Does Sentiment Measured Through Language Models Encompass a Broader Expanse of Information From the Options Market?

ABSTRACT Language models have emerged as powerful tools for measuring textual sentiment. Do sentiment indices derived from language models and textual data truly capture richer informational content? Drawing on a theoretical model, this paper illustrates how market sentiment embedded in news‐texts shapes investors' risk aversion and influences their inclination to speculation. This dynamic drives more frequent trading activities, ultimately exerting an impact on the options market. Using news data from The Wall Street Journal as the corpus, we employ language models to construct a market sentiment index. Our findings validate the predictions of the theoretical model: market sentiment exerts a significant direct effect on option prices and indirectly influences them through risk aversion as a mediating variable. Furthermore, empirical evidence reveals that uncertainty significantly moderates both the direct and mediated channels linking market sentiment to option prices. Capturing market sentiment precisely, language models are indispensable for asset pricing sentiment analysis.