Predicting new venture survival: A Twitter-based machine learning approach to measuring online legitimacy

Research indicates that interactions on social media can reveal remarkably valid predictions about future events. In this study, we show that online legitimacy as a measure of social appreciation based on Twitter content can be used to accurately predict new venture survival. Specifically, we analyze more than 187,000 tweets from 253 new ventures' Twitter accounts using context-specific machine learning approaches. Our findings suggest that we can correctly discriminate failed ventures from surviving ventures in up to 76% of cases. With this study, we contribute to the ongoing discussion on the importance of building legitimacy online and provide an account of how to use machine learning methodologies in entrepreneurship research.