USING MACHINE LEARNING TO ADVANCE THEORY: PRIDE IN WORK AS A NOVEL ANTECEDENT OF JOB SATISFACTION

Management researchers typically identify new constructs by analyzing existing theories, generalizing from observations, and making connections with neighboring social sciences. We propose that machine learning methods can serve as yet another tool for generating novel hypotheses. We used lasso regression and deep learning to identify potential antecedents of job satisfaction in the World Values Survey. Both models identified pride in work as the top predictor. A literature review revealed that no management research to date has conceptualized pride in work as a construct. A three-wave survey found that pride in work predicted job satisfaction above and beyond meaning in work, a long-known predictor, and pride experienced at work. Further, the effects of skill variety, task identity, and task significance on job satisfaction were simultaneously mediated by pride in work and meaning in work. The findings suggest that pride in work is a missing construct in the literature, both as an antecedent of job satisfaction and as a mechanism through which core job characteristics influences job satisfaction. This research thus demonstrates that machine learning methods can be used to identify novel constructs and to extend theory in the organizational sciences.