The Journal of Things We Like (Lots)
Select Page
Janet Freilich & Sepehr Shahshahani, Measuring Follow-On Innovation (Feb. 20, 2022), available at SSRN.

Ask any patent law student why we have a patent system, and they are likely to answer that patent law addresses a fundamental market failure: the free-riding by non-inventors on the inventions of others. A patent holder’s right to exclude others from making and using her patented invention addresses free-riding directly, restoring ex ante incentives to invest in innovation. But in solving the free-riding problem, patents create a second-order problem—one that is inextricably linked to the dynamics of innovation itself. Because all knowledge, and therefore all innovation, is cumulative, patents make innovations that build upon a patented feature more costly for parties other than the inventor, who must license an invention if they are to build upon it.

The problem of “follow-on” innovation has long preoccupied both economists and legal scholars. In their excellent paper, Measuring Follow-On Innovation, Janet Freilich and Sepehr Shahshahani contribute to this debate by bringing together both a deep understanding of patent law doctrine and precise econometrics research. In so doing, they make important contributions not only to the empirical literature, but also to our current theoretical thinking about the impact of patents on follow-on innovation.

Freilich and Shahshahani’s key empirical contribution is a refinement of the measure to capture the universe of follow-on innovations that are plausibly influenced by a patent. Two dimensions of patent law doctrine impact follow-on innovation. The first, patent breadth, determines how wide a net a specific patent casts—the more expansive the metes and bounds of a patent entitlement, the more follow-on innovations it will ensnare. The second, the collection of patent infringement doctrines, determines which types of activities in fact encroach upon the patent right. Their paper largely cabins the issue of patent breadth by focusing exclusively on measures of patent infringement.

Their article’s core claim is that existing empirical analyses of the impact of patents on follow-on innovation include activities that are non-infringing—and therefore, not the type of “patent related” follow-on innovation that those studies attempt to measure. The effect of this incorrect inclusion is large; in fact, Freilich and Shahshahani conclude that “little follow-on innovation is directly affected by the patent, with implications for theorizing the tradeoff between initial and follow-on innovation.” (P. 4.)

Freilich and Shahshahani focus largely on replicating and refining results in one important paper by Bhaven Sampat and Heidi Williams. In that study, Sampat and Williams measure follow-on innovation on patented and unpatented genes, using creative strategies to eliminate selection biases and concerns about claim scope.1 Freilich and Shahshahani’s refinement consists of eliminating non-infringing activities from the pool of follow-on innovations. Replicating Sampat and Williams’ methodology, they rely on articles published in a scientific journal that discuss research on the patented genes as their measure of follow-on innovation. Specifically, their refinement removes from the pool of follow-on innovations those scientific publications that fall under the following well-defined non-infringement categories: (1) extraterritorial activities; (2) activities by persons affiliated with the state government; (3) activities by persons affiliated with the federal government; (4) activities generating information for submission to the Food and Drug Administration (FDA) under the 35 U.S.C. § 271(e)(1) safe harbor provision; (5) and using or importing results generated from a patented technique.

Sampat and Williams found a small but statistically significant effect of gene patents on follow-on innovation. In light of other studies on gene patents,2 one might have expected that this paper’s correction would have uncovered a larger impact of gene patents on follow-on innovation. But the opposite was in fact the case. While Sampat and Williams found that gene patents had a small negative effect on follow-on innovation, Freilich and Shahshahani’s refinement found an even smaller effect. In short, with the new correction, patenting genes has an almost negligible impact on follow-on innovation.

Their refinement is important in no small part because it requires deep knowledge of patent infringement doctrines, including understanding which infringement doctrines produce predictable results ex ante that are unlikely to lead to litigation. This last point is critical: because the universe of follow-on innovations should include any invention for which a license would reasonably be required, their measure should only exclude those innovations for which we can be certain that a license would not have been sought ex ante. The importance of identifying clearly non-infringing activities also puts pressure on their empirical methodology. In this regard, one particular infringement area—the safe-harbor provision under 271(e)(1)—is worth discussing further.

The experimental use exception proves to be an important filter in their corrected measurement, eliminating 388 publications from the original pool of 2,771 follow-on publications. The 271(e)(1) safe harbor provision is meant to exempt from infringement activities “reasonably related to the development and submission of information”3 to the FDA. Courts have interpreted this provision broadly to include many drug-development activities, including early-stage research. Recent caselaw, as Freilich and Shahshahani point out, excludes two key types of activities from this safe harbor: basic scientific research and patented technology that is not itself subject to federally regulated approval (also known as the research tools exclusion).

