Abstract
We argue that the standard statistical models commonly used by management scholars to investigate the relationship between prior entrepreneurial experience and the probability of switching into entrepreneurship from wage employment are less capable of unveiling the true relationship, especially in the context of big data. In particular, because the immense volume of data means that almost everything can be significant, the statistical significance relying on p-values may not imply economic significance. In addition, in the context of big data, more flexible relationships than simple linear relationships (linear, curvilinear, cubic, etc.) are possible, yet the standard statistical models that pre-specify the linear relationships between the independent and dependent variables lack the capability of detecting such relationships. To illuminate these concerns, we re-visit this important relationship using two different models: logistic regression – a standard statistical model commonly used by management scholars, and random forests – a powerful machine learning tool for analyzing big data. Through comparing the discrepant findings of these two models, we assert the benefits of using contemporary approaches to handle big data in re- visiting fundamental questions in management.