Benjamin Büttner

The patent system is primarily designed to incentivize innovation by granting temporary exclusivity in exchange for the disclosure of technological knowledge. Yet its function may have broader, underexplored implications: it may also serve as a navigational tool that helps firms identify emerging technologies, reduce search costs, and source external talent. By providing a structured repository of knowledge, the patent system helps firms identify ’gaps’ in their knowledge. This paper investigates whether such exposure is associated with increased recruitment of high-skilled foreign workers, with a particular focus on examiner-added citations as an exogenous source of knowledge. Focusing on the flow of talent from China to the United States, we combine data on US patent citations to Chinese patents with H-1B visa applications by Chinese nationals from 2009 to 2017. We first estimate the relationship at the county-industry-year level to capture broader high-skilled labor market effects, and then examine whether the effect persists at the level of individual firms. Our findings reveal a robust, positive association of US examiner-added citations to Chinese patents and subsequent H-1B visa applications by US firms for Chinese nationals. The effect is strongest in technology-intensive sectors and regions where China is at the global frontier and is particularly pronounced among California-based firms with high innovation intensity and highly competitive industries. These results suggest that the patent system, beyond its intended legal functionality, acts as an infrastructure for reducing search barriers and costs: by surfacing previously unfamiliar but relevant knowledge, it helps firms identify knowledge and capability gaps and guides global talent acquisition.

This paper examines whether patents can substitute for restricted-access scientific publications as a source of knowledge. We focus on cases where scientific content is replicated or disclosed in patent documents, enabling downstream users to access findings without paywall barriers. Using matched patent–paper data and citation outcomes, we show that such disclosure through patents disproportionately benefits resource-constrained actors, suggesting that patents may partially mitigate access inequalities in science.

Sir Isaac Newton’s famous remark, ’If I have seen further, it is by standing on the shoulders of giants’, captures a well-established understanding of scientific and technological progress: Ideas build upon what came before. The cumulative nature of innovation depends on the ability to identify, access and understand previous contributions. Sharing scientific information enables problem-solving and accelerates innovation, yet making disclosure effective in practice requires coordinated effort. Maintaining global databases, standardizing reporting, and ensuring timely access demand substantial investment and policy interventions. Disclosure alone is not enough. Disclosure makes information formally public, yet diffusion determines who can locate, access, and apply it. Knowledge that cannot be found or interpreted remains effectively inaccessible. This dissertation examines how such barriers shape diffusion and whether the patent system can mitigate them. The patent system grants temporary exclusivity in exchange for detailed technical disclosure, creating a paradox: An instrument designed to restrict use simultaneously generates one of the largest repositories of freely accessible codified technical knowledge. This exclusivity–disclosure bargain makes the system both a diffusion channel and a mechanism of restriction. It also provides a unique data source. Because inventors must disclose all prior art relevant to their inventions, patent citations function as ’paper trails’ of how knowledge travels and is applied. By linking patents to scientific articles, firm-level data, and visa information, I measure diffusion patterns that would otherwise remain unobservable. The empirical analysis combine econometric methods with large-scale data science approaches, including deep learning for image and text recognition. Three structural barriers are studied empirically. Language constitutes the first barrier. Using a difference-in-differences design around the 2013 introduction of machine translation of Google Patents, I show that making Chinese patents accessible in English increased follow-on use by US inventors, with the strongest effects in frontier technologies and among smaller actors that lack resources to access foreign-language content. Access restrictions constitute the second barrier. I develop a deep-learning image-recognition pipeline that identifies scientific articles and patents disclosing the same underlying knowledge, producing a large dataset of patent–paper pairs. When the scientific article is behind a paywall, the linked patent provides an alternative channel through which the same knowledge becomes freely available. Closed-access papers with a linked patent receive substantially more citations from both scientists and inventors, with the largest effects among small firms and resource-constrained individuals. Discovery and search costs constitute the third barrier. Patent examiners independently add citations to prior art during the patenting process exposing applicants to new foreign knowledge. I use examiner-added citations to Chinese patents as exogenous exposure to foreign knowledge and show that US firms respond by filing works visa applications for Chinese nationals. The effect is strongest in sectors where China holds a technological advantage, consistent with firms identifying capability gaps through the patenting process and recruiting international talent to access unfamiliar knowledge. The three empirical chapters show that lowering barriers to knowledge diffusion generates measurable gains. When barriers fall, researchers and inventors respond immediately, indicating that the core constraint is not demand for knowledge but the difficulty of reaching or interpreting existing work. Some barriers are inherent, such as language or technical complexity, but can be reduced through technologies like machine translation. Others arise from institutional or commercial choices, such as paywalls or restrictive visa policies, and require policy intervention. Creating the conditions for broad and timely diffusion therefore requires active efforts to lower barriers across all disclosure channels, particularly for resource-constrained researchers, firms, and institutions. Lowering these barriers expands the pool of actors who can build on existing knowledge, which is essential for sustained scientific progress and innovation.