To identify publications that fall under the 271(e)(1) exception, the authors use a proxy: whether the publication discussed a potential therapeutic application. This proxy is cleverly designed but both potentially under- and over-inclusive, as the authors recognize. Some exempt research projects may not explicitly address therapeutic applications. On the other hand, it is plausible that scientific publications engaging only in basic scientific research mention therapeutic applications for the purpose of seeming more attractive to funders or publishing venues. The research tool exclusion could also lead to over-inclusivity concerns. Because correctly identifying research that is exempt under 271(e)(1) is likely to be crucial in future studies of follow-on innovation in the biomedical sciences, further refining this measure would be a fruitful avenue of future research.

Freilich and Shahshahani next turn to exploring how their refinement might help reconcile other studies on follow-on innovation—many of which found larger effects of patents on downstream innovation. They begin with an explanation rooted in sociological factors. University researchers, who are likely overrepresented in the sample of follow-on innovations measured by published scientific papers, may be largely unaffected by details of patent law doctrine. This effect can be explained by a robust underlying social norm that fosters the free sharing of information and materials, and a disregard for patent entitlements. The situation might be quite different for communities embedded in biotech start-ups and established pharma companies, where uncertainties about the content of 271(e) and doctrines designed to police claim breadth may lead to choices of research projects away from those that require the use of patented materials.

Their explanation reveals another interesting mechanism worth exploring in future studies on follow-on innovation: the impact of patents is likely to be mediated by communities’ social norms. This makes it crucial to understand those social norms and the likely distinct mechanisms by which communities of university researchers, biotechnology startups, and established pharmaceutical companies (among others) decide which follow-on research projects to pursue and which ones to abandon.

Finally, Freilich and Shahshahani consider how their measure may inform existing theories of innovation. As they recognize, “the proper measure of follow-on innovation depends on the hypothesized mechanism through which a patent might affect downstream innovation.” (P. 29.) Two quite different theoretical perspectives on the role of patents in follow-on innovation have emerged from the literature. The first one conceptualizes patents as increasing incentives for patent holders to invest in follow-on innovation. Under this theory, patents do not diminish follow-on innovation; rather, patents tend to concentrate follow-on research in the hands of fewer patent-holding inventors. In contrast, the second theory emphasizes the access costs that patents impose on follow-on innovators other than the patent holder.

Freilich and Shahshahani’s combination of detailed legal analysis with econometric research opens the door for testing a number of additional questions about the mechanisms of follow-on innovation. For example, one interesting question that emerges from these two theories is whether innovation by multiple parties is likely to be qualitatively different from concentrated innovation by a few patent-holding pioneers. Here, it would be interesting to further parse the authors’ refinement to test whether patents influence the kinds of follow-on innovation that take place. More specifically, do patents change the balance between incremental and breakthrough innovation? Network theories of innovation suggest that breakthrough innovation benefits from knowledge recombination across firm boundaries. By increasing access costs outside the firm, patenting may lead to more incremental and less breakthrough follow-on innovation.

A single paper cannot settle the debate between these two theories of follow-on innovation. But by emphasizing the importance of infringement doctrines in measuring follow-on innovation, and by showing how this measure modifies the experimental results of several prior articles, the authors move us towards a more precise answer, an answer that is likely to vary by industry and to impact basic and applied research in different ways.

Download PDF
  1. Bhaven Sampat & Heidi L. Williams, How Do Patents Affect Follow-On Innovation? Evidence from the Human Genome, 109 Am. Econ. Rev. 203 (2019).
  2. Fiona Murray & Scott Stern, Do Formal Intellectual Property Rights Hinder the Free Flow of Scientific Knowledge?: An Empirical Test of the Anti-Commons Hypothesis, 63 J. Econ. Behav. & Org. 648 (2007).
  3. 35 U.S.C. § 271(e)(1).
Cite as: Laura Pedraza-Fariña, How Do Patents Influence Cumulative Innovation?, JOTWELL (January 24, 2023) (reviewing Janet Freilich & Sepehr Shahshahani, Measuring Follow-On Innovation (Feb. 20, 2022), available at SSRN), https://ip.jotwell.com/how-do-patents-influence-cumulative-innovation/.