Global Mobile Inventors (GMIs) are patent inventors who have filed patents in more than one country during their career. These inventors carry tacit knowledge across borders, driving innovation diffusion and connecting research communities worldwide. This site visualizes those movements to help understand how the dynamics of innovation are changing around the globe over time, in part given this important rise in the global movement of inventors.
For more information, see: Bahar, D., Choudhury, P., Miguelez, E., & Signorelli, S. (2024). Global Mobile Inventors. Journal of Development Economics, 171, 103357. https://doi.org/10.1016/j.jdeveco.2024.103357
How global inventor mobility has evolved over five decades.
Geographic distribution of GMI inflows, outflows, and net movement by country.
Select a country on the map to see detailed statistics
Visualizing the flow of inventors between countries. Width represents the number of Global Mobile Inventors moving between origin and destination countries.
Select a flow on the diagram to see detailed statistics
Select a time period to see the top migration corridors for inventors.
Explore how GMI inventors collaborate with local inventors. The force-directed layout naturally positions well-connected (central) inventors toward the middle, while peripheral inventors drift to the edges. GMIs (cyan) use the "always" definition: once an inventor patents in multiple countries, they are considered a GMI forever. Countries shown are the 60 with the most GMIs.
This feature is in beta. We are actively refining the network metrics and visualization. The technology filter does not apply to this section.
Eigenvector centrality from complete network
How this network is built: We include all GMI inventors who patented in the selected country during the period, plus their direct co-inventors. The visualization shows a random sample; centrality statistics above are computed from the complete network.
This interactive visualization explores Global Mobile Inventors (GMIs) — inventors who patent in multiple countries over their careers, potentially facilitating the emergence of new technologies in the countries they arrive to. It is based on the findings of the research paper "Global Mobile Inventors" published in the Journal of Development Economics in 2024 by Bahar, Choudhury, Miguelez, and Signorelli.
The research shows that GMIs are "superstar" inventors: they have longer careers, patent more frequently, and produce higher quality patents than the average inventor. While fewer than 0.5% of patents in the early 1970s involved a GMI, by 2015 roughly one in three patents involved at least one GMI.
The research also shows that the first few patents filed within a country-technology pair are twice as likely to be invented by a team including a GMI with prior experience in that technology — tangible evidence that GMIs facilitate technology-specific diffusion of knowledge across nations.
The data for this website is derived from the USPTO PatentsView database (November 2024 release), covering over 9 million granted patents from 1976–2025, belonging to over 4 million disambiguated inventors. This latest release includes more data than the one used in the original study, and is based on PatentsView's 2020 disambiguation update which uses advanced "Tree Grafting" algorithms to more accurately track inventor careers, resulting in higher GMI counts (approximately 1.5–2.4x depending on period) due to better tracking, not methodology changes.
The identification of GMIs for this visualization follows exactly the same methodology as in Bahar et al. (2024). A GMI is identified when an inventor's country of residence changes between consecutive patent filings, using the patent's priority date.
Technology Filtering: The technology filter uses the Cooperative Patent Classification (CPC) system, organized hierarchically by 8 sections (e.g., "A - Human Necessities") and ~600 subclasses (e.g., "G06F - Electric Digital Data Processing"). Each GMI movement is assigned to the primary CPC classification of the associated patent. Note that approximately 5% of patents in PatentsView lack CPC classification data; this percentage is higher (~11%) for the 2021-25 period, as recently granted patents may not yet have CPC assignments. Technology-filtered statistics therefore represent the subset of GMI movements with available CPC data.