ESG in Corporate Filings: An AI Perspective

Using Artificial Intelligence (AI) techniques, we quantitatively examine the mentions of ESG terms in the U.S. corporate filings with the SEC over the 2019-20 period. We find that in our sample of companies, the management tends to focus ESG discussions in corporate filings along three dimensions: 1) diversity, 2) hazardous materials, and 3) greenhouse gasses. However, overall, the companies and their stakeholders approach ESG from the following three “ESG pillars”: 1) a combination of greenhouse gasses, data security and inclusion, 2) a tradeoff between emissions and product quality, and 3) product labeling. These dimensions can be effective ESG ratings aspects for companies. We further measure the investor appreciation of the corporate ESG communication through a cross-sectional analysis of the ESG discussions on the contemporaneous and subsequent stock returns. Using standard linear regression methodology, we find that the markets are much more responsive to the broad spectrum of ESG pillars, as opposed to the management “party line.” However, when we account for potential correlation of ESG communication dimensions using AI-based techniques, the statistical superiority of the full communication disappears and only the official corporate message carries through to the investors. We find that the strategic ESG messaging in corporate filings matters and significantly impacts corporations’ forward-looking returns.