Text, Tone, and Legal Language: Analyzing Mutual Fund Disclosure Sentiment

Abstract

Financial disclosures provide more information than simply their regulated content. Disclosure tone and textual attributes predict firm performance, earnings persistence, stock volatility, capital costs, and retail investor behavior. But do these theories apply to mutual fund disclosures? Following Loughran & McDonald, we develop customized dictionaries specific to mutual funds. We then introduce a novel sentiment scoring framework that generates a transparent sentence- and disclosure-level score for our sample of 164,602 mutual fund summary prospectuses (497k) from 2010 to 2020. Descriptive analysis validates our dictionary by showing meaningful and statistically significant differences across disclosure sections, CRSP categories, and time. Principal risk (PR) sections are more negative (and uniformly so) than Investment Strategy (IS) sections across time. Our model results further validate the sentiment scoring framework and illustrate content differences between the Principal Risk and Investment Strategy, while highlighting the role of legal language in setting the disclosure tone. Our context-sensitive approach provides researchers and regulators with a tool to better assess not only what fund disclosures are saying, but how they say it. These persistent differences in PR and IS estimated information signals map onto content regulations and suggest a rich field for future research.

Keywords: mutual funds, institutional investors, disclosure, textual analysis, text mining, sentiment

JEL Classification: G1, G14, G20, K22