agentleFS
Barrios88Barrios88/barrios-skills/skills/writing-and-review/econ-writing-plus/SKILL.md

econ-writing-plus

Skill63 starsChanged 5 months agoMITView on GitHub

Supplements economic-writing, econ-humanizer, academic-paper-writer, research-grants, and referee-report with advanced econ-specific guidance — (1) identification-strategy-specific writing conventions for 13 methods (RCT, DiD, IV, RDD, Synthetic Control, Synthetic DiD, Structural, Descriptive, Bunching, Shift-Share/Bartik, Event Studies, ML Causal), (2) modern DiD estimator guidance (Callaway-Sant'Anna, Sun-Abraham, de Chaisemartin-D'Haultfoeuille, Goodman-Bacon decomposition), (3) AEA replication-package standards, (4) AI-use disclosure policies (AEA, Econometric Society), (5) ERC grant structure, (6) field conventions for macro, trade, finance, (7) title evaluation framework, (8) economic-significance translation rules. USE when drafting/reviewing identification-strategy descriptions, AEA-journal submission prep, ERC proposals, or when the existing econ-writing skills don't cover method-specific narrative conventions. Synthesizes Cochrane, McCloskey, Head, Bellemare, Shapiro, plus modern DiD literature (Goodman-Bacon 2021, Callaway & Sant'Anna 2021, Sun & Abraham 2021, de Chaisemartin & D'Haultfoeuille 2020, Borusyak-Hull-Jaravel 2022, Goldsmith-Pinkham-Sorkin-Swift 2020, Arkhangelsky et al. synthetic DiD).

The file

We split this skill into the rules it gives an agent

7 rules. About 0 are common, 7 are distinctive, and 0 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈7

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 0

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    When in doubt: default to the broader econ skills for general structure (abstract, intro, conclusion, review).

    line 13
  2. Distinctive:

    If a DiD paper uses staggered treatment adoption, do NOT present only two-way fixed effects.

    line 43
  3. Distinctive:

    Every empirical paper submitted to AEA journals (and increasingly Econometrica, EJ, field journals) must include a replication package …

    line 74
  4. Distinctive:

    Core rule: If AI drafted a paragraph, read it as if a careless RA wrote it — verify every fact, citation, and number.

    line 108
  5. Distinctive:

    Use 3-year or 5-year panel intervals (not annual) with specific justification

    line 144
  6. Distinctive:

    Never report a coefficient without translating it.

    line 190
  7. Distinctive:

    When testing multiple outcomes or subgroups:

    line 205

≈ “Common” is an estimate: at least 1 in 5 of the 372,429 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Barrios Skills — John Barrios's curated workflow for economists and accountants. Prioritize reproducible empirical work, clear identification language, and journal-ready output.

Econ Writing Plus

Advanced, method-specific additions to the existing economic-writing skill set. Use alongside economic-writing, econ-humanizer, research-grants, pre-submission-review, and referee-report — not instead of them.

When in doubt: default to the broader econ skills for general structure (abstract, intro, conclusion, review). Invoke this skill for:

  • Writing or reviewing an identification section — see identification-strategies.md
  • Preparing AEA / Econometrica / EJ submission compliance
  • ERC grant proposals (the existing research-grants skill only covers NSF/NIH/DOE/DARPA)
  • Field-specific conventions in macro, trade, or finance
  • Title evaluation
  • Translating statistical significance into economic magnitudes

1. IDENTIFICATION STRATEGIES (Primary Reference)

See identification-strategies.md for detailed writing conventions for each of the following:

  • Randomized Controlled Trials (RCTs) — ITT vs. LATE, attrition, spillovers, take-up
  • Difference-in-Differences (DiD) — parallel trends, event study, modern estimators
  • Instrumental Variables (IV) — instrument relevance, exclusion restriction, complier population, Anderson-Rubin CIs
  • Regression Discontinuity (RDD) — running variable, McCrary density test, bandwidth sensitivity, LOCAL interpretation
  • Synthetic Control — donor pool, pre-treatment fit, placebo inference
  • Synthetic DiD (Arkhangelsky et al.) — doubly robust, unit + time weights
  • Structural Estimation — identifying vs. functional-form assumptions, model fit, counterfactuals as payoff
  • Descriptive / Measurement — explicitly disclaim causality, data construction as contribution
  • Bunching Estimation (Saez, Kleven) — counterfactual distribution, optimization frictions
  • Shift-Share / Bartik — share vs. shift exogeneity (Goldsmith-Pinkham et al.; Borusyak-Hull-Jaravel), leave-one-out
  • Event Studies — pre-period coefficients, normalization, staggered-event estimators
  • ML for Causal Inference — Causal Forests, double/debiased ML, cross-fitting
  • Multi-strategy papers — name a primary, frame others as complementary

