← 論文一覧に戻る

資産価格評価における炭素リスク：従来型・機械学習・説明可能AIのアプローチ

VARLIK FİYATLANDIRMASINDA KARBON RİSKİ: GELENEKSEL, MAKİNE ÖĞRENMESİ VE AÇIKLANABİLİR YAPAY ZEKÂ YAKLAŞIMLARI (原題)

San Arca

OpenMETU (Middle East Technical University)ジャーナル2026-10-05#AI×ESG経営インパクト: 資金調達対象セクター: finance
原典: https://hdl.handle.net/11511/121050

🤖 gxceed AI 要約

日本語

本研究はFama-French5ファクターモデルに炭素強度を表すBMGファクターを加え、米国株のリターンと炭素リスクの関係を検証した。時系列回帰・GRS検定・Fama-MacBeth検定に加え、XGBoostとSHAPを用いてBMGファクターの寄与を解釈する。FF5+モデルは5つのポートフォリオ分類のうち4つでGRS p値を改善し、炭素リスクは特に鉱業・石油ガス採掘など炭素関連セクターで追加的な説明力を持つことが示された。

English

This study augments the Fama-French five-factor model with a Brown-minus-Green (BMG) carbon intensity factor to examine carbon risk and US stock returns. Beyond time-series regression, GRS and Fama-MacBeth tests, it applies XGBoost and SHAP to interpret the BMG factor's predictive contribution. The FF5+ model improves GRS p-values in four of five portfolio sorts, with carbon risk adding explanatory power especially in carbon-intensive sectors like mining and oil and gas extraction.

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基準や有報での気候関連開示が進む日本では、炭素リスクを資産価格に織り込む実証手法は投資家対応やポートフォリオ構築の参考になる。特にSHAPによる要因解釈は、開示情報を投資判断に接続する際の説明可能性の枠組みとして示唆的である。

In the global GX context

As ISSB/TCFD-aligned disclosure becomes mandatory, this work offers empirical evidence that carbon intensity carries priced risk, supporting climate-risk integration into asset pricing and transition-finance valuation. The explainable-AI angle connects disclosure data to investor decision-making, relevant to CSRD and SEC climate disclosure debates.

👥 読者別の含意

🔬研究者:炭素ファクターの追加的説明力と機械学習・SHAPによる解釈手法を、資産価格研究の実証設計として参照できる。

🏢実務担当者:炭素集約度が株価リターンに織り込まれる可能性を示し、ポートフォリオの炭素リスク管理や投資家説明に活用できる。

🏛政策担当者:気候関連開示が資産価格に反映される経路を示し、開示制度設計の政策的根拠として参考になる。

📄 Abstract（原文）

The main purpose of this study is to examine the relationship between carbon risk and the returns of US stocks, using both a conventional asset pricing model and complementary machine-learning-based models. The study is grounded in the Fama-French five-factor (FF5) asset pricing model, to which a Brown-minus-Green (BMG) factor is added to represent carbon intensity. The empirical design of this study is founded upon the construction of portfolios and factors using financial firm-level financial and environmental data. Firms are allocated to portfolios on the basis of size, book-to-market ratio, operating profitability, investment, and carbon intensity. The combined model, comprising the Fama-French five-factor asset pricing model together with the BMG factor, is then evaluated through time-series regression, the Gibbons-Ross-Shanken (GRS) common alpha test, and the Fama-MacBeth cross-sectional test. In addition to these conventional econometric analyses, this study uses Extreme Gradient Boosting (XGBoost) and SHapley Additive exPlanations (SHAP). SHAP values are used to show the manner in which the BMG factor contributes to the model’s predictions. Simplified results show that the FF5+ model yields a higher GRS p-value than FF5 in four out of five of the portfolio rankings; the most significant increase in R-squared is 0.028 in the Size-Carbon Intensity 2×3 portfolios. The findings indicate that conventional risk factors continue to play a fundamental role in explaining returns but carbon-related risk may contain additional information, especially within specific sectoral and portfolio breakdowns. Also, the sectoral findings suggest that additional explanatory power of the BMG factor is more apparent in carbon related sectors such as Mining, Quarry and Oil and Gas Extraction. In summary, the findings indicate that carbon risk becomes increasingly apparent in asset pricing, but its impact remains contingent upon sector, portfolio structure, and period.

🔗 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。