An Empirical Analysis of Company Culture: Using Glassdoor Data to Measure the Impact of Culture and Employee Satisfaction on Performance

This paper examines the impact of culture and employee satisfaction on company performance; it considers which elements of culture are most important in specific industries: manufacturing, technology, and finance. Additionally, it explores whether these elements of productive cultures are also associated with employee satisfaction. It uses data from the MIT Sloan Management Review/Glassdoor Culture 500 database, which applied machine learning to analyze 1.2 million Glassdoor reviews. This data quantitatively measures nine dimensions of culture: agility, collaboration, customer focus, diversity, execution, innovation, integrity performance, and respect—assigning each company a score for every cultural dimension. Two depende­­­nt variables are used to measure company performance, stock growth and ROA. The Glassdoor Company Employee Company Satisfaction Rating was used to as the dependent variable for employee satisfaction rating was used as a dependent variable for satisfaction. When industries are combined, it is concluded that customer focus, innovation, performance rewards, and integrity all increase company stock performance, whereas collaboration decreases stock performance. However, when regressions are run individually for individual industries, culture has different marginal effects. Collaboration was found to be positively linked to performance in the manufacturing industry, but was associated with lower performance outcomes in the technology industry. For both technology and manufacturing, customer focus and respect were most positively associated with performance. Additionally, employee satisfaction is highly correlated with company performance. Specifically, similar cultural variables (innovation, respect, customer focus, and performance rewards) that are significant for company performance are also positively associated with overall employee satisfaction. The paper concludes that further research should be conducted on a larger, more diverse dataset. This data set should include more observations for every industry and control for job titles to see if these findings hold when controlling more accurately for industry-related effects