Predicting corporate innovation using machine learning and social media data

Abstract

This study explores the potential of employee reviews on social media to predict corporate innovation performance. We investigate these relationships using a novel social media dataset and an explainable machine learning approach to assess the predictive value and importance of various employee treatment policies in driving corporate innovation. In addition to traditional patent-based innovation measures, we employ a text-based innovation metric derived from 10-K filings. Our findings reveal that several employee ratings on social media provide valuable insights for predicting corporate innovation. Specifically, we highlight the importance of flexible working hours and employee stock or equity options in predicting patent counts, patent citations, and text-based innovation. Other significant predictors of patent-based innovation include employees' career growth prospects and pride in the company. Furthermore, we find that the ability to work remotely is a strong predictor of text-based innovation but is less significant for patent counts and citations. Our findings reveal notable differences in the key determinants of various types of innovation, contributing to a deeper understanding of how employee experiences associate corporate innovation outcomes.