In the last 15 years, there has been substantial pushback against the patent system. Critics claim that patent rights are becoming an impediment to innovation rather than an incentive, and an instrument to extract rents through patent litigation. One of the main culprits identified by critics is ineffective screening by the patent office, granting patents to unworthy inventions (Jaffe and Lerner 2004). These issues have also been prominent in public debates (Federal Trade Commission 2011), US Supreme Court decisions (e.g. eBay Inc. v. MercExchange in 2006, and KSR Int'l Co. v. Teleflex Inc. in 2007) and Congressional legislation to reform the patent system, most importantly the Leahy-Smith America Invents Act of 2011. Further, some scholars have called into question the desirability of having a patent system at all, arguing that its costs outweigh its benefits (Boldrin and Levine 2008, Bessen and Meurer 2009).
Of course, the real policy choice is not binary – it is not the current patent system or no system. What we need is a framework that allows us to gauge the severity of the patent-quality problem, to evaluate whether the current system is welfare-enhancing and, most importantly, to assess whether a better-designed patent system would deliver greater welfare gains. In recent research (Schankerman and Schuett 2020), we develop a framework designed to provide quantitative answers to these key policy questions. We focus on four policy instruments, of which three are within the control of the patent office: the intensity of examination, pre-grant (application) fees, and post-grant (renewal) fees. Only the fourth policy instrument – review by the courts of issued patents challenged by a competitor – lies outside the scope of the patent office.
In our model, there are some inventions which would not be developed without the possibility of obtaining a patent while others do not rely on patenting options. This depends on whether the profits the inventor can appropriate without a patent are large enough to cover the development costs of the invention. We refer to the former inventions as high types and to the latter as low types. We assume that development costs and private values are heterogeneous and that the inventor has private information about the type of the invention. This asymmetric information creates a need for screening as patent protection increases the profit for both types and owners of low-type inventions also have a private incentive to seek a patent.
The patent office examines all applications and approves all high types, but screens out low types at a rate that depends on the intensity of the examination.1 If the application is approved, the inventor can offer a license contract to a competitor of the same industry, and the competitor chooses whether to challenge the validity of the patent in court. For most of the analysis, we assume that courts are perfectly able to distinguish low from high type inventions and will hence only uphold high-tpye patents. We use this benchmark to highlight the limitations of relying on courts, even if they never make mistakes.
We calibrate the model by choosing the parameter values that match the equilibrium predictions of the theoretical model to the observed values of the grant rate in the US patent office, the litigation rate and the patentee win rate in the courts, and the R&D and total factor productivity growth per invention, which are constructed using disaggregated sector-level data and other information. This calibration allows us to measure the severity of the patent-quality problem, to assess the value of the patent system, and to evaluate the welfare impact of policy reforms. We then use the model to study two types of reforms: reforms to the patent office fee structure and examination intensity, and reforms to the litigation procedures for patent challenges and the antitrust treatment of patent licensing.
For convenience, Table 1 summarises selective findings from the baseline (current) patent regime and various policy reforms we study. We highlight some of these findings in the following discussion.
Table 1 Some key findings under the current and counterfactual patent regimes
The theoretical analysis and calibration of the model reveal striking findings about the current patent landscape. First, the patent-quality problem is real. Under current patent policy, 57% of patent applications are low-type inventions which, from an economic point of view, should not be granted. The patent office screens out about half of these applications. This implies that two out of five patents are issued on low-type inventions, causing unnecessary social costs. Nonetheless, we find that even under current patent policy, the patent system is welfare-enhancing – total welfare with patent rights net of the welfare from low-type inventions (which would occur anyway) is positive.
The second important finding is that even if courts are perfect, they are not able to fix the problem caused by low-type patents. There are two reasons for this. The first is that litigation is costly, and not all patents are sufficiently valuable to make them worth litigating. Second, we show that even for patents worth litigating, in equilibrium, low types may mimic high types by charging a high license fee or pre-empt challenges by setting a low license fee.2 This is achieved with high royalties and a negative fixed fee from the patentee to the licensee (so-called ‘reverse payments’). As royalties determine deadweight loss, such pre-emptive behaviour causes social harm comparable to low types charging high license fees.
Our quantitative analysis also indicates that most low-type patents escape scrutiny in the courts – only 17% of low-type patents (and 10% of all patents) are sufficiently valuable to even be at risk of a challenge; of those that are, about 64% strategically pre-empt challenges. This result raises serious doubts about over-reliance on the court system to weed out bad patents.
Third, we show that the private incentives to challenge a patent can be either smaller or larger than the social incentives. The conventional wisdom is that the incentives to challenge are insufficient due to the positive externalities from a successful challenge (Choi 2005, Farrell and Shapiro 2008). While this point is valid, it is incomplete both because the challenger takes only his own litigation cost into account and because the private gains from a successful challenge can be larger than the deadweight loss avoided. Our quantitative analysis shows that, under the current patent regime, the net social benefits from challenges are negative: litigation costs far exceed the deadweight loss they eliminate.
Fourth, whether the patent system is socially beneficial or not depends critically on the patent policy design. We show theoretically that a well-designed patent system is welfare-improving relative to no patent system. This result relies on the judicious use of both pre-grant fees and examination intensity. Our quantitative analysis complements this result by showing that the converse is also true: a poorly designed patent system reduces welfare. According to our simulations, a patent system with only registration (i.e. no patent office examination) is associated with 6.7% lower welfare than the current system, and its social value is negative. Thus, it is worse to have a registration system that gives temporary monopoly patent rights to all applicants than it is to have no patent rights at all, even if those (registered) patent rights are subject to review by perfect courts.
