WORDS BEFORE VOLATILITY: DO TEXT-BASED RISK DISCLOSURES PREDICT FUTURE STOCK RETURN VOLATILITY?

Sammanfattning

Textual analysis of 10-K filings has long been used to examine the relationshipbetween corporate disclosures and market outcomes. Prior research shows thattextual measures of sentiment and uncertainty are associated with stock returns,volatility, and other indicators of firm risk (Li, 2008; Loughran and McDonald,2011; Campbell et al., 2014). However, traditional dictionary-based approaches relyon predefined word lists and may fail to capture contextual meaning in financialdisclosures.This study examines whether large language models (LLMs) provide a more informative measure of disclosure-based risk than traditional dictionary-based methods.Using Item 1A (Risk Factors) and Item 7 (Management’s Discussion and Analysis) disclosures from U.S. 10-K filings, we construct LLM-based volatility scoresand compare their performance with dictionary-based uncertainty measures. Theanalysis combines pooled OLS regressions, firm and time fixed effects models, andout-of-sample forecasting tests.The results show that LLM-based measures exhibit stronger explanatory powerthan dictionary-based measures in pooled cross-sectional regressions and are positively associated with subsequent realized stock return volatility. However, theserelationships largely disappear once firm and time fixed effects are introduced, indicating that textual measures primarily capture persistent differences across firmsrather than time-varying changes in risk. Consistent with this interpretation, neither LLM-based nor dictionary-based measures improve out-of-sample volatilityforecasts beyond standard financial control variables.The findings suggest that while LLMs provide a richer representation of disclosurebased risk, their value lies primarily in characterizing cross-sectional differences infirm risk rather than improving the prediction of future stock volatility.