The Commercial Potential of Science and Its Realization

Résumé du livre

The research of universities and government labs--what we succinctly call “science”-- contributes to technical advancement and economic growth. However, challenges in the commercialization process can prevent the full realization of the potential contribution of science. A first step in understanding the nature and extent of those challenges is identifying scientific contributions that offer commercial potential. Second, it is important to understand what factors may limit society's ability to capitalize on those contributions fully. In this paper, we argue that a key barrier to understanding where commercial opportunities lie and the frictions limiting their fruition is the inability to accurately measure a scientific discovery's ex-ante commercial potential. To address this issue, we introduce and validate a new text-based measure of commercial potential for scientific articles using a large language model (BERT). Using this measure, we analyze the mechanisms driving the commercialization of over 124,000 articles at a leading research university. Our measure predicts which science is disclosed, investments by the TTO, patent filings, startup formation, and licensing--even after accounting for field and author-level differences shown to predict commercialization in past research. We find that the greatest friction in the commercialization process arises early in the process, when scientists decide whether to disclose their science to the TTO; a sizable majority of high potential articles are never disclosed. To explore whether undisclosed science may still yield commercial value, we examine firms' use of it outside of the TTO channel and find that a majority remains uncited in corporate patents. Additionally, we apply our measure to nearly 4 million articles in the U.S. and find that articles with high commercial potential are more likely to be incorporated into patents that are renewed. Commercialization rates are, however, 40% lower for high-potential discoveries originating from institutions with historically low commercial impact, holding constant other factors related to the science, author, field, and university.