Measuring open innovation practices through topic modelling: Revisiting their impact on firm financial performance

Despite the popularity of open innovation in recent years, studies examining the impact of open innovation upon firm performance have shown mixed results. Previous empirical work on this topic is often based on surveys or archival sources, usually done either in isolation or in aggregate through employing proxy measures. In contrast, we employ an unsupervised learning technique (i.e., topic modelling) utilizing natural language processing to extract information on companies' open innovation practices, creating an initial keyword basket for future development. We then revisit the relationship between open innovation practices and financial performance of firms. The results show that a firm's overall openness level is associated with improved financial performance. More granular practices developed from our approach, however, show variations. The inverted U-shaped relationships are observed in specific open innovation practices but not in all, partly supporting the existence of the openness paradox from prior literature. The complementarity between internal R&D and individual open innovation practices also varies by practice. Further, the influence of these open innovation practices also varies by sector. Our findings prompt us to conclude that open innovation's impact on financial performance is nuanced, and that there is no uniform set of best practices to practice open innovation effectively.