← 論文一覧に戻る

「報告排出量を超えて:企業の隠れたカーボンフットプリントを明らかにする」再現パッケージ

Replication package for "Beyond Reported Emissions: Uncovering Hidden Corporate Carbon Footprints" (原題)

Pothin, Florian, Rondeau, Fabien, Razafindrabe, Tovonony, Macherey, Arthur

Zenodoプレプリント2026-10-29#AI×ESGOrigin: EU経営インパクト: 調達リスク対象セクター: cross_sector
DOI: 10.5281/zenodo.23757920
原典: https://zenodo.org/records/23757920

🤖 gxceed AI 要約

日本語

本パッケージは、CDP開示における企業のScope 3報告を15カテゴリ別に補正する研究の再現コードと集計結果を提供する。各報告を活動除外・境界不完全・適正開示に分類し、勾配ブースティングモデルで欠損規模を推計、報告値にアンカーした上限の下で補正フットプリントを再構成する。対象は2013〜2023年の13,675社・40,037社年。コードはMIT、集計結果はCC BY 4.0で公開され、原データは非再配布。

English

This replication package accompanies a study that corrects corporate Scope 3 disclosures category by category across the 15 GHG Protocol categories. Each firm report is classified as activity exclusion, boundary incompleteness, or sound disclosure; missing magnitudes are estimated with gradient-boosted models trained on sound disclosures and recomposed under a cap anchored on reported footprints. The sample spans 13,675 firms and 40,037 firm-years of CDP data (2013-2023). Code is MIT-licensed and derived aggregates are CC BY 4.0; upstream data are not redistributed.

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

📝 gxceed 編集解説 — Why this matters

日本のGX文脈において

SSBJ・有報でのScope 3開示義務化が進む日本企業にとって、報告値の過小性をカテゴリ別に可視化・補正する手法は、投資家向け開示の信頼性検証やサプライチェーン排出把握の実務に直結する。再現可能な形で公開された点は国内の開示インフラ整備にも示唆が大きい。

In the global GX context

As ISSB/SSBJ and CSRD push Scope 3 into mandatory disclosure, this work offers a transparent, replicable method to detect and correct under-reporting in CDP data — directly relevant to assurance, greenwashing detection, and comparability debates in global disclosure scholarship.

👥 読者別の含意

🔬研究者:Scope 3開示の測定誤差をMLで補正する再現可能な手法と、CDP大規模パネルの構築手順を学べる。

🏢実務担当者:自社Scope 3報告のカテゴリ別欠落リスクを点検し、開示の頑健性を高める際の参照枠になる。

🏛政策担当者:Scope 3開示の過小報告をどう検証・規律するか、開示制度設計の論拠として活用できる。

📄 Abstract(原文)

Replication code and derived aggregate results for Pothin, F., Rondeau, F., Razafindrabe, T. and Macherey, A., "Beyond Reported Emissions: Uncovering Hidden Corporate Carbon Footprints", International Journal of Finance & Economics, forthcoming. The paper builds a category-level correction of corporate Scope 3 disclosures. For each of the 15 GHG Protocol categories it classifies a firm's report as an activity exclusion, a boundary incompleteness or a sound disclosure, estimates the missing magnitude with gradient-boosted models trained on sound disclosures, and recomposes a corrected footprint under a cap anchored on the reported footprint. The sample covers 13,675 firms and 40,037 firm-years of CDP disclosures, 2013-2023. The package contains (i) the Python code that builds the firm-year panel from the raw CDP files, runs the classification, the estimation engine, the correction and the bootstrap inference, and produces every table and figure of the paper and its Supplementary Appendices; (ii) the derived aggregates behind the reported results (sector, region, category, year and sample statistics, model performance, regression coefficients and 90% bootstrap intervals), with no firm-level rows and no firm identifiers; (iii) a README giving the order of execution, the mapping from scripts to tables and figures, the checks run when the package was assembled, and the known limits to reproducibility; and (iv) WITHHELD.md, which lists every licensed or proprietary input (CDP, Bloomberg, CoGEM/Pladifes, ADEME, a private revenue snapshot) and how to obtain it. Code: MIT licence. Derived aggregates: CC BY 4.0. Upstream data are not redistributed and remain under their providers' terms. Funding: Florian Pothin's doctoral research was supported by a CIFRE agreement between Toovalu and Université de Rennes, funded by ANRT (grant 2022/1627).

🔗 Provenance — このレコードを発見したソース

🔔 こうした論文の新着を逃したくない方は キーワードアラート に登録(無料・3キーワードまで)。

gxceed は公開メタデータに基づく研究支援データセットです。要約・翻訳・解説は AI 支援で生成されています。 最終的な解釈・検証は利用者が原典資料に基づいて行うことを前提とします。