Patent rights and cumulative innovation: Part II
When courts, libraries, and secrecy laws shift patent rights, what actually happens to follow-on innovation?
The previous post on The Patentist reviewed the theoretical literature on patent rights and cumulative innovation. We discovered that there is no one-size-fits-all patent system. Because technologies and industry structures differ widely across sectors, a policy choice that works well in one setting may be counterproductive in another. While economic models are helpful for clarifying the mechanisms at work and mapping out the trade-offs, they cannot, on their own, tell us how large the effects are in the real world. For that, we need data. This post moves from theory to evidence, delving into the empirical literature.
Ideally, we would randomly assign inventions to a stronger or weaker patent protection (or no protection at all), let time pass, and study differences in follow-on inventions. This experiment is, of course, not possible, and scholars have used a variety of creative methods to explore the question.
Plausibly random patent invalidation
A first line of enquiry has exploited randomness within patent offices or courts. Specifically, Alberto Galasso and Mark Schankerman have exploited the random assignment of judges to three-judge panels at the U.S. Court of Appeals for the Federal Circuit (CAFC), knowing that some judges are more lenient than others. In other words, some patents were upheld because they had a lenient panel, while others weren’t because of a tougher panel, producing random variation in outcomes that enables the study of patent rights (or the lack thereof) on cumulative innovation. The authors find that, on average, the judicial removal of patent rights leads to about a 50% increase in follow-on innovation by other parties (as measured by patent citations), providing causal evidence that patents block cumulative innovation. This blocking effect is highly heterogeneous, concentrated in complex technology fields (like computers, communications, and medical instruments).
In a similar vein, Fabian Gaessler and colleagues have used post-grant opposition at the EPO, exploiting random variation in whether the original granting examiner participates in the opposition proceedings. (They argue and find evidence that the opposed patent is more likely to be upheld if the granting examiner participates.) According to the authors, patent invalidation increases follow-on innovation by 16% on average. Crucially, the invalidation effect is heterogeneous with respect to the value of the original innovation. For low-value patents, invalidation predominantly increases the number of low-value follow-on inventions outside the patentee’s product market. For high-value patents, invalidation primarily increases the number of high-value follow-on inventions in the patentee’s product market. This finding aligns with the prediction that the blocking effect is caused by licensing failure due to excessive transaction costs for low-value patents and rent dissipation for high-value patents.
These two studies focus on a set of patents that were “important” enough for third parties to challenge the patent’s validity—precisely those for which we expect large effects on follow-on innovation. However, only a fraction of patents are challenged, and other research approaches complete the picture.
Looking beyond patentable inventions
A second line of enquiry has focused on the broader blocking potential of patents, looking beyond technological innovation. For instance, Fiona Murray and Scott Stern study whether patent rights hinder the free flow of scientific knowledge. They focus on patent-paper pairs—inventions documented in a scientific article and protected by a patent—and exploit the substantial delay between the scientific publication and the eventual patent grant. Focusing on publications in Nature Biotechnology, they compare the citation rate of the scientific paper in the pre-grant period with that in the post-grant period, using similar “non-patented” papers as a control. They find that the scientific article citation rate declines by approximately 10% to 20% after the patent grant, relative to control papers that are not subject to patent rights.
In follow-up work, Fiona Murray, Scott Stern, and coauthors examined a natural experiment in which NIH agreements reduced IP restrictions and high access costs associated with genetically engineered mice, a key research tool in life sciences. Access was facilitated, among other measures, through a simple, standardized one-page Material Transfer Agreement and an institution-wide license. The mice were further made available through a nonprofit research mouse repository, on an open-access basis. The sudden increase in openness led to a significant boost in follow-on scientific research, as evidenced by a 30% average increase in annual citation rates to the original mouse articles. The authors further find that the increase in cumulative innovation was driven by greater exploration and the entry of new researchers into the field, and suggest that IP restrictions had stifled research diversity.
Bhaven Sampat and Heidi Williams further broadened the analysis, focusing not only on scientific research but also on commercial investment. They compared follow-on research associated with genes claimed in accepted versus rejected patent applications. On average, gene patents have had no quantitatively important effects on follow-on scientific research (as measured by scientific publications) or commercial investments (as captured by clinical trials and diagnostic tests). The authors suggest that cross-firm licensing contracts in this market seem to have operated at least somewhat efficiently to mitigate the blocking effect of patents on downstream research.
Cumulative innovation… by the incumbent
The research and innovation capacity of a single firm is necessarily inferior to the combined capacity of all other firms. This is perhaps why most of the literature has focused on cumulative innovation at large. However, some studies have also examined the effect of patent rights on the patent-owning firm’s cumulative innovation. In a sense, if patent rights are important for protecting a market niche, not securing patent rights means a more competitive market… and less revenue to invest in follow-on research by the incumbent.
Deepak Hegde and colleagues look at exactly that question in the context of startups. They find that receiving a broader patent scope (because a more lenient examiner examined the patent) significantly spurs cumulative innovation for the patent holder. Specifically, each additional granted claim increases subsequent patent applications and grants by about 6% and boosts the quality of follow-on inventions, increasing citations to subsequent patents by 6% per patent.
A study by Ashish Arora and colleagues estimated the causal effect of patent protection on a firm’s commitment to further internal scientific research by analyzing sudden, unanticipated reductions in the effective strength of granted patents, termed ‘priority disclosures.’ This occurred when a competitor’s patent application, filed earlier but disclosed later, unexpectedly became prior art, weakening the focal patent’s protection. The main finding was that firms responded to this decrease in protection by reducing follow-on investments in that specific research trajectory, observing an average 16% drop in internal citations to the associated scientific work.
