brycewang-stanford Replication & Data Policy (jcf-replication-and-data-policy)
When to trigger
- Drafting the data-availability statement at submission
- Deciding what code/data to deposit and where, given vendor licenses
- Preparing the archive that supports the paper's tables and figures
JCF data policy (verified; re-confirm on the official guide)
JCF follows Elsevier "Research Data" Option C. Authors are required to state data availability at submission and are encouraged to deposit research data — including software, code, models, and algorithms — in a relevant repository and cite/link it, or provide a statement explaining why the data cannot be shared.
There is no dedicated JFE-Data-Archive-style mandatory replication-package mandate for JCF (待核实 / none found): the policy is the standard Elsevier data-availability-statement framework, not a finance-society mandatory code/data archive.
What this means in practice
- Most JCF data come from licensed vendors (Compustat, CRSP, SDC/Refinitiv, DealScan, BoardEx, ISS, Execucomp). You cannot redistribute the raw data.
- For licensed data, share code + variable-construction details + access instructions, and write a statement that data are available from the named vendor under license.
- For self-collected or public data (e.g., EDGAR text), deposit it with a DOI (Mendeley Data, Zenodo, openICPSR) and cite the deposit.
Build checklist
- A data-availability statement drafted for submission (Option C wording)
- One master script (
run_all) regenerating every table/figure - Software/package versions pinned; seeds set for bootstrap/simulation
- Variable-construction documentation for each licensed source
- A DOI deposit for any shareable data/code, cited in the paper
- A clear statement of why any data cannot be shared (license/proprietary)
Source-by-source sharing decisions
Source | Deposit raw? | What to deposit instead
Compustat / CRSP / CCM | No (license) | Pull scripts, link rules, filter code, variable formulas
SDC Platinum / Refinitiv | No | Deal-screen criteria, query dates, cleaning code
DealScan | No | Facility-match code; loan-level filter documentation
BoardEx / ISS / Execucomp | No | Classification rules; year-by-year coverage notes
Hand-collected governance | Usually yes | The dataset itself with a DOI + collection protocol
EDGAR / public filings text | Yes | Scraper code, parsed extracts, dictionary files
Derived firm-year variables | Check terms | Many vendors bar derived-data sharing — confirm first
When in doubt about derived data, describe the construction completely rather than deposit — a license breach is worse than an opaque archive.
Worked statement: mixed licensed and hand-collected inputs
Illustrative Option C wording for a governance paper combining BoardEx with hand-collected charter amendments:
Data availability: Board and director data were obtained from BoardEx under license and cannot be
redistributed; access instructions and full variable-construction code are provided at [DOI]. The
hand-collected charter-amendment dataset and all analysis code are deposited at [repository, DOI].
Compustat and CRSP inputs are available from WRDS under subscription.
Adapt the names and repositories; the structure — what is licensed, what is deposited, where the code lives — is what the editor checks.
Archive skeleton that survives a JCF revision
archive/
README.md # inputs, licenses, runtime, expected outputs
run_all.do|.R|.py # single entry point; table/figure numbering matched
build/ # pull + clean scripts per source (compustat.do, boardex.R, sdc.py)
analysis/ # one script per exhibit (t3_main_did, f2_event_study)
handcollected/ # shareable data + collection protocol
output/ # regenerated exhibits land here
Number analysis scripts by exhibit so an R&R that renumbers tables is a rename, not an archaeology project. JCF revisions routinely add robustness exhibits; an archive organized per-exhibit absorbs them without breaking run_all.
Anti-patterns
- Promising "data available on request" without an Option C statement.
- Attempting to upload raw Compustat/CRSP/SDC extracts (license violation).
- Leaving the archive to acceptance — build it alongside the analysis.
- A README that omits how to obtain the licensed inputs.
Output
【DAS】Option C statement drafted? [Y/N]
【Archive】run_all + versions + seeds + var-construction? [Y/N each]
【Deposit】DOI for shareable data/code? [Y/N]; licensed-data note? [Y/N]