Innovation Networks, Asset Pricing, and Machine Learning

Linda Zhao and Junhui Cai, Statistics, The Wharton School, and Wu Zhu, University of Pennsylvania

Abstract: Firms or agents are linked via various linkages, which significantly impacts business cycle and asset pricing at the macro level, and firm decision and investor behaviors at the micro level. This project uses detailed firm-level data to construct dynamic firm-to-firm innovation networks from 1919 to 2018 and explores the economic value as well as innovation opportunities and risks in innovation networks. At the macro level, we highlight the crucial role of innovation diffusion across firms in propagating the shocks into the whole economy and amplifying uncertainty of the future economy. At the micro level, we emphasize that managing the idiosyncratic shocks exposure to firms in the center of the innovation network is crucial to mitigate future growth uncertainty. We further examine how the link-complexity of the innovation network affects the efficiency of investors’ information acquisition and shapes investors’ decisions. Finally, if linkages between firms are unobservable, we propose to adopt a deep learning method to estimate a deep state-space model with high dimensionality. Overall, this proposed project provides new insights on how firms make innovations through standing on the shoulders of giants, and how innovation diffusion between firms affects the overall economy and complicates information presented to investors.