Replication package for "The paradox of openness in reverse: What sponsors lose when they relicense open-source software"

Replication package for the manuscript "The paradox of openness in reverse: What sponsors lose when they relicense open-source software." The package contains the analysis code, the event registry with verified switch dates, the affiliation evidence log, and a verification harness that re-derives every numeric claim in the manuscript and its three supplementary appendices from the stored artefacts and checks each against the text in a single run (365 claims, all passing). Contents (295 files):- reproduce.py and a 1.2 MB anonymised reproduction core (contributor-by-month panel with surrogate identifiers, the three classification variants, and the event registry), which regenerates the headline dose-response, the embodied-knowledge outcome and the symmetric-control specification in under a minute with no network access;- the full analysis code and clone scripts, which rebuild the 1,002,645-commit corpus from the public git histories of 45 repositories;- verify_paper.py, the verification harness;- the evidence log for 496 verified contributor-employer classifications, each row carrying the public source consulted. Data provenance and ethics: all inputs are public. Commit metadata is published with each repository under its own licence and is content-addressed, so every record can be confirmed by its hash against the source repository. Affiliation verdicts rest on public professional records consulted for one fact each, the contributor's employer at the event date; no individual was contacted and no non-public data were collected. The reproduction core carries surrogate identifiers and no names, addresses or commit messages. Quick start: unzip, then run "python3 reproduce.py" (Python 3.11, pandas, numpy). See README.md for the full audit route.