← 論文一覧に戻る

上場前のESG開示準備シグナル:インドIPO目論見書のテキストマイニング分析

Signaling ESG disclosure readiness before listing: text-mining evidence from Indian IPO prospectuses (原題)

Vattappoyil Safas, Mohsin Khan

Frontiers in Sustainability📚 査読済 / ジャーナル2026-09-24#AI×ESG経営インパクト: 資金調達対象セクター: cross_sector
DOI: 10.3389/frsus.2026.1954577
原典: https://doi.org/10.3389/frsus.2026.1954577

🤖 gxceed AI 要約

日本語

インドのメインボードIPO目論見書310件を対象に、Sentence-BERTとESGタクソノミーを用いてESG関連記述の強度とテーマ構成を分析した。ESG文は平均11.4%を占め、ガバナンスが75.5%と圧倒的で、社会・環境開示は限定的かつ偏在していた。取締役会規模・独立性がガバナンス開示と関連し、上場時点での信頼性シグナル仮説と整合する。

English

Using Sentence-BERT and an ESG taxonomy, this study analyzes ESG language in 310 Indian mainboard IPO prospectuses. ESG sentences average 11.4% of text, with governance dominating at 75.5% while social and environmental disclosure remain limited. Board size and independence are associated with governance disclosure, consistent with a credibility-signaling channel at listing.

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ESG開示が進む中、上場時点での開示準備状況を定量化する手法は、新規上場企業のIR戦略や投資家の目論見書評価に示唆を与える。インドの事例は新興国市場における開示インフラ整備の参考となる。

In the global GX context

This study contributes to global disclosure scholarship by examining ESG communication at the IPO stage, a critical juncture for capital market credibility. It offers a replicable NLP methodology for assessing disclosure readiness, relevant to ISSB/CSRD implementation and emerging market disclosure infrastructure.

👥 読者別の含意

🔬研究者:AI×ESG開示分析の手法と、IPO文脈におけるシグナリング理論の実証例を提供する。

🏢実務担当者:上場準備企業のESG開示戦略において、ガバナンス偏重の現状と環境・社会開示の強化余地を示唆する。

🏛政策担当者:新興国市場におけるESG開示規制の設計や、上場審査での開示要件検討に参考となる。

📄 Abstract(原文)

Introduction Environmental, social, and governance (ESG) disclosure has become central to how firms establish credibility with capital markets, yet its role at the point of market entry remains underexamined. This study investigates how Indian firms communicate ESG readiness in mainboard initial public offering (IPO) prospectuses, measuring the intensity and thematic composition of ESG-related language rather than treating disclosure as a proxy for ESG performance or quality. Methods The analytical sample comprises 310 non-financial operating-company IPO prospectuses issued from FY2021-22 to FY2025-26. Prospectus text was cleaned, segmented into sentences, and embedded using Sentence-BERT. Each sentence was matched by cosine similarity to reference sentences derived from a Baier et al.-based ESG taxonomy comprising three pillars, ten categories, and forty subcategories; sentences scoring 0.45 or above were classified as ESG-related and assigned to the corresponding pillar and subcategory. In addition to the textual analysis, the study also carried out firm-level regressions relating pillar-specific disclosure to issuer characteristics and board structure, controlling for sector, year and document length. Results ESG-related sentences constitute, on average, 11.382% of total sentences across the sample. Governance disclosure is significantly more prevalent (8.605% of clean sentences) than social (1.462%) or environmental (1.314%) disclosure, with governance accounting for 75.5% of all ESG-classified sentences. Disclosure intensity and thematic emphasis vary meaningfully across financial years, industry sectors, and ESG subcategories. Supplementary regressions reveal pillar-specific rather than uniform associations between issuer characteristics and ESG disclosure, while board size and board independence are jointly associated with governance disclosure in a parsimonious specification. Discussion These findings are consistent with, but do not establish, a credibility-related signaling channel centered on governance disclosure. In the IPO prospectuses governance-related content dominates ESG communication at listing, while environmental and social disclosure remain comparatively limited and unevenly distributed across sectors and time.

🔗 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。