Born to Invent? Social Origins and Environmental Exposure of Inventors in Historical Britain, 1891-1911, with Myra Mohnen （Working Paper）
[DownloadFullPaper]
Abstract: Who becomes an inventor, and what do they choose to invent? Using a novel dataset linking patent records to complete-count census data in England and Wales between 1851 and 1921, comprising 56.9 million observations of men aged 18–80 including 15.5 million father–son pair observations, we show that family background and local environment are associated not only with entry into invention but also with the technological domains inventors enter. Both father's occupation and parish-level inventor density strongly predict who becomes an inventor, with sons of professionals having 1.6 times the odds of sons of farmers. To examine the technological fields in which inventors patent, we apply the large language model PatentBERT to predict missing patent technology codes. We find that inventors patent disproportionately in fields closely aligned with their fathers' occupations and the occupational structure of their residential parishes and birth counties. This pattern persists among migrants who relocated to different counties, suggesting that early-life exposure is linked to lasting technological specialization. These findings imply that unequal exposure may shape not only the supply of inventors but also the direction of inventive activity.
Gender Gap in Patent Opposition (Working Paper)
Abstract: Patent opposition refers to challenges made after a patent has been granted in the U.S. Are patents led by female inventors more likely to face opposition? This paper examines the gender bias in patent opposition. The study uses a comprehensive dataset of approximately 6 million USPTO granted patents between 1976 and 2023. By applying advanced Natural Language Processing (NLP) techniques and training gender-specific neural networks, this paper estimates the opposition probability of female-led inventors based on the patent claims and performs a counterfactual analysis investigating what would have been the opposition probability had this same patent been granted to the opposite gender. The results show an overall gender bias in granted patent opposition against female-led patents. The analysis estimates that female-led patents are approximately 15.2% overchallenged relative to the actual female opposition rate, while male-led patents are 9.5% underchallenged. The bias against female-led patents can be mitigated through specific channels. The presence of a star inventor dramatically reduces bias by 34% of the baseline gap, suggesting that external validation from elite collaborators helps address the credibility deficits female inventors face. In the broader environment, female representation in the technology field they innovate in also plays a critical role: bias diminishes dramatically in fields with higher female representation. However, a paradox emerges in high-stakes contexts: affiliation with prestigious institutions and pursuit of patents with global market impact increase bias for female inventors by 15% and 11% respectively, indicating that rather than signaling quality or enhancing credibility, visibility in competitive environments amplifies bias against female inventors.
AI-assisted Teams Outperform AI-led Teams but Not Human-only Teams in Assessing Research Reproducibility in Quantitative Social Science'', with A. Brodeur, D. Valenta, A. Marcoci et al., Proceedings of the National Academy of Sciences, 123(22), e2524747123, 2026.