-way
fixed effects — yields a positive and statistically significant coefficient on [X]
(β = [value], SE = [value], p < [threshold]). Consistent with Hypothesis 1,
provinces/countries/units with higher levels of [X] experience [Y-direction] [Y] outcomes.
A one-standard-deviation increase in [X] is associated with a [magnitude] change in [Y],
equivalent to [X% / comparison].

5. Robustness Section

Standard robustness battery (both tracks):

## Robustness Checks

### 5.1 Alternative Outcome Measures
[Baseline DV] → [Alternative 1] (Appendix Table X): [Result — "holds / attenuates to β=X"]
[Baseline DV] → [Alternative 2] (Appendix Table Y): [Result]

### 5.2 Alternative Treatment Measures
[Baseline IV] → [Alternative measure 1]: [Result]
[Baseline IV] → [Time-invariant version]: [Result]

### 5.3 Sample Restrictions
Excluding [subgroup] (Appendix Table Z): [Result]
Excluding [outlier provinces/countries]: [Result]
Winsorized at [1%/5%]: [Result]

### 5.4 Placebo Tests
[Temporal placebo — pre-treatment period]: [Result, null as expected]
[Spatial/unit placebo]: [Result, null as expected]

### 5.5 Alternative Estimators / Specifications
[DID → Staggered DID with Callaway-Sant'Anna 2021]: [Result]
[OLS → Poisson / Log-log / IV]: [Result]
[No clustering → Wild cluster bootstrap (Cameron et al. 2008)]: [Result]

6. Discussion / Conclusion

Discussion Section (PolSci):

Address in order:
1. Scope conditions — where does the argument apply / not apply?
2. Alternative mechanisms — rule out or acknowledge remaining threats
3. Implications for theory — what does this change in our understanding?
4. Policy implications — practical takeaways (1 paragraph, not over-sold)

Conclusion (both tracks) — 3-5 paragraphs:

¶1: Restate the question and core finding (not a copy of the abstract)
¶2: Mechanism / theoretical contribution
¶3: What we still don't know — honest limitations
¶4: Broader implications for [literature] / [policy] / [future research]

Activation Commands

When user asks to write a specific section, respond with:

1. Draft — produce a complete first draft of the section
2. Outline — produce a structured outline with key arguments for each paragraph
3. Polish — improve an existing draft for style, clarity, and journal norms
4. Review — evaluate the section against journal standards and suggest revisions

Default mode: Draft (unless user specifies otherwise)

Journal-Specific Requirements

AER / QJE / JPE (Economics)

APSR (Political Science)

JOP (Journal of Politics)

Anti-AI Writing Patterns (Built-in)

This skill produces writing that avoids:

• ❌ "This paper explores / investigates / sheds light on..."
• ❌ "Crucially, importantly, notably, interestingly" as sentence starters
• ❌ Passive voice where active is more direct
• ❌ Hedging every sentence ("may suggest," "could potentially indicate")
• ❌ Over-long sentences with multiple nested clauses
• ❌ "In conclusion, in summary, to summarize" at the start of conclusion sections

Preferred:

• ✅ Direct, declarative sentences: "We find X. Y because Z."
• ✅ Precise quantification: "a [0.66 percentage point / 12%] increase"
• ✅ Active voice: "The analysis shows" not "It is shown that"
• ✅ Varied sentence length: mix short punchy with longer explanatory

References

• references/econ-writing-guide.md — AER/QJE/JPE submission tips, common referee complaints
• references/polisci-writing-guide.md — APSR/JOP submission tips, theory section templates

Related Skills