Geocoding of worldwide patent data

How can we provide geographic coordinates for inventor and applicant locations in patent documents worldwide?

We care because accurate geographic data are essential for studying patenting activity and for understanding innovation at local levels. Traditional patent office data can lead to biases and incomplete analyses because they do not capture all innovation activities occurring across different jurisdictions. Furthermore, precise location data facilitate research on knowledge spillovers, regional disparities in innovation, and policy-making relevant to regional development.

We compiled a dataset of 18.8 million first filings of patents from 46 countries between 1980 and 2014. We used multiple patent databases (including PATSTAT and REGPAT) to recover missing address information via imputation methods from subsequent patent filings and ensured the data were geocoded accurately. Dataset construction involved rigorous data cleaning, geolocation, regionalization, and final assembly processes that included technical validation steps.

Our new dataset covers 81 percent of all first filings worldwide over the defined period, allowing geographic activity to be analyzed more granularly. Technical validation indicated a low miss-assignment rate, with only about 2 percent potentially being incorrect. Importantly, the dataset is structured to connect seamlessly with existing patent databases for enhanced usability in research and policy analysis.

• 18.8 million patent documents from 46 countries filed between 1980 and 2014
• Data combined from nine national, regional, and international patent databases
• Coverage of 81 percent of all first filings globally during the study period

• The imputation of missing address data is accurate and represents the true localization of inventors and applicants
• Geocoding services provide reliable geographic coordinates with a low error rate
• The distribution of patents across jurisdictions is representative of actual innovative activity