jue-replication-package

Replication Package (jue-replication-package)

When to trigger

• The paper contains empirical, simulation, or experimental work — JUE's mandatory replication policy applies
• The data is geocoded, restricted, or proprietary (addresses, parcels, admin records) and you need a confidentiality path
• You are at acceptance and must upload the replication folder before publication
• A spatial pipeline (geocoding, GIS joins, distance/ring construction) is undocumented and not reproducible
• A coauthor assumes the deposit can wait until after publication — it cannot

JUE's mandatory replication policy (检索于 2026-06；以官网为准)

JUE will publish a paper only if the data and code are clearly documented and readily available for replication. At acceptance, before publication, authors of papers with empirical work, simulations, or experiments must upload a replication folder to a major data repository. The minimum deposit is:

1. The dataset(s) used to generate all estimates reported in the paper;
2. A description of how intermediate datasets and programs were used to build the final dataset(s);
3. The complete set of code from raw data to final results — except where code would reveal confidential information about the data.

Proprietary / confidential data: authors may request an exemption from providing data and/or code. The request must be made in the submission cover letter to the editor. Even with an exemption, document the pipeline and provide whatever code does not reveal confidential information so the path to results is auditable.

Spatial-data deposits are harder than they look

Urban papers carry data hazards that generic replication checklists miss:

• Geocoding is not reproducible by default. Pin the geocoder, version, and date; deposit the crosswalk from addresses to coordinates (or a de-identified version) so distance/boundary variables can be regenerated.
• Restricted address/parcel data. Provide a synthetic or aggregated extract plus exact instructions to obtain the restricted file (agency, application, access tier), and deposit all code that runs against it.
• GIS layers and shapefiles. Deposit or cite the exact boundary/shapefile vintages (tracts, school zones, transit lines); boundaries change across years and silently break replication.
• Spatial construction code. Ring/donut buffers, distance-to-boundary, market-access, and spatial-weight matrices must be in the code, with the distance cutoffs and projection (CRS) stated.

Building the folder

1. One-click master script (run_all) raw → cleaned → estimates → exhibits, with a stated software/version environment.
2. README mapping each table/figure to the script that produces it; list data sources with access tiers and the GIS-layer vintages.
3. Separate raw, intermediate, and final data so reviewers can re-run from any stage.
4. Confidential path: if exempt, the README states what is withheld, why, and how a researcher with access can reproduce it; the cover-letter exemption is filed.
5. Verify on a clean machine before deposit — the most common failure is hard-coded local paths and unstated CRS/geocoder versions.

Restricted-data access paths common in urban work

Urban papers lean on data that cannot simply be posted; document the path, not just the absence:

• Census/admin microdata in an RDC/FSRDC enclave — state the dataset, the project-approval route, and deposit all code that runs inside the enclave plus disclosure-cleared output; a researcher with clearance must be able to reproduce.
• Proprietary real-estate / transaction data (CoreLogic, Zillow ZTRAX-type, MLS) — cite the vendor and vintage, deposit construction code, and provide a synthetic or aggregated extract where the license permits.
• Geocoded individual records — deposit the de-identified crosswalk or the construction code that maps protected addresses to the spatial variables (distance, ring, boundary side), withholding only the raw identifiers.
• Mobility / GPS / cell-phone data — document the provider, the spatial and temporal resolution, and the aggregation that produces the analysis file.

In every case the cover-letter exemption names what is withheld and why, and the README states exactly how an authorized researcher reproduces the results.

Checklist

• [ ] Replication folder targets a major repository; deposit planned for acceptance (before publication)
• [ ] All datasets behind reported estimates included (or exemption requested in the cover letter)
• [ ] Intermediate-data construction described; raw → final pipeline complete
• [ ] Geocoder/version/date pinned; address↔coordinate crosswalk (or de-identified) deposited
• [ ] GIS layer/shapefile vintages cited; CRS/projection stated
• [ ] Spatial construction code (buffers, distance, weights, market access) included with cutoffs
• [ ] run_all master script verified on a clean machine; README maps exhibits to scripts
• [ ] Confidential-data exemption documented if applicable
• [ ] Restricted-access path (RDC/vendor/enclave) named with the route to reproduce
• [ ] Folder assembled during the R&R as a correctness check, not left to acceptance

Anti-patterns

• Assuming the deposit waits until after publication (JUE requires it at acceptance, before publication)
• Geocoding with an unpinned service, so distance/boundary variables cannot be regenerated
• Depositing code but not the GIS-layer vintages, so boundaries silently differ across years
• Hard-coded local paths and an unstated CRS that break the spatial pipeline on another machine
• Claiming proprietary data without filing the cover-letter exemption or documenting the access path
• A README that lists files but never maps each table/figure to the code that makes it

Why early assembly pays off

Although JUE requires the deposit only at acceptance, assembling the folder during the R&R (or before submission) protects you twice. First, the act of building a one-click run_all from raw geocoded data to final maps routinely surfaces silent errors — a wrong CRS, a stale shapefile vintage, a dropped buffer — that would otherwise survive into the published result. Second, an acceptance deadline is the worst time to discover a restricted-data path cannot be reproduced or that a vendor license forbids posting an extract; the exemption and access instructions take time to negotiate. Treat the replication folder as a correctness check on the spatial pipeline, not a post-acceptance formality.

Output format

【Policy trigger】empirical/simulation/experimental? deposit at acceptance planned? [Y/N]
【Datasets】all reported-estimate data included / exemption requested in cover letter
【Pipeline】raw→intermediate→final documented; run_all verified clean? [Y/N]
【Geocoding】geocoder+version pinned; crosswalk deposited? [Y/N]
【GIS layers】shapefile vintages cited; CRS stated? [Y/N]
【Spatial code】buffers/distance/weights/market-access included with cutoffs? [Y/N]
【Confidential path】exemption + access instructions documented? [Y/N/NA]
【Next skill】jue-referee-strategy