Quantifying Open Innovation: A Machine-Learning Textual Analysis Approach

Abstract

While the concept of open innovation has drawn the attention of scholars and practitioners alike, it still remains difficult to capture it particularly in a quantitative, longitudinal setting where data about open innovation activities is largely absent. This impedes an adequate assessment of open innovations’ long-term implications for firms’ innovation and financial performance. While researchers have extensively examined the former link, the latter is controversially discussed against the backdrop of the costs and benefits of open innovation but remains largely unexplored. In this study, we develop and validate a longitudinal, text-based measure for firms’ open innovation activities, and probe related performance implications in a longitudinal, cross-industry setting. Combining machine-learning content analysis to create an open innovation dictionary, we analyze the 10-K annual reports of 8,491 publicly listed firms in the U.S. between 1994 and 2016. Our results support our theorizing that curvilinear relationships that take an inverted U-shape exist both between open innovation and innovation performance, and open innovation and financial performance.