Crowdsourced Employee Sentiment and Abnormal Stock Returns

Abstract

Previous literature has found crowdsourced employee sentiment obtained from Glassdoor.com is related to stock returns. Evidence has shown this data can suffer from some abnormalities which may limit its usefulness. To account for these discrepancies, we utilize textual analysis through the multinominal inverse regression method to create monthly firm specific expected employee sentiment indexes for US public firms from the period 2008-2019. We test the implications of the previous literature to determine if our expected employee sentiment index is related to the cross section of abnormal returns. We find evidence that our expected employee sentiment index is related to the cross section of abnormal returns. Further, we find that employee sentiment as estimated similar to the previous literature is no longer related to stock returns. These result are robust to different specifications.

Keywords: Glassdoor, Textual Analysis, Employee Satisfaction, Abnormal Returns.

JEL Classification: G12,G41,G39