/aejmic-replication-package

name: aejmic-replication-package
description: Use when assembling the proof appendix and any code/data deposit for an American Economic Journal: Microeconomics (AEJ: Micro) manuscript under the AEA Data and Code Availability Policy. Covers proof appendices for pure theory plus numerical/structural/experimental code; it does not run your estimation.

Replication Package: Proofs + Code (aejmic-replication-package)

For AEJ: Micro the "replication package" has **two faces**: the **proof appendix** that makes every theory claim verifiable, and, for any paper with **data, code, experiments, or numerical results**, an AEA Data and Code Repository deposit. Pure-theory papers still deposit any **numerical/simulation code** used to generate examples or figures.

When to trigger

• Proofs are scattered, abbreviated, or rely on "it can be shown"
• The paper has numerical examples, simulations, structural estimation, or an experiment with no deposit prepared
• You are preparing for the AEA Data Editor check (administered before publication)
• A referee or editor flags reproducibility

The proof appendix (every AEJ: Micro paper)

• **Self-contained proofs of all stated results.** Key proofs belong in the paper (main text or appendix); do not exile a load-bearing proof to supplementary material.
• **Lemma scaffolding:** state and prove auxiliary lemmas before the main theorem; reference them precisely.
• **Verify, do not assert:** no "it can be shown that" for a claim the result depends on; complete the argument or cite a precise source.
• **Match the statement:** the proof establishes exactly what the proposition claims (no gap between the body statement and what is proved).

Code / data deposit (papers with data, code, experiments, or numerical results)

The AEA operates a **Data and Code Availability Policy** administered by the **AEA Data Editor** (currently Lars Vilhuber — 检索于 2026-06，以官网为准), with materials deposited to the **AEA Data and Code Repository on openICPSR**. Build it as you go.

• **One master script** (`run_all`) regenerating every table, figure, and numerical example from inputs.
• **Pin versions:** `requirements.txt` / `conda` (Python), `renv.lock` (R), `Project.toml` / `Manifest.toml` (Julia), recorded Stata `ssc`/`net` versions.
• **Set and report seeds** for any simulation, bootstrap, or randomization.
• **README** mapping each exhibit to the script that produces it; document any restricted-data or partial-reproduction scope.
• **Pure-theory papers:** deposit the code behind numerical examples / figures even when there is no dataset.
• **Experiments:** include instructions, z-Tree/oTree code, raw and analysis data, and pre-registration links.

Checklist

• [ ] All stated results have self-contained proofs; none rely on "it can be shown"
• [ ] Auxiliary lemmas stated and proved before they are used
• [ ] Each proof matches exactly what its proposition claims
• [ ] (If any data/code/numerics) one master script regenerates all exhibits
• [ ] Versions pinned; seeds set and reported
• [ ] README maps every exhibit to its script; restricted/partial scope documented
• [ ] Pure-theory numerical-example code deposited even with no dataset
• [ ] Experiment materials (instructions, code, data, pre-registration) included

Anti-patterns

• A "Proof." that asserts rather than argues the load-bearing step
• A load-bearing proof hidden in an un-checked supplementary file
• Numerical figures with no deposited code ("available on request")
• Unpinned dependencies / unset seeds — results not reproducible by the Data Editor
• Deferring the whole package to acceptance, then scrambling under the Data Editor deadline

Worked vignette (illustrative)

A persuasion paper has a clean Proposition 2 but its proof says "concavifying the value function yields the cutoff." For the appendix: state the auxiliary lemma (the value function's concave closure equals the indirect utility), prove it, then derive the cutoff explicitly — no hand-wave. The two numerical figures are generated by `make_figures.py`; deposit it with a fixed seed and a README line mapping Figure 3 → `make_figures.py`, even though there is no dataset.

Output format

【Proof appendix】all results proved, self-contained, no "it can be shown"? [Y/N]
【Lemma scaffolding】auxiliary results proved before use? [Y/N]
【Code/data deposit needed?】[yes — data/structural/experimental/numerical | theory-only numerics]
【Master script + pinned versions + seeds】[Y/N]
【README exhibit→script map】[Y/N]
【Next step】aejmic-referee-strategy then aejmic-submission

Supplementary resources

• [`../../resources/code/`](../../resources/code/) — runnable Stata/Python skeleton for the empirical/structural subset
• [`../../resources/README.md`](../../resources/README.md) — when the code kit applies vs. theory proof-appendix craft