Fifth, one design feature which our theory identifies as desirable is to frontload fees, i.e. to rely on pre-grant rather than post-grant fees. This is in contrast to current practices by the major patent offices around the world, which tend to backload fees. The intuition for the optimality of frontloading is that low types have a stronger preference for post-grant fees than high types because low types have a smaller chance of passing examination. Although our simulations show that the impact of frontloading alone is small, a policy that accompanies frontloading with reinvestment of the additional fee revenue it generates into screening (a policy which is revenue-neutral), enables the patent office to raise detection of low types from 50% to 58% and increases welfare by 1.1%. This is still well below the detection rate in the welfare-maximising patent policy, where both examination intensity and fees are optimally set. Such optimal policy would screen out 86% of low types and increase welfare by 3.9%.
Sixth, there is a trade-off between screening and innovation. High fees help deter low-type applications, but they also discourage high-type investment. The optimal policy is a case in point. Optimal fees are about ten times higher than current levels, leading to an 8% drop in high-type innovation. In effect, the optimal policy trades off losses in high-type innovation for savings in the social costs associated with patents. If such large increases in fees seem politically infeasible, it is reassuring that much of the welfare gain from the optimal policy can be achieved through more modest increases in the patent office budget, accompanied by revenue-neutral increases in fees. We find that doubling (tripling) examination resources per application would yield 49% (66%) of the gains from the optimal policy.
For a broader view of the welfare effects of different patent policies, Figure 1 presents a heat map of iso-welfare curves as a function of examination intensity (horizontal axis) and pre-grant fees (vertical axis, logarithmic scale), with post-grant fees set to zero. The coloured bar on the right-hand side of the figure illustrates the welfare level relative to the current system. We depict the optimal patent policy along with the other reforms discussed above, and plot the locus of revenue-neutral policy options. This can be helpful in thinking about politically feasible policy choices (in terms of fees, for example) and understanding the welfare cost of not being more ambitious.
Seventh, we assess the impact of the Patent Trial and Appeal Board (PTAB), which was established by the US Congress as part of the America Invents Act in 2011. The PTAB is an administrative procedure for challenging granted patents within the patent office, which is much cheaper (but less thorough) than a challenge through the courts. It has become one of the most contentious aspects of the patent system. Critics claim that the PTAB has made it too easy to revoke patents – or creates “patent death squads", as former chief judge Rader of the US Court of Appeals for the Federal Circuit colourfully put it. However, we estimate that the PTAB upholds 92% of high type patents, and thus revokes only 8%, while invalidating 82% of low types. Hence, the net effect of the introduction of the PTAB is a 1.6% welfare increase.
Finally, the social value of the patent system is much larger if it is complemented by antitrust limits on licensing. The theory shows that prohibiting negative fixed fees in licensing contracts (so-called reverse payments) lowers the royalty rates charged by low-type inventors and thus reduces prices and deadweight loss, but it also increases the number of patent challenges. In 2013 the US Supreme Court (FTC v Actavis, Inc.) ruled that such ‘reverse payment’ agreements may be illegal under the antitrust laws but, acknowledging this trade-off between litigation and higher prices, applied a rule of reason test to such cases. Our quantitative analysis shows that the net welfare effect of prohibiting negative fixed fees is positive and very large – a 3.8% gain relative to the current patent system without the restriction. This indicates that a tough policy towards reverse payments may be warranted, perhaps an even tougher one than the current rule of reason approach adopted by the Supreme Court in the Actavis case.
The framework developed in this research can be adapted to study other patent reforms – e.g. a loser-pays rule for allocating legal costs and introduction of litigation insurance – and perhaps even calibrated for patent regimes in other countries for a broader assessment of the impact of patent rights.
Bessen, J and M Meurer (2009), Patent Failure: How Judges, Bureaucrats, and Lawyers Put Innovators at Risk, Princeton University Press.
Boldrin, M and D Levine (2008), Against Intellectual Monopoly, Cambridge University Press.
Choi, J P (2005), “Live and Let Live: A Tale of Weak Patents”, Journal of the European Economic Association 3: 724-733.
Farrell, J and C Shapiro (2008), “How Strong Are Weak Patents?” American Economic Review 98: 1347-1369.
Federal Trade Commission (2011), The Evolving IP Marketplace: Aligning Patent Notice and Remedies with Competition.
Jaffe, A B and J Lerner (2004), Innovation and Its Discontents: How Our Broken Patent System is Endangering Innovation and Progress, and What to Do About It, Princeton University Press.
Schankerman, M and F Schuett (2020), “Patent Screening, Innovation, and Welfare,” CEPR Discussion Paper 15301.
1 From an economic perspective, patents should only be issued on high type inventions, for which the cost of developing the idea exceeds the profits the inventor can appropriate without a patent. This requirement is justified on both normative and descriptive grounds. Normatively, this is the patentability requirement that a social planner would choose in our setup since low type inventions would be developed anyway, so giving them patents causes unnecessary social costs. From a descriptive perspective, courts rely on empirical proxies to determine whether or not an invention would have been developed absent a patent. The most important proxy -- non-obviousness -- has been rationalized by courts and legal scholars in exactly this way. In the landmark case Graham v. John Deere (1966), the US Supreme Court stated: “The inherent problem was to develop some means of weeding out those inventions which would not be disclosed or devised but for the inducement of a patent.” This rationale for non-obviousness was reaffirmed in KSR v. Teleflex in 2007. In our baseline model, these legal standards are assumed to be perfect proxies for the underlying economic standard of patentability, but the results are robust when we allow the courts to make errors.
2 It has been alleged that this is what so-called `patent trolls' do (Federal Trade Commission 2011). In our model, trolling behaviour is endogenous and the extent to which it occurs depends on the design of patent policy.