What a delight it has been to receive comments on my lead essay from such excellent scholars! I am grateful to all three respondents for their thoughtful remarks.
The debate seems to have split into two streams—Zorina Khan disputes the premises of my argument for reform, while John Duffy and Christina Mulligan have discussed instead whether one of my proposed reforms is appropriate. I will therefore address these points separately and in turn.
Is the Explosion of Patents a Myth?
Zorina Khan begins her essay with a claim to which I object: She says that the U.S. Constitution “bestows secure property rights on patents.” I don’t believe this is a fair reading of the plain text of the Constitution, which states that “The Congress shall have Power To… promote the Progress of Science and useful Arts, by securing for limited Times to Authors and Inventors the exclusive Right to their respective Writings and Discoveries.” Congress has the power, therefore, but not the obligation, to set up a patent system. It is well within Congress’s constitutional authority to abolish the patent system altogether, or (as I have proposed instead) to limit the patent system’s scope to non-software inventions. From a legal perspective, there is not, as Khan claims, a special burden of proof that falls on those who seek a less expansive patent system.
What is the best way to evaluate whether there has been an explosion in patenting activity after the creation of the Federal Circuit? In my essay, I cited econometric studies that found structural breaks in court rulings and patent applications consistent with such an effect. In addition, I showed visually the raw increase in the number of patents granted by the USPTO over time, which displays a startling increase shortly after the Federal Circuit was created. Khan takes raw patent filing data, makes two adjustments—controlling for population growth, and separating out foreign sources of patents—and shows that per-capita American-originated patent filings have increased much less dramatically (although they are now at an all-time high).
Do these data adjustments make sense? If what matters is the number of restrictions imposed on individuals and businesses bringing products to market, they do not. If there are now four times as many patents being granted annually, that’s four times as many ways businesses can get dragged into a lawsuit. It certainly doesn’t matter whether the holder of a patent is American or not; American businesses can be sued by foreign plaintiffs in American courts. And even if we did want to adjust for population, since 1982, U.S. population has increased by about 37 percent, whereas the number of patents granted has more than quadrupled—adjusting for population doesn’t seriously affect my argument.
Neither population growth nor increases in foreign patenting can explain why Bronwyn Hall found a statistically significant structural break in patent applications in the early 1980s. Population grows pretty smoothly, and the increase in foreign patent applications was underway before the Federal Circuit was created. In addition, Khan’s data adjustments say nothing about the statistically significant effects on patent rulings that were discovered by Matthew Henry and John Turner. Wouldn’t it be odd, after the law became more friendly to patent interests, for patent applications not to increase?
Khan also casts doubt on whether there has been an explosion in patent litigation. The number of lawsuits per patent granted, she argues, has been roughly constant (although, again, they are at a local maximum). This data underscores my point, however. If patents are now being issued at over four times the rate that they were before the Federal Circuit was created, and the rate at which they are being litigated is roughly constant, then there are now approximately four times as many patent lawsuits as there were before the Federal Circuit came into existence. That is a large increase. Furthermore, this data does not capture what many firms have been complaining about—threatened lawsuits, often with little legal basis, that nevertheless result in settlements.
Finally, Khan is mischaracterizing my comments on patent “trolls” by implying that I believe that patent rights should vary depending on “the identity of the patentee.” While it is true that some patent reformers have suggested we need special laws that apply to non-practicing entities, that is not my view. There is nothing illegitimate about the secondary market in patents per se. Rather, as I said in my essay, the problem arises when some parties (including non-practicing entities) exploit asymmetries in the legal system. The troll problem should be alleviated by removing those asymmetries, for example by adopting a loser-pays rule and by ruling on the validity of the patent at issue before costly trial preparations begin.
Sometimes You Need a Blunderbuss
John Duffy and I agree on much about the patent system’s problems, so I am grateful to him for homing in on a remaining area of disagreement: whether it is desirable to eliminate a class of particularly problematic patents by subject matter, or whether we should instead rely on the enforcement of existing legal doctrines to eliminate the worst patents individually. I have always thought that Duffy’s proposed solution—stricter application of the non-obviousness standard to software patents—has a certain elegance. While I have some sympathy for his view, I remain convinced that eliminating the class would be preferable.
Christina Mulligan has already articulated an important response to Duffy; I would have expressed roughly the same concerns, only not as well. One of the key insights of public choice is that elegant academic ideas do not always work well when they meet the reality of interest group politics. If we are designing policy for this messy world, sometimes policy will have to be somewhat less tidy than we might otherwise prefer. Mulligan’s essay on rules versus standards provides an excellent way to think about these issues; please consider her arguments incorporated by reference.
Duffy calls the statutory elimination of software patents a “blunderbuss solution.” To an extent, I accept this characterization, but it is nevertheless worth asking exactly how imprecise this particular blunderbuss would be compared to Duffy’s more surgical response. The non-obviousness standard in § 103 says that one cannot patent an innovation that would have been obvious to “a person having ordinary skill in the art.” For software, the relevant art is programming a general-purpose computer. Someone with ordinary skill in this art is capable of quite a lot—the whole point of using a general-purpose computer is that one can compute anything at all that is computable. Would Duffy agree that rigorous application of the non-obviousness standard would eliminate, if not all, then almost all software patents? Insofar as our solutions yield similar invalidations in practice, my blunderbuss starts to look somewhat less haphazard.
Evaluation of policy options must also contain a cost element. If we agree that there are many low-quality software patents that should be invalidated, what is the cheapest way to invalidate them? Duffy’s solution requires socially costly litigation to determine whether particular software patents represent a non-obvious invention. Even granting that this would not require literally every software patent to be litigated for obviousness, it seems likely that it would result in thousands of trials. Whatever the drawbacks of simply abolishing software patents, it has the virtue of economizing on litigation costs. These cost savings must of course be weighed against the supposed gains of Duffy’s less aggressive standard, but they are significant.
It is worth speculating about exactly which software innovations would survive a more rigorous non-obviousness standard. I can think of a number of important, world-changing software innovations that are arguably non-obvious: for example, the invention of TCP, which created a virtual circuit over a packet-switched network, or the creation of the World Wide Web, which married (pre-existing) hypertext pages with the (pre-existing) Internet. For the sake of argument, assume these were non-obvious inventions. What is interesting about these innovations is not that they were not patented (patents on software were not generally allowed at the time these inventions were made), but that copyrights on the software implementations were not even enforced by the creators. Copyrights and patents are not the same thing, but we might consider the stance of an inventor toward the copyright on her innovative software as a proxy for the extent to which that inventor is motivated by directly monetizing her invention. In the case of at least these non-obvious software innovations, direct monetization does not appear to have been the goal.
Similarly, the most non-obvious consequential software invention of the post-State Street era is arguably Bitcoin. Skilled programmers had been trying for decades to create decentralized digital cash. Yet when Satoshi Nakamoto finally invented cryptocurrency, it was released under a non-restrictive license and not patented. Silicon Valley is now investing many millions of dollars in Bitcoin-related businesses, despite the currency’s tender age.
Admittedly, this is anecdotal evidence, but it is striking: to my eyes, the most consequential innovations that might remain patentable under Duffy’s proposal but would be ineligible for patent protection under mine do not appear to have been motivated by direct monetization. It’s possible that this pattern might not hold up in the future, but there seems to be something fundamentally generative about software that makes the most important innovations something that inventors want to share. Combined with the arguments above, this makes me skeptical that we would actually lose anything significant by abolishing software patents altogether. I am eager to understand why Duffy disagrees with this analysis.