---
name: brycewang-stanford/aeri-replication-package
source: https://app.decimal.ai/s/brycewang-stanford-aeri-replication-package@1/SKILL.md
source_sha256: 8abd6c4d68c4
---

# Replication Package — AEA Data Editor Ready (aeri-replication-package)

## When to trigger

- A conditional accept is likely or in hand and the deposit is not assembled
- Unsure what the AEA Data and Code Availability Policy requires
- Results were produced ad hoc and cannot yet be regenerated by one script
- Restricted/proprietary data raise an access question for the data editor

## The AEA reproducibility regime (检索于 2026-06；以官网为准)

AER: Insights enforces the **AEA Data and Code Availability Policy**, administered by the **AEA Data Editor** (Lars Vilhuber). Before publication, authors deposit **data, code, and documentation sufficient to permit replication of every result in the paper and approved appendices** in the **AEA Data and Code Repository (openICPSR)**; the data editor's team **verifies reproducibility before the paper is published**. The short format does **not** shrink this requirement — the Supplemental Appendix results are in scope too. Build the package as you go, not after the conditional accept, because the verification step is a gate to publication.

## What the package must contain

- **All code** to go from raw (or provided) data to every table, figure, and in-text number, organized and runnable.
- **Data** that are non-exclusive to the authors; where data are restricted, provide access instructions and the **code that would run on them**, plus a clear restricted-data statement.
- **A master script** (`run`/`00_master`) that regenerates every exhibit end-to-end.
- **A README** following the AEA/Social Science Data Editors **README template**: data sources and provenance, file-by-file description, computational requirements (software + versions), instructions to reproduce, and a **mapping from each exhibit to the code that produces it**.
- **Seeds set and reported** for any simulation, bootstrap, or randomization inference.
- **Disclosure** of data citations and a data availability statement.

## Short-paper specifics

- **Exhibit-to-code map is small but must be complete** — five exhibits plus the appendix; every in-text number too.
- **Online appendix is in scope.** Material you pushed out of the main text to fit the cap ([`aeri-robustness`](../aeri-robustness/SKILL.md), [`aeri-theory-model`](../aeri-theory-model/SKILL.md)) must still be reproducible.
- Keep a clean separation: `data/raw` (read-only) → `code` → `output`; see [`../../resources/code/`](../../resources/code/) for a runnable skeleton.

## Timing relative to the fast review

Because AER: Insights decisions are **conditional accept or reject** with no traditional R&R
([`aeri-referee-strategy`](../aeri-referee-strategy/SKILL.md)), a conditional accept moves quickly to the
data-editor stage. There is no long revision phase in which to build the package retroactively. Have the
deposit essentially ready **at submission**, so a conditional accept does not stall on reproducibility. The
data editor's check is a **gate to publication**, not a formality — non-reproducible code delays or blocks
the paper regardless of the editor's decision.

## Worked vignette (illustrative)

A fictional staggered-DiD paper has five exhibits and a dozen appendix tables. The package contains a
`00_master` script that runs `01_clean` → `02_descriptive` → `03_did_modern` → `08_robustness` →
`09_tables`, regenerating Figures 1–2 and Tables 1–3 plus every appendix table, with seeds set in the
bootstrap step. The README maps "Figure 1 ← code/03_did_modern.do, lines 40–80" and lists Stata 18 with
pinned `csdid`/`reghdfe` versions. The restricted payroll extract is described with an application link and
the exact code that would run on it. This is what passes the AEA Data Editor on the first pass.

## Checklist

- [ ] One **master script** regenerates every exhibit and in-text number
- [ ] **README** per the AEA/SSDE template: sources, file list, requirements, exhibit→code map
- [ ] Data non-exclusive, or a restricted-data statement + access instructions + runnable code
- [ ] Software and **package versions pinned**; seeds set and reported
- [ ] **Supplemental Appendix results reproducible** too (in scope)
- [ ] Deposit prepared for the **AEA Data and Code Repository (openICPSR)**
- [ ] Data citations and availability statement included

## Anti-patterns

- Assembling the package only after the conditional accept (verification is a publication gate)
- A README missing the exhibit→code mapping or version information
- Excluding appendix results from the reproducible set
- Hard-coded paths, unset seeds, or manual steps the master script does not capture
- Treating "short paper" as exempting any result from reproducibility

## Output format

```
【Master script】regenerates all exhibits + in-text numbers? [Y/N]
【README】sources / file list / versions / exhibit→code map present? [Y/N]
【Data status】public deposit / restricted (statement + access + code)?
【Seeds & versions】set and pinned? [Y/N]
【Appendix in scope】appendix results reproducible? [Y/N]
【Deposit target】AEA Data and Code Repository (openICPSR)
【Next step】aeri-referee-strategy (or aeri-submission if submitting)
```