UC Berkeley

This thesis examines the effect of the nature of a startup firm's underlying technological relationships, as measured by patent citations, on its commercialization strategy, likelihood of success, and value when successful. The innovation-lineage view of patent citations that has been articulated and explored in the literature to date has left several empirical puzzles and is insufficient for the task. I therefore propose and validate an economic view of patent citations. Using variation in whether a cited patent is in-term or expired at the time that a citing patent is applied for, I provide evidence that patent citations can and do convey economic information directly pertaining to a patent-holder's right to exclude. I then exploit whether citations are in-sector or not, and whether it is the number of cited owners or cited patents that explains meaningful variation in outcome measures, to uncover a fourth degree of heterogeneity in patent citations -- whether citations convey information about complements and substitutes. Using samples that almost comprise the population of successful startups and failed venture-capital-backed firms that experienced their liquidity event between 1986 and 2006, and a reference dataset of all U.S. patents from 1963-2006, I find that the economic information in patent citations is of first-order importance and that startups file for patents on technologies that are at least partly substitutes. Furthermore, my results are consistent with start-ups that achieve initial public offerings patenting more radical inventions and acting as a force for creative destruction. The substitution effect of some patents explains both why receiving more citations does not necessarily make patents more valuable and why positive value effects are more likely to be found in the tail of citation count distributions. Overall, the economic view provides richer information than the innovation-lineage view, has strong construct validity, and provides a foundation for further research using patent-citation-based measures.