Discrimination against foreigners in the U.S. patent system

Is there a systematic bias against foreign applications in the grant of U.S. patents?

We care because understanding potential discrimination within the U.S. patent system is crucial as it may deter foreign inventors from seeking patents, affecting international trade and R&D investment. The existence of a bias against foreigners could undermine the national treatment principle essential for fair competition in innovation, particularly as global patent systems rely on equitable treatment to encourage multinational firms to protect their inventions. Furthermore, insights into discrimination can guide policy reforms aimed at equalizing access to the patent system, thus fostering a more inclusive innovation environment.

We conducted an empirical analysis of approximately 1.5 million U.S. patent applications filed from 2002 to 2012 using data that are primarily sourced from the PATSTAT database, the USPTO PAIR database, and relevant APIs for name classification. We employed statistical techniques, including linear probability models controlling for fixed effects of patent agents, examiners, and art units, to assess the grant probabilities for domestic versus foreign applications.

We find that foreign patent applications are approximately 10 percentage points less likely to be granted than domestic applications, a result that remains significant after controlling for application quality and patent agent effectiveness. Additionally, we find that the disparity in grant rates is not due to intentional discrimination but is largely explained by differences in financial resources and the level of effort applicants invest in the patent prosecution process. Notably, while foreign inventors do exhibit lower grant success, they also have lower engagement in the prosecution process, leading us to conclude that the patent system operates with 'disparate impact' rather than intentional bias.

• 1.5 million U.S. patent applications from 2002-2012 sourced from PATSTAT and USPTO PAIR databases, encompassing applications with twin counterparts in eight jurisdictions
• Random sample of 1 million observations for each patent family type used in analysis

• We assume that the determination of grant probability relies on fixed effects controlling for differences among patent agents, examiners, and art units
• We assume that the quality of patent applications is consistent across jurisdictions and comparable using the twin patents methodology
• We assume that financial resource availability through GDP per capita and applicant portfolio size significantly influences patent grant outcomes