Can GenAI fill banks’ emissions data gaps?
GenAIは銀行の排出データギャップを埋められるか？ (AI 翻訳)
Cristina Angelico, Enrico Bernardini
🤖 gxceed AI 要約
日本語
本稿は、ユーロ圏主要銀行の排出データに大きな限界があることを示す。データプロバイダー4社のデータを分析した結果、スコープ3排出量にデータギャップ、プロバイダー間の不一致、高いボラティリティが確認された。次に、3つのGenAIツールを用いて、GenAIがデータギャップを埋められるか検証。GenAIベースの排出データは従来データと相関し、部分的にギャップを埋める可能性があるが、品質や一貫性に関して同様の問題を抱え、再現性や透明性にも課題がある。
English
This paper finds significant limitations in emissions data for euro area banks from four providers: gaps, inconsistencies, and high volatility in scope 3 emissions. It tests three GenAI tools and finds that GenAI-based data correlates with traditional sources and can partially fill gaps, but suffers similar quality issues and raises concerns about replicability and transparency.
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
日本ではSSBJ基準や有報での気候関連開示が進む中、金融機関の投融資先からのスコープ3排出データの質が課題となっている。本稿の知見は、GenAI活用の可能性と限界を示唆し、日本の開示実務やデータ品質向上策に示唆を与える。
In the global GX context
As ISSB, CSRD, and SEC climate rules push for scope 3 disclosure, data gaps remain a critical bottleneck. This paper provides novel evidence on the limitations of both traditional and GenAI-based emissions data, underscoring the need for standardized measurement and regulatory progress to improve underlying information.
👥 読者別の含意
🔬研究者:Provides empirical evidence on data quality issues in bank emissions data and a systematic evaluation of GenAI as a complementary data source.
🏢実務担当者:Banks and data providers can use findings to assess current data limitations and explore GenAI tools for gap-filling, while being aware of quality and transparency trade-offs.
🏛政策担当者:Highlights the urgent need for clear measurement standards and regulatory developments to improve climate disclosure by non-financial corporations in bank portfolios.
📄 Abstract（原文）
This paper presents new evidence highlighting significant limitations in emissions data for major listed banks in the euro area, sourced from four leading providers. Notable issues are data gaps, inconsistencies across providers and high volatility over time in scope 3 emissions, which are often inexplicably lower than scope 2 emissions and do not correlate with the banks’ exposure to high-emitting sectors. The paper then examines whether Generative Artificial Intelligence (GenAI), which relies on broader and diverse information sets, can help bridge the existing data gaps by comparing the outcomes from three GenAI tools. We find that GenAI-based emissions data, either estimated or retrieved, are correlated with data from traditional sources and may therefore help to partially fill current gaps and identify anomalies in the available data. Nevertheless, they suffer from similar issues in terms of quality and consistency to those documented for the data supplied by professional providers. Additional concerns regard replicability and transparency, underscoring the need for initial caution in their use. Despite the current limitations, looking forward, GenAI may become a valuable complementary data source as models that are fine-tuned for this task are developed. Future improvements will also depend on the availability of more reliable underlying information, which in turn requires parallel progress in defining simple, clear and actionable measurement standards as well as regulatory developments to promote climate disclosure by non-financial corporations within banks’ portfolios.
🔗 Provenance — このレコードを発見したソース
- semanticscholar https://doi.org/10.2139/ssrn.6666661first seen 2026-05-05 22:05:58
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。