This paper studies the effects of the NSF Science Development Program on universities and local innovation, combining historical data from scientific publications, doctoral dissertations, and patents. Introduced in 1965, the program awarded large institutional grants to natural science and engineering departments at U.S. research universities. I exploit top-ranked universities excluded from the program as a comparison group in a difference-in-differences research design. First, I find that Science Development awards increased faculty size, the number of PhDs awarded, and publications at funded universities. Second, I find a patenting increase in commuting zones hosting funded universities, primarily attributable to incumbent private firms and driven by commuting zones with established R&D-intensive sectors. I find a larger effect in technology fields with high exposure to local universities' research. I provide evidence indicating two main mechanisms behind the patenting increase: greater reliance on scientific knowledge in patenting and the employment of local PhD graduates in industrial R&D.

We study the effects of the 1944 GI Bill on the probability to become an inventor. The GI Bill provided large subsidies to WWII veterans’ college education, covering tuition fees at any U.S. university and providing a monthly stipend. Using a fuzzy regression discontinuity design and novel linked inventor-census data, we find a positive effect of GI Bill eligibility on the probability to become an inventor. This result is driven by inventors patenting for companies or government agencies and by inventors active in the computing and electronics fields. We find positive effects also when focusing on inventors patenting high-impact technologies. We provide evidence on the individual role of college education—through GI Bill subsidies—in increasing access to inventor careers in two steps. First, we find a positive effect of GI Bill eligibility on college attendance and on college completion. Second, we find that GI Bill-eligible individuals who resided far from a higher education institution were less likely to access an inventor career. We explore the heterogeneity of our effects by parental background for both inventor and college outcomes. We find a positive effect of GI Bill eligibility for the children of middle-school and college educated individuals, as well as for the children of parents employed as craftsmen or as clerical and sales workers and as professional or technical workers.

We study the innovation effects of the Agreement on the Free Movement of Persons, signed by Switzerland and the European Union in 1999. We exploit a quasi-experimental setting created by Switzerland’s implementation policy, which initially eased entry restrictions only for commuters from neighboring countries, thereby inducing a large inflow of "cross-border inventors" in regions close to the border. We find that this policy increased patenting in these regions relative to comparable ones farther away from the border. Additionally, we find no evidence indicating the displacement of native inventors or a reduction in the patenting activity of Switzerland's neighboring countries. We also provide evidence suggesting that cross-border inventors contributed to Swiss patenting by enabling R&D laboratories to expand and hire inventors with valuable skills, who increased the productivity of incumbent inventors they collaborated with, but not of local peers outside direct collaborations.

Patents disclose knowledge, yet this disclosure is often insufficient for the knowledge to be put into practice and used for cumulative innovation. Firms rely on workers possessing tacit knowledge or specific skills to effectively build on the ideas of others. In this study, we examine the effects of expanding Swiss firms' access to the German labor market on the diffusion of knowledge developed in Germany to Switzerland. We investigate the impact of a reform implemented in 2002, which eliminated the restrictions Swiss firms previously faced in hiring German cross-border commuters. We find that following the reform's implementation, German patents originating from locations within close commuting distance to the Swiss-German border are more heavily cited by Swiss applicants. Moreover, we observe an increase in the number of new Swiss patents that are textually similar to patents from the German border region. Knowledge diffusion effects are particularly pronounced for cumulative innovation at an intermediate technological distance to the original German invention. Such inventions introduce at least one new technology field of application while having at least one common field. Additionally, we find that the effects are concentrated in fields where Switzerland is relatively closer to the knowledge frontier than the neighboring German regions.

We apply modern machine learning methods to extract patent-to-patent citations from the text of USPTO patent documents. Overall, we are able to recover around 49 million “in-text” citations, 37 million of which are not found among traditional front-page ones. We show that in-text citations bring a different type of information compared to front-page citations. First, we observe weak relationships between bibliometric measures derived from each citation type, such as forward citation counts and various kinds of self- citation. Second, the positive relationship between inventions’ reliance on science and their impact is stronger when measuring the latter using in- text citations. Lastly, in-text citation pairs exhibit higher textual similarity and are more geographically localized than front-page ones.