Employer ratings in social media and firm performance : Evidence from an explainable machine learning approach

Abstract

This study examines the ability of crowdsourced employee opinions about their workplace to reveal value-relevant information about corporate culture. We investigate the employee-friendly (EF) corporate culture values that are strongly associated with firm value and operating performance using a unique social media dataset of approximately 250,000 crowdsourced employee reviews to evaluate 18 distinct characteristics of a firm's corporate culture. The explainable machine learning model is used to examine the nonlinear associations and relative importance of employee-friendly cultural values. We find that several employee-friendly corporate culture features are associated with firms' value (Tobin's Q) and operating performance (ROA). Our findings reveal two features whose association is clearly superior to other EF culture variables in our explainable machine learning model: pride in the company for Tobin's Q and job security for ROA. Based on the SHAP values, their effects are positive, significant, and relatively linear.