Abstract
Previous research on breakthrough innovations has used patent data to identify them and assess their impact. The main proxy for breakthroughs uses forward citation counts, where patents at the top of the distribution are considered breakthroughs. Scholars have found this metric correlates with the economic value of patents (i.e., stock market valuations), yet, it does not tell us much about their technological content. We propose a new methodology – topic modeling of patent texts – to distinguish cognitive from economic breakthroughs. In our test case analysis of 2,826 nanotechnology patents, we find that cognitive breakthroughs are more likely to be highly cited, yet the mechanisms that produce cognitive and economic breakthroughs are quite different. Moreover, patents that are cognitive as well as economic breakthroughs have a bigger and more enduring impact on future inventions. This approach gives us traction in understanding the emergence and evolution of technologies over time.