From “Buttons to Biotech” to “Progress and Potential” and Beyond: the USPTO Studies on Gender Gap in Patenting

The United States Patent and Trademark Office (USPTO) has been measuring women inventorship for decades and is currently deepening the research. Learn the main results and new methods they developed to predict the attribution of inventors’ gender and patents’ topic.

Applicants in the United States of America (US) are some of the most active users of the patent system. However, the country has a long history of diversity gaps in innovation, similar to other developed economies. And like most IP offices in these developed economies, the USPTO is currently developing active research and policy agendas to strengthen women’s contribution to patenting.

In his presentation at WIPO’s Online Seminar Series on the IP Gender Gap in the Americas region, held in October 14, 2021, the USPTO Chief Economist, Andrew Toole, describes his office’s efforts in measuring women’s contribution to patenting.

1. Analyzing Gender in Patenting at the USPTO

In 1996, the USPTO published its first statistical analysis of women inventorship entitled “Buttons to Biotech”. The study examined the share of women inventors across different intellectual property (IP) instruments – including patents, industrial designs, plant patents and others - granted between 1977 and 1996. It focused on women inventors of US patents, defined as patents whose first applicants reside in the US. Data showed that for every 100 patents granted in 1977, two had at least one woman inventor. This share improved over the years: for every 100 patents granted in 1998, less than ten had at least one woman inventor.

Slightly over two decades later, in 2019, the USPTO published its second report, titled “Progress and Potential: Profile of Women Inventors in U.S. Patents”. It focuses specifically on women inventors named on US patents granted from 1976 through 2016. Data indicates that the progress continues over time: for every 100 patents granted in 2016, 21 had at least one woman inventor.

In both studies gender is assigned in a binary manner, and based on the sex of inventors at birth. However, the methodologies used to identify the gender of women inventors in IP documents differs significantly across reports. In the 1996 study, researchers manually checked the names of each inventor of the granted US IP instrument and assigned the genders based on the names. In sharp contrast, in the 2019 study, researchers were able to attribute the genders of the inventors in an automatized way: they combined WIPO's World Gender Name Dictionary with a database from IBM called "Global Name recognition database". This approach is automates the disambiguation of inventors’ names and assigns probabilities to identify their genders. Given the increasing volume of IP filings in recent years, this method saves time and effort, and is arguably more efficient.

2. It’s all about the methods

One of the important insights from carrying out this analysis is the need to develop an entity resolution policy to disambiguate inventors' names in documents.

As there are several methods on attributing gender in documents worldwide, the USPTO wanted to find which method was the “most accurate algorithm” for assigning the gender of patent inventors. Nine different methods were tested. The accuracy of the alternative algorithms was tested against a ground truth dataset containing the self-reported gender of USPTO examiners drawn from employment records.

Based on preliminary findings, WIPO’s WGND 2.0 performed the best among the different gender attribution methods. More information can be found in the upcoming USPTO report.

3. Going Forward

Dr. Toole’s team has used a machine-learning algorithm to predict whether patents belong to the Artificial Intelligence (AI) topic. The current model analyzed over 11 million patents and performed very well. The resulting database, the Artificial Intelligence Patent Dataset, and its companion journal article can be downloaded and used free of charge. The results are comparable to the work of the Canadian IP office on the same topic.

Are you interested in producing statistics and visualizations on the gender gap in US patenting? You can use the API called PatentsView maintained by the USPTO, it is completely open access. If you want to learn more about the US experience, you can also check our selection of innovation and IP and Innovation related studies on the country.

Other stories you may enjoy

 

In Canada, 1 in 8 AI Inventors is a Woman

There are several ways to get innovation and IP data with gender breakdown. Check which one suits your case better.

Analyzing the U.K.’s leaky pipeline

The term “leaky pipeline” refers to the progressive reduction in women’s participation at the different stages of the career progression in the science, technology, engineering and mathematics (STEM) fields. In many instances, this gap is most prominent at the inventor level. How can we measure how “leaky” is the “pipeline”? See how the UKIPO researchers measured the gender gap in patent data, and find out what they found in their investigation.