Comparison of the coverage of the USPTO’s PatentsView and the EPO’s PATSTAT patent databases: a reproducibility case study of the USPTO General Patent Statistics Reports

Abstract

Abstract Patent statistics from various sources are often cited in science, technology, and innovation (STI) studies and STI policies, usually as a proxy for measuring output and activity related to technological innovation. But how reliable are those statistics? The reproducibility of numerical data originating from these information sources may be low, as the query methodology and data coverage of those sources is often omitted or unclear in documents that report on patent-based data collection and comparative analysis. Given the importance of reliable, reproducible statistics as the foundation of evidence-informed policymaking, this study aimed to conduct a reproducibility exercise for the patent statistics produced by USPTO’s Patent Technology Monitoring Team (PTMT) and compare them with statistics derived from PatentsView and PATSTAT patent databases. Comparing the two databases, four types of coverage were investigated, namely: intellectual property (IP) types, geographic (countries and territories) coverage, owner/inventor coverage, and technological coverage. Although there were similarities in the historical trends between the coverage of PATSTAT and PatentsView, the absolute counts data presented high levels of variability. PatentsView had highly varying coverage compared to PTMT, which may be due to excessive generation of secondary fields derived from the original data. Additionally, PATSTAT covered earlier dates (from 1790) compared to PatentsView (from 1976). As for tracking of relative trends in patent output, either database could be chosen, but the choice will highly influence the outcomes of patent-based bibliometric studies.