從自願到強制：臺灣永續報告書政策轉變對文本品質之影響

Abstract

本研究旨在探討強制性資訊揭露對企業永續報告書品質的影響。利用臺灣金融監督管理委員會自 2015 年起，依產業類別及資本額分階段逐步擴大永續報告書強制編製範圍的獨特政策作為準自然實驗場景，本研究以 2008 年到 2023 年間臺灣上市櫃公司為樣本，運用文本分析衡量永續報告書的四項文本屬性（長度、硬資訊、明確性、黏性）。在計量方法上，本研究除了以傳統雙向固定效應模型（TWFE）進行基準分析外，更採用了堆疊式雙重差分法（Stacked DiD）及 Sun & Abraham（2021）估計策略等前沿方法，以克服 TWFE 在交錯採納設計下的潛在偏誤。本研究發現，強制性編製規定對於永續報告書的長度、硬資訊比例與明確性，幾乎未產生統計上顯著的影響。最穩健且一致的發現是，強制性規定顯著且持續地降低了報告的內容黏性。這些結果表明政策主要驅動了報告的形式更新，而非品質躍升。

This study investigates the impact of mandatory disclosure on sustainability reporting quality and leverages a unique quasi-natural experiment in Taiwan. The setting is provided by the policy of Taiwan's Financial Supervisory Commission (FSC), which has expanded the scope of mandatory sustainability reporting in a phased, staggered manner based on industry and firm capital size since 2015. Using a sample of all Taiwanese listed companies from 2008 to 2023, the research employs textual analysis to measure four textual attributes: length, hard information, specificity, and stickiness. Methodologically, in addition to using the traditional Two-Way Fixed Effects (TWFE) model for baseline analysis, this study adopts cutting-edge methods such as Stacked Difference-in-Differences (DiD) and the Sun & Abraham (2021) estimator to overcome the potential biases of TWFE under a staggered adoption design. The finding shows that the mandate had nearly no statistically significant effect on report length, hard information, or specificity. The most robust and consistent finding is that the mandatory requirement significantly and persistently decreases report stickiness. These results indicate that the policy primarily drove "formalistic updates" rather than a "substantive quality leap."