Most research on cumulative innovation is theoretical and inconclusive. In a classic article, Suzanne Scotchmer asserted that where cumulative innovation is important, "fully efficient incenties" are "impossible" because they require awarding both the initial and follow-on innovator "the entire social surplus of his innovation." Daniel Hemel and I argue that this is wrong: for welfare-enhancing R&D projects (where the expected social benefit of that project is greater than the expected cost), incentives are "fully efficient" so long as the private benefit from patent and non-patent incentives (such as first-mover advantage) exceeds the post-tax cost. Awarding the entire social surplus to patentees will often lead to costly over-incentivization errors. But this still does not answer the empirical question of whether the existence of an early patent will promote follow-on innovation (à la Kitch's prospect theory) or impede it (e.g., by raising transaction costs). Only a couple of studies, such as Heidi Williams's work on genes, have attempted to address this question.
In this new article, Galasso and Schankerman examine 1357 Federal Circuit decisions covering 1258 patents from 1983 to 2008, and they recorded whether at least one claim in each patent was invalidated. Their primary measure of cumulative innovation was citations to the patent after the Federal Circuit decision. But wait, you might be asking—won't this tell us more about the difference between valid and invalid patents than about the effects of patents on follow-on innovation? Yes—this is known as an endogeneity problem, and Galasso and Schankerman have an elegant solution: they take advantage of the random assignment of judges to cases and judges' differing propensities to invalidate patents to create an instrumental variable, the "Judges Invalidity Propensity." The key results are summarized by the authors:
We find that patent invalidation leads to a 50 percent increase in subsequent citations to the focal patent, on average, but the impact is highly heterogeneous. Patent rights appear to block follow-on innovation only in the technology fields of computers, electronics and medical instruments. Moreover, the effect is entirely driven by invalidation of patents owned by large patentees that triggers entry of small innovators, suggesting that patents may impede the 'democratization' of innovation.In contrast with the results in high-tech fields, the authors "find no evidence of blocking in non-complex fields such as chemicals, mechanical, and pharmaceuticals," such that "patent rights block cumulative innovation only in very specific environment."
These results are significant (in both the technical and natural sense), although relying on patent citations is not without problems. Defining cumulative innovation in terms of patent citation counts biases the definition of innovation away from research that isn't patented. And while the authors control for heterogeneity in value using the number of claims and the number of citations prior to the Federal Circuit decision, as David Abrams described at PatCon3, citations are not a good measure of value.
Most problematically, patent citations might not be a good measure of the prior art that the researchers actually built on (and the authors recognize this concern in n.14). Most citations (63%, according to one study) are added by patent examiners, and many applicant citations are found by patent attorneys, not the researchers themselves. And although applicants are technically required to disclose all known prior art (including both valid and invalid or expired patents), they might be more hesitant about citing valid patents and potentially exposing themselves to licensing demands from those patentees—the effects in this paper could stem from such strategic behavior. It would be interesting to examine more post-2001 patents (when examiner and applicant citations began to be reported separately) to show whether applicants are significantly more likely to cite invalid and expired patents than examiners.