Low-Quality Patents in the Eye of the Beholder: Evidence from Multiple Examiners

How can we measure patent quality across different patent offices and discern the sources of low-quality patents?

We care because concerns about low-quality patents are prominent, as they may hinder technological progress and the efficient allocation of resources in innovation. If patent offices grant low-quality patents due to lenient standards or inconsistent decisions, this can create uncertainty for firms around patent validity, impacting their willingness to invest in research and development. Furthermore, understanding patent quality is essential for policy formulation aimed at improving patent systems globally.

We developed a formal model to estimate the quality of patents based on 408,133 inventions with linked patent applications from five major patent offices (EPO, USPTO, JPO, KIPO, CNIPA) covering over a million applications. Our model distinguishes between systematic office standards and inconsistencies in decision-making using a probit model to analyze grant probabilities.

Our findings suggest that systematic differences in office standards account for a significant portion of mixed grant-refusal outcomes, with only up to 10% of granted patents being inconsistently issued according to an office's own standard. Notably, the JPO has the highest refusal rates at about 19% compared to others, illustrating the stricter standards at play. Additionally, we estimate that much lower rates of internal inconsistencies exist compared to previous litigation studies, offering a new perspective on patent quality.

• A population of 408,133 inventions examined at least twice in the EPO, USPTO, JPO, KIPO, and CNIPA between 2000 and 2006, with a total of 1,064,513 patent applications analyzed.

• Patent offices have varying de facto standards for granting patents.
• Examiner consistency is affected by the subjective nature of the evaluation process.
• Patents are submitted to multiple offices without selection bias based on their quality.
• The model accurately captures the relationship between inventive merit and grant probability.