Patent Citations and Stock Performance: Constructing a Dynamic Industry Classification

Abstract

Commonly used industry classifications do not reliably predict stock co-movement because companies change their core structure quickly relative to the slow and haphazard updating of widely used subjective industry classifications. We construct an objective industry classification that clusters companies using patent citations to better measure changes in company structure over time. Our classification predicts daily stock co-movement between 5% and 25% better than SIC and NAICS classifications and is a statistically significant addition to a 3-factor Fama-French model. We validate our use of patent citations to classify companies by showing that companies which cite each other’s patents are significantly more likely to merge.

Keywords: classifications, stocks, industry, cluster, return, growth, Industrial Organisation, patents

JEL Classification: C38, O3, G00, G3