‘Moving On’—investigating inventors’ ethnic origins using supervised learning

Abstract Patent data provides rich information about technical inventions, but does not disclose the ethnic origin of inventors. In this article, I use supervised learning techniques to infer information. To do so, construct a dataset 96′777 labeled names and train an artificial recurrent neural network with long short-term memory (LSTM) predict origins based on names. The trained achieves overall performance 91.4% across 18 origins. model investigate 2.68 million inventors provide novel descriptive evidence regarding their composition over time countries technological fields. global has become more diverse last decades, which was mostly due relative increase Asian Furthermore, prevalence foreign-origin is especially high in USA, also increased other high-income economies. This mainly driven by inflow non-Western into emerging high-technology fields for countries.