Development of a Job Satisfaction Index Based on Employees’ Online Reviews of Their Employers

This study investigated whether job satisfaction can be estimated using employees’ narrative reviews of their workplace. We analyzed more than 250,000 company reviews using natural language processing (NLP) and machine learning techniques, and we developed two context-based text analysis algorithms that predicted the cognitive component of job satisfaction (i.e., employees’ ratings of their jobs) and the emotional component of job satisfaction (i.e., whether the job review was intended to be positive or negative). Our predictive models categorized 90% of the reviews as either positive or negative, or provided indices which correlated strongly with job ratings (i.e., correlation values up to .80). Findings supported the idea that employees’ attitudes towards their jobs can be successfully predicted using employees’ reviews. Despite having a series of limitations, the study highlights the potential of assessing psychological variables using NLP-based methods. Given that our algorithms provide accurate predictions of employees’ attitudes, our practical implications include an open-access resource for predicting job satisfaction using employees’ narratives.