Working Paper: NBER ID: w22244
Authors: Gatan De Rassenfosse; William E. Griffiths; Adam B. Jaffe; Elizabeth Webster
Abstract: A low-quality patent system threatens to slow the pace of technological progress. Concerns about low patent quality are supported by estimates from litigation studies suggesting that the majority of patents granted by the U.S. patent office should not have been issued. This paper proposes a new Bayesian method for measuring patent quality, based on twin patent applications granted at one office but refused at another office. Our method allows us to distinguish whether low-quality patents are issued because an office implements a (consistently) low standard, or because it violates its own standard. The results suggest that quality in patent systems is higher than previously thought. In particular, relative to the own standard of each office, the percentage of mistakenly granted patents is under 10 percent for all offices. The Japanese patent office has a greater percentage of mistakenly granted patents than those of Europe, the United States, Korea and China, largely because it has a higher standard.
Keywords: Patent Quality; Bayesian Method; Patent Examination; Intellectual Property
JEL Codes: K41; L43; O34
Edges that are evidenced by causal inference methods are in orange, and the rest are in light blue.
| Cause | Effect |
|---|---|
| systematic differences in standards across patent offices (L15) | grant outcomes (I28) |
| examiner mistakes (K40) | grant outcomes (I28) |
| Japanese Patent Office (O34) | mistakenly granted patents (O34) |
| Japanese Patent Office (O34) | patents that would not have been granted by the strictest office (O34) |
| office policy choices (L49) | grant outcomes (I28) |
| low-quality patents prevalence (L15) | litigation studies (K41) |