Text-implied Risk and the Cross-section of Expected Stock Returns Job Market Paper

This paper establishes an econometric framework to construct a systematic risk factor from textual data, by linking a beta pricing model with a language model based on machine learning techniques. In this framework, the distributions of stock returns and words in associated documents are determined by a common underlying systematic risk factor (text-implied risk). The exposure to the text-implied risk of a stock is measured by a text-based risk measure, document risk score (DRS). I construct a factor mimicking portfolio TXT for the text-implied risk by sorting stocks based on DRS and estimate its risk premium. I nd signi cant positive risk premium for TXT , and the adjusted R of the Fama-MacBeth cross-sectional regression increases signi cantly by adding TXT into a four factor model that includes the Fama-French three factors and a moment factor.