jle-replication-package

Replication Package & JLE Data Policy (jle-replication-package)

When to trigger

• The paper is empirical (or has simulations/experiments) and you are heading toward acceptance, or want to build the package early
• An R&R or conditional acceptance asks you to prepare the data + code deposit
• You need to write the README and a data-availability statement to JLE standard
• Some data are restricted (sealed court records, proprietary firm data) and you must plan the deposit around that

Why this matters at JLE

JLE's stated policy is to publish empirical/simulation/experimental papers only if the data are clearly and precisely documented and readily available to any researcher for replication, and authors of accepted papers must provide the data, programs, and other computation details sufficient to permit replication, prior to publication (检索于 2026-06；以官网为准; verify on journals.uchicago.edu/journals/jle/data-policy). This is a pre-publication requirement, not an acceptance-day formality. JLE does not run the AEA-style openICPSR pipeline; the obligation is documentation and availability — so build a package any researcher could re-run, and treat it as a gate, not a chore. Law-and-economics data raise distinctive access issues (court records, sealed cases, regulatory filings, proprietary market data) that you must plan for explicitly.

What the package must contain

Component	Requirement
Data files	All data used to produce the results, documented; or, for restricted data, a precise access path
Analysis + transformation code	Every script from raw legal/regulatory data → cleaned data → each table/figure
Master script	One run_all that regenerates every exhibit from the inputs
README	Data sources and provenance, computational requirements, run instructions, and an exhibit-to-code map
Data-availability statement	Provenance and access terms for each dataset (court system, agency, vendor), stated in the paper
Legal-data documentation	How statutes/cases were coded, the coding protocol, and inter-coder checks for hand-coded doctrine

Handling restricted or proprietary legal data

• Sealed court records, individual case files, and proprietary market data often cannot be redistributed. Deposit all code regardless, and document the exact access procedure (court/agency/vendor, application steps, cost, approximate wait, any data-use agreement).
• Provide a synthetic or public extract so the code runs and a verifier can check the logic where the real data cannot be shared.
• For hand-coded legal variables (a damages-cap index, a liability-standard classification), deposit the coding protocol, the source documents list, and inter-coder reliability so the measurement is reproducible.
• Flag any restricted-data or exemption situation early in the process, not at the pre-publication check.

Reproducibility hygiene (build as you go)

• Pin versions: Stata version + recorded ssc/net package versions; requirements.txt/conda env (Python); renv.lock (R).
• Set and report seeds for every simulation, bootstrap, and randomization/permutation step.
• No absolute paths — one root macro/variable; relative paths thereafter.
• Exhibit-to-code map in the README: Table 2 → code/05_did.do, Figure 1 → code/06_event_study.R, etc.
• Run it clean on a fresh checkout before depositing.

Adapt the vendored skeleton in ../../resources/code/ (master script → clean → descriptive → DiD/IV/RD → mechanism → robustness → tables) as the package backbone.

Checklist

• One run_all master script regenerates every table and figure from the inputs
• All data documented and available (or restricted-data access path fully documented + synthetic extract provided)
• README with complete exhibit-to-code map and computational requirements
• Data-availability statement for every dataset, with provenance and access terms
• Hand-coded legal variables: coding protocol + sources + inter-coder reliability deposited
• Software/package versions pinned; seeds set and reported; no absolute paths; clean fresh run
• Restricted-data / sealed-records situations flagged early, not at the pre-publication check

Anti-patterns

• Treating the deposit as an acceptance-day task — the policy requires materials available before publication
• Depositing code with absolute paths or unrecorded package versions (will not reproduce)
• Hand-coded legal variables with no coding protocol or reliability check (measurement not reproducible)
• Restricted court/market data discovered at the check with no access documentation or synthetic extract
• Unset seeds making bootstrap/permutation results non-reproducible

Common legal-data provenance situations

Law-and-economics data come from a handful of recurring sources, each with its own access and documentation pattern. Name yours and document accordingly:

Source	Access reality	What to document
Court records (PACER, state dockets)	often public but fee-gated or rate-limited; some sealed	the query/scrape procedure, date pulled, sealed-case handling
Administrative / regulatory filings (SEC EDGAR, agency dockets)	usually public	the form types, vintage, and any parsing code
Enforcement / litigation databases (vendor)	proprietary, license-restricted	the vendor, license terms, and a synthetic extract
Statute / case coding (hand-built)	you create it	the coding manual, sources, and inter-coder reliability
Linked administrative microdata (sealed)	DUA-restricted, non-redistributable	the application path, DUA terms, wait time, synthetic schema

The reproducibility obligation is the same in every row: someone who legitimately obtains the source must be able to re-run your code and recover every exhibit.

Worked vignette (illustrative)

A paper on judge assignment uses individual case records the court provides only under a data-use agreement. The author cannot redistribute them, so the package deposits: (i) all cleaning and analysis code; (ii) a documented access path (the court's data-request form, the DUA terms, the ~10-week wait); (iii) a synthetic case file with the same schema so run_all executes end-to-end and a verifier can confirm the logic; and (iv) the judge-leniency leave-out construction script. The hand-coded ruling-type variable ships with its coding manual and a 200-case inter-coder reliability table. Any researcher who obtains the records can reproduce every exhibit — the JLE standard.

Output format

【Master script】run_all regenerates all exhibits from inputs? [Y/N]
【Data】documented + available, or restricted path + synthetic extract? [state]
【README】exhibit-to-code map + computational requirements complete? [Y/N]
【DAS】provenance + access terms for every dataset? [Y/N]
【Legal coding】protocol + sources + inter-coder reliability deposited? [Y/N/NA]
【Reproducibility】versions pinned + seeds + no absolute paths + clean fresh run? [Y/N]
【Restricted/sealed】flagged early? [Y/N/NA]
【Next step】jle-referee-strategy (or jle-submission)