Jhe Replication Package

Replication Package (jhe-replication-package)

When to trigger

• The analysis uses restricted health data (Medicare/Medicaid claims, hospital discharge, EHR, registry, survey-linked admin) and you must document an access path
• A Data Availability Statement and code package need assembling for submission
• A referee or editor asks whether the results are reproducible despite data being non-public
• You are deciding what can be deposited publicly versus what stays behind a Data Use Agreement / IRB

What JHE actually requires (and what it does not)

JHE follows the Elsevier Research Data policy: authors are encouraged to deposit data in a repository and to state data availability at submission — this is not the AEA's mandatory pre-publication openICPSR build-and-verify gate (检索于 2026-06；以官网为准). That softer baseline is precisely why credibility here rests on how well you document the path to reproduction, not on a public dataset that often cannot exist. The health-economics reality is that most rich data is restricted by privacy law (HIPAA), Data Use Agreements, and IRB. The professional standard JHE referees increasingly expect: code is shareable even when data is not, the Data Availability Statement is honest and specific, and another researcher could in principle obtain the same data and re-run the code.

Build the package around the data's access tier

Documentation that earns trust

1. Data Availability Statement, honest and specific. Name the data, the holder, the access mechanism, and the realistic path/cost/wait — not "available on request" with no detail.
2. Share the code regardless of data access. Cleaning, construction, and analysis scripts in run order, with a master script and a manifest mapping each exhibit to the code that makes it.
3. Pin the environment. Software versions, packages, seeds; for skewed-spending estimators (two-part, GLM) state the exact implementation so results are reproducible.
4. Separate construction from analysis so a reader with the same DUA can rebuild the analysis sample.
5. Respect privacy in the package itself. No re-identifiable cells, small-cell suppression honored, no PHI in logs or example outputs.
6. Map every table and figure to its script so the reproduction is auditable exhibit by exhibit.

Checklist

• Data Availability Statement names the data, holder, access mechanism, and realistic path/cost/wait
• Code shared even though data is restricted; master script + exhibit-to-code manifest present
• Environment pinned (versions, packages, seeds); skew/zero estimators reproducible
• Construction code separated from analysis code; sample-build documented
• Privacy honored: small-cell suppression, no PHI in code/logs/outputs
• Public-data portions are genuinely one-command reproducible
• DUA / IRB constraints stated so a referee understands what cannot be posted

Why this matters more at JHE than at the AEA journals

Because JHE only encourages deposit rather than running a mandatory pre-publication build-and-verify, the burden of credibility shifts onto the author. There is no Data Editor who will catch a broken pipeline before publication, which cuts both ways: less friction at acceptance, but a higher reputational cost if a reader later cannot reproduce your work from the materials you provided. Treat the package as your own quality control — a colleague should be able to run the public portions today and, with the documented access path, rebuild the restricted portions later. That standard, not the letter of the policy, is what protects the paper.