Ruveyda Gozen, London School of Economics
Michael Andrews, University of Maryland
Enrico Berkes , The Ohio State University
Who are America’s inventors? While patent documents record inventors’ names and places of residence, they provide no other information about inventors’ identities. Researchers must therefore use several techniques to determine inventor demographics. In this paper, we compare two of the most common methods to identify inventor gender, and then document how these different methods affect our inferences about the prevalence of female inventors and the types of inventions they tend to create. One commonly used method is to use inventors’ first names to infer gender. To operationalize this method, researchers usually rely on pre-existing gender guesser routines that tend to be calibrated using modern data. A second method is to use inventors’ names and locations to link patentees to another data source that includes demographic information, such as linking to the U.S. decennial population censuses. We consider multiple variations for each method, for instance how to handle androgenous names and how to handle cases in which one patent record plausibly links to multiple records in the U.S. census. Different methods produce quite different results for the share of U.S. patents attributed to women inventors, ranging from about 0.7% to more than 7.5%. While the shares differ widely, we verify that different methods are correlated with one another, both across time and across geographic space. We further conduct an audit to determine which women inventors can be identified using first names but not when linking to the census, which women inventors can be identified when linking to the census but not when using first names, and which are identified using both methods. Finally, we measure the extent to which the distribution of patent technology classes, average patent quality, and patent novelty change when using different methods. Throughout, we document several novel empirical facts about patenting by women in the U.S.
No extended abstract or paper available
Presented in Session 166. Labor, Patents, and Regulation: Insights from IPUMS Full Count Census Data