Replication package for “Full-Retirement Heterogeneity in the Voluntary Carbon Market: Project Types, Registries, and Composition”
Description
Licence scope: MIT applies only to original project code; CC BY 4.0 applies only to original project documentation, aggregate analytical outputs, tables, and figures. No licence is asserted over third-party material, which is not included; see LICENSES.md. This replication package supports the article “Full-Retirement Heterogeneity in the Voluntary Carbon Market: Project Types, Registries, and Composition.” It contains original analytical code, canonical non-row-level outputs, final tables and figures, deterministic exporters for the validated robustness synthesis (Table B1) and corrected Y3 table, input integrity hashes, and documentation of provenance and transformations. The analysis uses a fixed CarbonPlan OffsetsDB snapshot dated 1 June 2026. The canonical primary sample contains 27,249 project-by-vintage observations representing 6,025 projects; Y2 full retirement is the primary outcome. The corrected Y3 sample contains 10,439 observations and 3,743 projects, with 300/300 valid project-cluster bootstrap results. CarbonPlan/registry raw data, row-level derived panels, and diagnostics containing individual project or row identifiers are not redistributed because source-registry rights have not been affirmatively cleared. The package preserves the article's identification limits: results are associational, the timing profile does not identify a causal age effect, and registry/project-type nesting constrains interpretation.
Files
Steps to reproduce
See README.md in the uploaded archive. For deterministic verification without estimation, run: python 04_CODE/CORRECTED_PREFERRED_20260826/export_canonical_submission_tables.py --check-only. A full rerun requires lawfully obtained inputs matching the SHA-256 values in INPUT_FILES_NOT_REDISTRIBUTED.csv and follows run_all.py.