Abstract

Policy makers look to propel growth through innovation, however beyond only targeting the total quantity of innovation, policymakers are increasingly concerned with its direction. Teams drive science and their organisation will be key to meeting technological challenges facing modern economies. This paper constructs a high-dimensional knowledge space and studies how team knowledge composition advances the innovation frontier by altering the direction of progress. I define a spatial mapping of inventor teams and patents to reconceptualise knowledge production. I estimate a set of key latent variables to empirically approximate the knowledge space using a model of Latent Dirichlet Allocation on patent texts. The paper tackles a set of hypotheses on which systematic evidence is lacking. I show that there exists a hump shape relationship between the quantity of prior work and the probability a team produces a breakthrough and that patents which are intellectually close to contemporary inventors are more likely to spark breakthroughs as inventors jump on the bandwagon. In addition, by adding the right inventor, teams can pivot and change the direction of their future patents by increasing their innovation capabilities. I provide causal evidence for these hypotheses using premature inventor deaths as a quasi-natural experiment.