査読報告：深層機械学習を用いたバングラデシュ金融機関におけるマルチ資本開示とサステナビリティ・パフォーマンス
Peer Review Report For: Multi capital disclosure and sustainability performance in Bangladeshi financial institutions using Deep Machine Learning [version 1; peer review: 1 approved, 1 approved with reservations] (原題)
(著者不明)
🤖 gxceed AI 要約
日本語
バングラデシュ上場金融機関62社・1,178社年を対象に、知的資本・人的資本・自然資本の各開示がサステナビリティ・パフォーマンスに与える影響を検証。パネル回帰とシステムGMMに加え多層パーセプトロン深層学習を補完的に用い、3種開示すべてが1%水準で正に有意、知的資本開示が最大の予測因子であることを示した。開示の質がグリーン融資量よりもサステナビリティ成果を規定することを示唆する。
English
Using 1,178 firm-year observations from 62 listed Bangladeshi financial institutions (2005-2023), the study tests how intellectual, human, and natural capital disclosure affect sustainability performance. Panel regressions, system-GMM, and a supplementary multi-layer perceptron deep-learning model all show positive, significant effects (1% level), with intellectual capital disclosure the leading predictor. Disclosure quality outweighs green-credit volume, supporting a substantive rather than nominal disclosure act.
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
新興国・気候脆弱国における統合報告とマルチ資本開示の実証は、SSBJ基準や統合報告書での資本区分開示を進める日本企業にとって、開示の「実質性」を測る比較材料となる。AIによる開示評価手法は日本での開示分析高度化の参考にもなる。
In the global GX context
Adds emerging-market empirical evidence to the integrated-reporting and multi-capital disclosure discourse relevant to ISSB/ISSB-aligned frameworks, showing disclosure quality can matter more than green-finance volume. The ML-based validation offers a template for assessing substantive versus nominal disclosure under CSRD/ISSB regimes.
👥 読者別の含意
🔬研究者:マルチ資本開示とサステナビリティ成果の因果的検証に、GMMと深層学習を組み合わせた手法設計の参考になる。
🏢実務担当者:統合報告での資本別開示の質向上が、グリーン融資量の拡大よりも成果に効く可能性を示す。
🏛政策担当者:新興国での開示規制設計において、開示の実質性を促す枠組みの重要性を示唆する。
📄 Abstract（原文）
Purpose This study empirically investigates the joint and independent effects of intellectual capital disclosure (ICD), human capital disclosure (HCD), and natural capital disclosure (NCD) on the sustainability performance (SUSP) of listed financial institutions in Bangladesh, a bank-centric, climate-vulnerable emerging economy. Design methodology approach Drawing from a composite theoretical framework based on the Resource-Based View, Signalling Theory and the Legitimacy/Stakeholder approaches, the study engages a balanced sample of 62 listed financial institutions, including 30 commercial banks, 22 non-bank financial institutions and 10 insurance companies, providing 1,178 firm-year observations over 2005-2023. Analysis is conducted through panel (firm- and year-fixed effects) regressions, dynamic system generalized method of moments (system-GMM) and a supplementary multi-layer perceptron deep learning model. Findings We find that ICD, HCD, and NCD have positive and substantive effects on SUSP (at the 1% significance level), with the impact of intellectual capital disclosure as the leading predictor, followed by human and natural capital disclosures. Relationships remain robust under winsorisation, lagged regressors, sub-sample data, pandemic-free data, and a non-linear deep-learning framework (out-of-sample R 2 = 0.932), thereby validating a substantive (but not nominal) disclosure act. Board independence is consistently positive, while green funding intensity demonstrates directionally positive but statistically subordinate effects, indicating that disclosure quality supersedes green-credit volume in shaping sustainability outcomes. Originality value The results contribute to the integrated reporting discourse by confirming the triadic disclosure construct and provide recommendations for policy-makers, banks and climate finance practitioners in institutionally constrained emerging-market settings.
🔗 Provenance — このレコードを発見したソース
- openalex https://doi.org/10.5256/f1000research.202919.r529870first seen 2026-11-05 05:24:01
🔔 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。