企業の気候目標を強制検証排出量に対して監査する：公開データパイプラインと推定値の仕様感応性（再現パッケージ）

Unofficial AI-generated summary based on the public title and abstract. Not an official translation.

🔬研究者:Provides a reproducible pipeline and data for auditing climate targets, with insights on specification sensitivity and clustering choices.

🏢実務担当者:Offers a method to validate climate claims against verified emissions, useful for internal audit and disclosure credibility.

🏛政策担当者:Demonstrates how mandatory registries can be leveraged for climate-target verification, informing greenwashing regulation and disclosure standards.

Replication package for the manuscript “Auditing corporate climate targets against mandatory verified emissions: a public-data pipeline and the specification sensitivity of the resulting estimates.” The study audits corporate climate-target claims against third-party-verified emissions held in two mandatory registries — the European Union Transaction Log and the US EPA Greenhouse Gas Reporting Program — and documents how few published claims can in fact be adjudicated against registry evidence, and how sensitive the resulting estimates are to the level at which standard errors are clustered. Every input is public. No licensed, proprietary or restricted data were used at any stage. No personal data are present: all identifiers are corporate, facility or regulatory. All sources were retrieved in the window ending 29 July 2026. Contents data/ — 20 files, 37 MB. Raw registry and filing extracts exactly as retrieved, plus every derived analysis file, including the balanced installation-year panel (63,162 observations, 5,742 installations, 2013–2023) on which every estimate in the paper rests, the parsed claim corpus of 1,779 assertions with adjudication-tier assignments, and the entity-resolution tables with their similarity scores. code/ — acquisition scripts for the EU Transaction Log, EPA GHGRP, EPA ECHO and SEC EDGAR; the deterministic rule-based claim parser; and the panel-construction scripts. figures/scripts/ — the reporting pipeline, which is the authoritative implementation. Running it end to end regenerates the results file, every table and every generated figure from the deposited data. figures/results/results.json — the single source of truth. Every number in the manuscript text, tables and figures is read from this file; nothing is transcribed by hand. DATA_DICTIONARY.md — variable definitions, coding schemes and transformations for every file and column, including the six-tier adjudication coding scheme and the entity-resolution procedure. MANIFEST.md — every file with its SHA-256 checksum, row count, column count and provenance. Reproduction check The reporting pipeline was executed end to end from a clean extraction of this archive. gen_results.py completed in approximately 40 seconds and regenerated results.json ; all 960 leaf values in the regenerated file are identical to the deposited copy . The check was deliberately run on library versions newer than and different from those pinned in requirements.txt (pandas 2.3.3 rather than 3.0.2, numpy 2.2.6 rather than 2.4.4, scipy 1.15.3 rather than 1.17.1), so the reported results are not dependent on a specific dependency resolution. Random seed 20260729. The only stochastic components are company-level cluster bootstraps: 2,000 draws for the group-time aggregate, 20,000 for the domicile comparison. Reproduce pip install -r requirements.txt python figures/scripts/gen_results.py # -> figures/results/results.json python figures/scripts/gen_tables.py # -> tables/tab_*.tex python figures/scripts/plot_figs.py # -> figures/fig_*.pdf All three scripts resolve paths relative to the archive root, so these commands work from a fresh extraction with no arguments. Primary sources Science Based Targets initiative dashboard export (39,379 target records, 15,264 companies) European Union Transaction Log, retrieved through the public API maintained by the euets.info project, which mirrors the register US EPA Greenhouse Gas Reporting Program (Envirofacts) US EPA Enforcement and Compliance History Online US SEC EDGAR, Forms 10-K and 20-F The SBTi export and the EPA enforcement database are refreshed on a rolling basis and carry no version identifier, so the retrieved copies are archived here to fix the analysed vintage. Licence Data and documentation are released under CC BY 4.0. The code in code/ and figures/scripts/ is additionally released under the MIT licence included as LICENSE in the archive.

• Zenodo https://zenodo.org/records/22040987first seen 2026-08-22 04:12:22 · last seen 2026-09-04 04:36:37