In knowledge we trust: learning-by-interacting and the productivity of inventors

Innovation rarely happens through the actions of a single person. Innovators source their ideas while interacting with their peers, at different levels and with different intensities. In this paper, we exploit a dataset of disambiguated inventors in European cities to assess the influence of their interactions with co-workers, organizations' colleagues, and geographically co-located peers, to understand if the different levels of interaction influence their productivity. Following inventors' productivity over time and adding a large number of fixed effects to control for unobserved heterogeneity, we uncover critical facts, such as the importance of city knowledge stocks for inventors' productivity, with firm knowledge stocks and network knowledge stocks being of smaller importance. However, when the complexity and quality of knowledge is accounted for, the picture changes upside down, and closer interactions (individuals' coworkers and firms' colleagues) become way more important.