Modern DiD estimator guidance (critical for current submissions)

If a DiD paper uses staggered treatment adoption, do NOT present only two-way fixed effects. Use:

  • Goodman-Bacon (2021) decomposition to show which 2×2 comparisons drive the TWFE estimate (exposes negative-weight problem)
  • Callaway & Sant'Anna (2021) for heterogeneous treatment effects over time
  • Sun & Abraham (2021) for event-study specifications under staggered adoption
  • de Chaisemartin & D'Haultfoeuille (2020) for the no-sign-reversal assumption

Present results from BOTH traditional TWFE and a robust estimator. If they differ, explain why (negative weights, treatment effect heterogeneity). The event-study plot should come from the robust estimator, not the TWFE version.

Adapting the introduction by paper type

Paper TypeHook StrategyParagraphs 4-6Key Threat
RCTPolicy relevance of interventionITT and LATE estimatesAttrition, spillovers, external validity
DiDPolicy change / natural experimentMain DiD + event studyParallel trends, anticipation
IVThe instrument and why it's cleverOLS vs. IV comparisonExclusion restriction, weak instruments
RDDThe cutoff and its stakesRD estimate + bandwidth sensitivityManipulation, other discontinuities
Synthetic ControlTreated unit + eventSynthetic vs. actual trajectoryPre-treatment fit, donor pool
Synthetic DiDPolicy change + few treated unitsSynth-DiD vs. DiD vs. SCParallel trends, synth-control fit
StructuralQuestion requiring a modelCounterfactual resultsModel assumptions, external validity
TheoryPuzzle the model resolvesMain proposition + intuitionRobustness to assumptions
DescriptiveWhy the fact mattersKey patterns with magnitudesMeasurement validity, sample selection
BunchingPolicy kink/notch + affected groupElasticity + bunching plotOptimization frictions, manipulation
Shift-ShareShock + local exposureMain estimate + leave-one-outShare exogeneity, shock exogeneity
Event StudyEvent + stakesPlot + key coefficientsPre-trends, anticipation
ML/CausalPrediction or heterogeneity questionML vs. parametric comparisonOverfitting, interpretability

2. AEA REPLICATION PACKAGE STANDARDS

Every empirical paper submitted to AEA journals (and increasingly Econometrica, EJ, field journals) must include a replication package meeting the Data Editor standards.

README structure (Social Science Data Editors template):

  1. Data Availability & Provenance — who owns the data, how to access
  2. Dataset List — every file, source, description
  3. Computational Requirements — software, packages, runtime, hardware
  4. Programs Description — what each script does, execution order
  5. Replicator Instructions — step-by-step, reproducible without manual intervention

Directory structure:

data/raw/        <- original downloaded data, never modified
data/analysis/   <- cleaned analysis-ready datasets
code/            <- all scripts, numbered in execution order
results/         <- generated tables and figures

Rules:

  • Cite every dataset in the References section with standard in-text citations
  • Code must reproduce all results without manual intervention (one config file for paths is the only exception)
  • Map every table and figure to a specific program file
  • For restricted-access data: include a Data Availability Statement explaining the application procedure, wait times, and monetary costs
  • Include LICENSE.txt (default: CC-BY 4.0)

3. AI-USE DISCLOSURE

AEA policy: AI cannot be listed as an author. Disclose AI use during submission. Econometric Society: Co-authors must sign a responsibility statement accepting responsibility for all content.

What to disclose: drafting assistance, code generation, literature search, data analysis suggestions. Typically no disclosure needed: spell-check, grammar tools, LaTeX formatting.

Core rule: If AI drafted a paragraph, read it as if a careless RA wrote it — verify every fact, citation, and number. You are responsible for all AI-generated content.


4. ERC GRANT STRUCTURE

(Complements research-grants, which covers NSF/NIH/DOE/DARPA.)

ERC reviewer priorities differ from NSF:

  • PI track record is weighted heavily — foreground publication record, citation impact, career trajectory
  • "High-risk, high-gain" framing is mandatory — ERC explicitly funds frontier risk-taking, not incremental extensions
  • Societal relevance section should connect to real policy questions, not abstract "advancing knowledge"

Proposal structure:

  1. Extended synopsis (5 pages): question, approach, expected impact — written for a smart non-specialist
  2. Scientific proposal (14 pages for StG/CoG, 15 for AdG): detailed research plan, methodology, feasibility
  3. CV + track record (2 + 2 pages): highlight your contributions, not co-authors'
  4. Budget justification: link every cost to a specific research activity

Common mistake: writing like a finished paper. A proposal sells the plan, not findings. Emphasize what you WILL learn.


