---
name: restat-replication-package
description: Use when assembling the data-and-code replication package for a The Review of Economics and Statistics (REStat) manuscript under the journal's Data and Code Availability Policy — the deposit to the REStat Harvard Dataverse with a README that permits replication. Builds the package; it does not run the primary analysis.
---

# Replication Package & Data/Code Policy (restat-replication-package)

## When to trigger

- The paper is approaching acceptance and a data/code deposit will be required
- You use proprietary or restricted data and must plan for the policy before paying the fee
- A README, master script, or directory structure needs to be assembled
- You are unsure what REStat's Data and Code Availability Policy actually requires

## The REStat data/code bar (refreshed 2026-06-20; live-check the official policy page)

REStat operates a **Data and Code Availability Policy** for empirical papers. The signature features that distinguish REStat from its siblings:

- **Deposit to the REStat Harvard Dataverse** (`dataverse.harvard.edu/dataverse/restat`) — REStat uses a **proper Dataverse archive**, NOT openICPSR (which the AEA journals use) and NOT the JAE Data Archive. Getting the repository right is a load-bearing REStat detail.
- **Post data and code, with documented README files, before publication.** The goal is to **provide enough information to permit replication** of the results in the paper.
- **Proprietary / restricted data:** indicate this **at submission**. When data cannot be posted, you must **document the data and provide instructions for how other researchers can obtain it**. Practical note: if your paper uses proprietary data, the journal asks you **not to pay the submission fee until the editorial office confirms your data comply with the policy**.
- Policy generally applies to papers from **volume 92 onward** per the REStat Harvard Dataverse API description.

## Package architecture (build it as you go)

```
restat-replication/
  README.(md|pdf)        # the heart of the deposit — see contents below
  data/
    raw/                 # source data (or access instructions if restricted)
    processed/           # built analysis files, regenerated by code
  code/
    00_run_all.do/.R/.py # one master script regenerating EVERY exhibit
    01_build.*           # raw → processed
    02_analysis.*        # processed → estimates
    03_exhibits.*        # estimates → tables/figures
  output/
    tables/  figures/    # exactly what appears in the paper + appendix
```

## README contents (what permits replication)

- **Data availability statement:** each dataset, its source, terms, and (if restricted) exact access instructions; flag proprietary data clearly.
- **Computational requirements:** software + versions (Stata 18 MP / R 4.x / Python 3.x), packages with versions, OS, approximate run time and hardware.
- **Instructions:** how to run `00_run_all` from raw data to every table/figure; what each script produces.
- **Mapping:** a table linking each exhibit number in the paper/appendix to the script and line that generates it.
- **Seeds:** set and report seeds for any simulation / bootstrap / randomization inference.

## Checklist

- [ ] Target archive confirmed: **REStat Harvard Dataverse** (not openICPSR, not JAE)
- [ ] One master script regenerates every table and figure from raw (or processed) data
- [ ] README has data-availability statement, computational requirements, run instructions, exhibit↔script map
- [ ] Software/package versions pinned; seeds set and reported
- [ ] Proprietary/restricted data flagged **at submission**; access instructions documented
- [ ] (Proprietary data) submission fee **not paid** until editorial office confirms policy compliance
- [ ] Every number in the paper traces to deposited code (consistency with `restat-tables-figures`)

## Anti-patterns

- Depositing to openICPSR or the JAE archive — REStat uses its own **Harvard Dataverse**
- Treating the package as a post-acceptance chore — assemble it alongside the analysis
- A README that lists files but gives no run instructions or exhibit↔script mapping
- Unpinned package versions / unset seeds — results that will not reproduce
- Discovering only at acceptance that proprietary data cannot be posted (declare at submission)
- Paying the submission fee on a proprietary-data paper before compliance is confirmed

## Output format

```
【Archive】REStat Harvard Dataverse confirmed? [Y/N]
【Master script】00_run_all regenerates all exhibits from data? [Y/N]
【README】availability stmt + compute reqs + instructions + exhibit↔script map? [Y/N]
【Versions/seeds】software+packages pinned; seeds set/reported? [Y/N]
【Proprietary data】flagged at submission; access instructions; fee held? [Y/N / n/a]
【Traceability】every paper number ← deposited code? [Y/N]
【Next step】restat-referee-strategy
```