The role of invention disclosure
So far, we have focused on how the sudden invalidation, restriction, or narrowing of patent rights affects subsequent innovation. An equally important dimension of the patent system is disclosure: simply putting an invention’s technical details in the public domain may itself spur follow-on research. As we explained in a previous post, patent documents must disclose enough detail for examiners to assess novelty, for inventors to define their claims, and for others to avoid infringing the protected technology. But the technical information they contain can also inspire new ideas. A growing body of empirical work shows that policies that enhance disclosure or lower the cost of accessing patent documents tend to stimulate subsequent innovation.
One major quasi-experimental approach leverages the American Inventors Protection Act (AIPA) of 1999, which required U.S. patent applications to be published 18 months after filing. Before AIPA, U.S. patents became public only at grant, so the reform shortened the secrecy period by roughly 1.5 to 2 years. Deepak Hegde and co-authors compared U.S. patents affected by AIPA with their ‘twin’ European patents (which were already published after 18 months). They find that earlier disclosure significantly increased knowledge diffusion: U.S. patents received more and faster follow-on citations, with citations rising by about 15% within ten years and the average delay to first citation falling by roughly 25%. The reform also appears to have reduced duplicative R&D—overlap declined between highly similar patents, and applications were less likely to be abandoned. Finally, firms that were more exposed to lengthy pre-AIPA grant delays increased their R&D spending by almost 4% after the reform. Overall, the benefits of cheaper access to knowledge seem to have outweighed any losses from free-riding, leading to more patenting and higher R&D investment.
Stefano Baruffaldi and Markus Simeth also use AIPA to study how disclosure timing affects knowledge diffusion. They find that earlier disclosure led to a 12% increase in forward citations, indicating stronger knowledge flows, but that this effect was concentrated in technologically proximate domains rather than in geographically close ones. Their conclusion is that faster disclosure does help diffusion, but its benefits depend on inventors’ ability to screen relevant information and on their absorptive capacity.
A second research design, used by Jeff Furman, Markus Nagler, and Martin Watzinger, exploits the gradual expansion of the U.S. Patent and Trademark Depository Library system between 1975 and 1997. Before the internet, these libraries were the only places outside the USPTO where the full technical content of patents was available, so their arrival sharply lowered the cost of accessing patent information in a given region. The authors show that after a patent library opened, local patenting rose by 8–20% relative to similar regions without a library. This increase was driven by knowledge transfer: new patents in those regions increasingly used technical terms that were new locally but already common elsewhere, indicating that the libraries—and, therefore, the patent literature—broadened the knowledge base of local inventors.
A third line of evidence comes from settings where disclosure is suppressed rather than accelerated. In work with Gabriele Pellegrino and Emilio Raiteri, I examine the impact of the U.S. Invention Secrecy Act of 1951, which allows the USPTO to withhold the publication of patent applications deemed detrimental to national security. Using a matching approach, we compare ‘treated’ prior-art patents (cited by an application subject to a secrecy order) with similar ‘control’ patents not linked to secrecy. We find that when an invention is kept secret for at least three years, related patents receive about 41–45% fewer forward citations than the controls—a negative effect that persists even after the secrecy order is lifted. These results complement Dan Gross’s analysis of World War II secrecy orders, which likewise shows that longer secrecy periods depress future citations. Together, these studies provide strong evidence that the patent system’s disclosure requirement actively supports cumulative innovation and that forced secrecy has lasting costs for the development of related knowledge.
Flying half-blind
Despite all the progress in theory and empirics, we are still seeing only part of the picture. Most work treats the decision to patent as given and then studies what happens when patent rights are stronger or weaker, or when disclosure occurs earlier or later. But in practice, firms also choose whether to patent at all or to rely on secrecy. Because the main empirical strategies all condition on inventions that were patented in the first place, they tell us how tweaks to rights affect follow-on innovation for that subset of inventions, not how the patent system compares with a world in which more (or fewer) inventions stay out of the patent record altogether.
A second blind spot concerns what ‘disclosure’ really means. Existing studies mostly treat it as a binary variable: whether the information is public or not, and how early others can see it. In reality, the quality and usability of disclosure vary enormously. Claims can be clear or deliberately vague, descriptions can be technically deep or superficial, and classification and drafting practices can make it easier or harder for others to find and understand relevant patents. We know that greater access to patent documents supports cumulative innovation, but we know far less about what ‘good’ disclosure looks like from the perspective of follow-on R&D across sectors.
Finally, the empirical literature leans heavily on patent citations and other noisy proxies to measure cumulative innovation. Citation practices have changed over time, search tools and examination procedures have evolved, and firms sometimes advise inventors to avoid reading patents altogether to limit legal exposure. This means that simple counts of citations can give a distorted picture of who is building on what, and by how much. Alternative outcome measures—such as products, clinical trials, software releases, or scientific outputs—exist only in a few domains. A key task for future research is, therefore, methodological: building richer, more robust indicators of cumulative innovation that combine multiple data sources and better capture the real economic and scientific value of follow-on works.
If you enjoy evidence-based takes on patents and innovation, join hundreds of readers who receive The Patentist directly in their inbox.
Please cite this post as follows:
de Rassenfosse, G. (2025). Patent rights and cumulative innovation. The Patentist Living Literature Review 10: 1–10. DOI: 10.2139/ssrn.5983416.