HETEROGENEOUS NETWORKS AND STARTUPS’ SUBSEQUENT FINANCING: THEORIZATION BASED ON MACHINE LEARNING

Startups’ interorganizational networks with venture capitalists (VCs) and technological partners largely affect their likelihood of obtaining external financing, and different networks (e.g., startup-VC investment network, VC syndication network, and startup-partner co-patenting network) may function differently and collectively. However, prior studies have mainly regarded different networks as homogeneous or isolated due to the difficulty dealing with heterogeneous, connected networks by econometrics methods. Following the “Algorithm Supported Induction for Building Theory” research paradigm, we construct heterogeneous, dynamic networks by using machine learning and self-develop an innovative hybrid composition layer to combine multiple networks. The results show that our model predicts startups’ subsequent financing more precisely than human investors or machine learning methods without the hybrid composition layer. We further find that a hybrid of investment network and syndication network outperforms that of investment network and co-patenting network, and theoretical propositions are developed. Our study contributes to the literature by highlighting the importance of heterogeneous networks in determining startups’ external financing and improving understanding of different stakeholders and their networks.