Employee-Friendly Corporate Culture and Firm Performance: Evidence from a Machine Learning Approach

Abstract

This study investigates which values of an employee-friendly (EF) corporate culture are the most important predictors of firm value and operating performance using a novel social media dataset of approximately 250,000 crowdsourced employee reviews of 18 different characteristics of a firm’s corporate culture. The extreme gradient-boosting model with SHAP (Shapley additive explanations) and SAGE (Shapley additive global importance) values is used to examine the predictive value and relative importance of employee-friendly cultural values and the potential nonlinearities of these relationships. We find that several employee-friendly corporate culture features contain value-relevant information for predicting firms’ value (Tobin’s Q) and operating performance (ROA). Our findings (SAGE values) reveal three features indicating that predictive importance is clearly superior to other EF culture variables in our machine learning model: employee reviews about the overall company culture, pride in the company, and job security. Based on the SHAP values, these effects are positive, significant, and closely linear. The effects are negative for low values of an EF corporate culture and strongly positive for high values. Specifically, we find that satisfaction with the overall company culture and organizational pride are the most important characteristics for predicting Tobin’s Q, whereas job security and the overall company culture are the most crucial predictors of ROA. Other significant predictors of Tobin’s Q are the attitude towards older colleagues, work–life balance, office/work environment, inclusive/diverse, and gender equality, while environmental friendliness, gender equality, workplace safety, and inclusive/diverse are important predictors of ROA.

Keywords: Corporate culture, Employee treatment, Cultural values, Firm value, Machine learning, Gradient boosting

JEL Classification: C63, D23, G30, M14, M41