Understanding Corporate Sustainability Disclosures from the Securities Exchange Commission Filings

Abstract

As sustainability becomes fundamental to companies, voluntary and mandatory disclosures or corporate sustainability practices have become a key source of information for various stakeholders, including regulatory bodies, environmental watchdogs, nonprofits and NGOs, investors, shareholders, and the public at large. Understanding sustainability practices by analyzing a large volume of disclosures poses major challenges, given that the information is mostly in the form of text. Applying machine learning and text analytic methods, we analyzed approximately 25,428 disclosure reports for the period of 2011 to 2020, extracted from the Securities and Exchange Commission (SEC) filings and made available at the Ceres website via application programming interfaces (APIs). Our study identified six industry clusters from the K-means and six main topics from the latent Dirichlet allocation (LDA) method that related to the disclosure of climate-change-related environmental concerns. Both methods produced overlapping results that further reinforce and enhance our understanding of climate-change-related disclosure at various levels, such as sector, industry, and topic. Our analysis shows that companies are concerned primarily with the topics of gas emission, carbon risk, climate change, loss and damage, renewable energy, and financial impact when disclosing climate-change-related issues to the government. The study has implications for corporate sustainability practices, the communication and dissemination of such practices to stakeholders at large and furthering our understanding of sustainability in general.