5. FIELD-SPECIFIC CONVENTIONS

Macroeconomics

  • Papers are longer (40–60 pages normal); the "under 40 pages" rule does not apply
  • Calibration tables are standard: parameter | value | source/target moment
  • Impulse response functions (IRFs) are the primary visualization, not regression tables
  • "Model Fit" section comparing model moments to data moments is expected
  • DSGE papers: describe steady state, log-linearization or solution method, shock specification
  • Phrase results as "the model generates X" rather than "I find X"

Trade

  • Gravity estimation: use PPML, multilateral resistance controls, explicit fixed-effects structure
  • General equilibrium counterfactuals are expected in structural trade papers
  • Use 3-year or 5-year panel intervals (not annual) with specific justification

Finance

  • Abstract limit is often 100 words (not 150) — check target journal
  • Fama-MacBeth regressions and portfolio-sort presentation are standard
  • Variable winsorization at 1%/99% is expected and must be reported
  • Chicago Manual of Style citation at some journals (differs from AEA)

Development / Applied Micro with RCTs

  • Pre-registration is nearly mandatory
  • Include a CONSORT-style flow diagram (enrollment → randomization → attrition → analysis)
  • Report cost-effectiveness alongside treatment effects
  • Balance tables belong prominently in the paper, not the appendix

6. TITLE EVALUATION FRAMEWORK

Score any candidate title 1–10 on each dimension:

  1. Clarity — Can a non-specialist understand the topic in one reading?
  2. Specificity — Are treatment/cause and outcome/effect both named?
  3. Length — Under 12 words is ideal; under 15 is acceptable
  4. Memorability — Would someone remember it at a conference?
  5. No methodology — Does it emphasize the finding, not the method? (Exception: you invented the method.)

Good:

  • "The Oregon Health Insurance Experiment: Evidence from the First Year"
  • "The China Syndrome: Local Labor Market Effects of Import Competition"
  • "Pollution and Mortality: Evidence from the 1952 London Fog"

Bad:

  • "A Difference-in-Differences Analysis of Education Policy" (methodology, not finding)
  • "On the Relationship Between Various Factors and Economic Outcomes" (says nothing)
  • "Essays on Labor Economics" (acceptable for dissertation, never for a paper)

Formulas that work:

  • "The Impact of [D] on [Y]: Evidence from [Context]"
  • "[D] and [Y]" (shorter, acceptable)
  • Theory: name the key mechanism, not the technique
  • Structural: "[Counterfactual Question]: Evidence from [Context]"

7. ECONOMIC SIGNIFICANCE (Not Just Statistical)

Never report a coefficient without translating it. Rules:

  • Translate to meaningful units: dollars, percentage points, standard deviations, policy benchmarks
  • Compare effect size to at least one of:
    • Mean of the dependent variable
    • A well-known intervention (e.g., "equivalent to 60% of the Moving to Opportunity effect")
    • A policy-relevant threshold (e.g., the poverty line)
  • For elasticities, state whether computed at the mean, median, or arc
  • Back-of-envelope calculations are encouraged — show the reader what the number means in aggregate
  • Null results: distinguish "precisely estimated zero" (report CI and rule out effects above X) from "imprecisely estimated" (wide CI). Discuss statistical power. If pre-registered, emphasize that.

8. MULTIPLE TESTING

When testing multiple outcomes or subgroups:

  • Acknowledge the problem explicitly
  • Report family-wise error rate (Bonferroni, Holm) or false discovery rate (Benjamini-Hochberg) corrections
  • At minimum, flag results surviving the correction
  • For heterogeneity: pre-specify subgroups based on theory, not data mining; report the number of subgroups tested

9. HOW TO USE THIS SKILL ALONGSIDE OTHERS

TaskPrimary skillAdd from this skill
Draft an introeconomic-writingPaper-type hook table (§1), title eval (§6)
Write identification sectioneconomic-writingFull identification-strategies.md
Rewrite for voiceecon-humanizer— (no overlap)
Pre-submission auditpre-submission-reviewAEA replication check (§2), AI disclosure (§3)
Referee responsereferee-report— (no overlap)
NSF/NIH grantresearch-grants—
ERC grantresearch-grants + this skill (§4)§4 is the delta
Finance papereconomic-writing§5 (100-word abstract, FM regressions)
Macro papereconomic-writing§5 (IRFs, calibration tables)
Staggered DiDeconomic-writing§1 modern-estimator guidance

Synthesized from the hanlulong/econ-writing-skill corpus (50+ sources: Cochrane, McCloskey, Head, Bellemare, Shapiro, Goldin & Katz, Glaeser, Kremer, Nikolov, Schwabish, Evans, Dudenhefer) plus the modern DiD and shift-share literatures.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.