Predicting Corporate Innovation Using Machine Learning and Social Media Data

Abstract

This study explores the ability of employee reviews in social media to predict the innovation performance of companies. We explore these links using a novel social media dataset and an explainable machine learning-driven research approach to examine the predictive value and importance of different employee treatment schemas with various types of corporate innovation. In addition to more traditional patent-based innovation measures, we employ a text-based innovation measure from 10-K fillings. We find that several employee ratings in social media contain value-relevant information in predicting corporate innovation. Specifically, we show 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 for both patent-based proxies include employees' career growth prospects, opportunities for professional development, and company pride. Our findings also indicate that text-based innovation is a strong predictor for patent counts and citations and that there are several notable differences between the meaningful predictors for different types of innovation.

Keywords: Social media analytics, Human resource policies, Corporate innovation, Machine learning, Textual analysis