The effect of Innovation Similarity on Asset Prices : Evidence from Patents’ Big Data

We analyze 7.7 million patents via textual analysis to develop a novel patent-to-patent measure of innovation similarity among firms. With this measure, we investigate whether stock price movements reveal information about firms' technological connectedness. We find that investors impound information about innovation similarity into stock prices, although not immediately and fully; stocks of technologically similar firms cross-predict one another's returns. This effect is distinct from the industry and supply-chain momentum and predictable lead–lag relations between large and small firms. A trading strategy that buys (shorts) stocks of technologically similar peers that earned high (low) returns during the previous month yields a 1.29% monthly return. Return predictability is stronger for firms with complex patents or with limited data dissemination about their technologies. Return predictability weakens with information transparency proxied by common analyst coverage of technologically similar pairs. In contrast with technological similarity determined with patent class filings, our Big Data text-based return predictability is still active. These results suggest investor attention constraints as a potential explanation for the technology momentum