The fault in the stars: exploring employee sentiment and stock returns through textual analysis
This study investigates whether employee sentiment from Glassdoor reviews is related to stock returns. Prior research relies on raw star ratings, but we show these measures are biased due to a 2012 change in Glassdoor's review process and a subsequent upward drift in scores. We compile over two million Glassdoor reviews for Russell 3,000 firms from 2008–2019. Using multinomial inverse regression, we estimate sentiment directly from review text to address star rating biases. We then form value-weighted stock portfolios sorted on changes in sentiment. Portfolios formed on textual sentiment changes deliver significant risk-adjusted returns, while those based on raw star ratings are inconsistent. Our evidence suggests that textual reviews contain more reliable and predictive information than numerical scores alone and that earlier findings based on star ratings may be overstated. Analysts and investors should be cautious when using Glassdoor star ratings in investment decisions. Textual reviews offer a viable measure of employee sentiment and firm fundamentals. This is the first study to document the change in Glassdoor collection procedures and subsequent higher average employer ratings. We demonstrate that textual analysis provides a viable method for the examination of employee sentiment. Our approach refines the link between employee satisfaction and asset pricing and offers practical insights for market participants.