Do the costs of U.S. patents exceed their benefits for publicly traded firms? This is the claim of John Turner, James Bessen, Peter Neuhäusler, and Jonathan Williams in their newly posted working paper, The Costs and Benefits of United States Patents. Bessen's prior work has very strong supporters and detractors, and I suspect the response to this paper will be similarly polarized. But I think it is worth understanding precisely what this paper is attempting to measure—and why, even if Turner et al. are correct that the costs they measure exceed the benefits they measure, this does not imply that we should scrap the patent system (though it may be cause for concern).
This study does not attempt to measure the full social benefits and costs of patents—rather, it measures their private benefit to public firm patentees and cost to those public firms involved in patent litigation. To illustrate the difference, consider the following hypothetical: A firm has an idea for a project that has an expected social value of $10 million (including the value of the resulting knowledge for related or follow-on projects), and an expected cost of $2 million. The net social value of having the project go forward is thus $8 million. Suppose that in a world without patents (or R&D tax incentives!), the firm's expected private value (from first-mover advantage) is $1 million, which is insufficient to offset the $2 million cost (boo!). But the firm can get a patent that raises the expected private reward to $3 million, and the firm now pursues the project (yay!).
That $3 million reward comes from supracompetitive prices on the resulting products (say, $1 million) and from license fees or litigation damages paid by other firms (say, $2 million in damages from a patent suit). One could think of this $3 million as a "benefit" (for the firm) or a "cost" (for consumers and other firms), though from an efficiency perspective, it's neither: it is simply a wealth transfer. The real social benefit of the patent system in this hypothetical is that it enabled the project to go forward, creating a social gain of $8 million. The real social costs are the litigation fees from the suit for damages (say, $1.5 million) plus other transaction costs, administrative costs, the costs of duplicative "racing" by other researchers, and the deadweight loss from the supracompetitive prices. Note that even if all of these social costs total $4 million (more than the $3 million private benefit!), the patent system is still welfare enhancing because these costs are less than the $8 million social benefit.
So what are Turner et al. attempting to measure? The "benefit" they try to measure is total annual patent rents for public firms—i.e., the $3 million private benefit to the patentee. And the "cost" they try to measure is the total cost due to litigation for public firms that are alleged infringers, which includes damages (transfers to the patentee that are part of that same $3 million) and litigation costs—i.e., the $2 million in damages plus the $1.5 million in litigation costs, for a total of $3.5 million. Thus, for this stylized hypothetical, the costs are greater than the benefits (as defined by Turner et al.), even though the social costs are less than the social benefit. Their conclusion that "private costs exceed private benefits overall" (for public firms) thus does not imply that social costs exceed social benefits. Nor does it imply that patents don't incentivize innovation, since the litigation costs focus on alleged infringers, not the innovating patentees (rather, rising patent damage awards could well increase innovation incentives). And it also doesn't mean that all private costs exceed all private benefits, since this study focuses on publicly-held firms.
But this does not mean that the study is wrong or that their estimates are not important—it only means that readers should keep in mind what, exactly, they are estimating. Estimating total annual patent rents is still very valuable, such as for comparing the amount we spend on patents to the amount we spend on grants or tax incentives, and there is too little work on this. In Patent Failure, Bessen & Meurer estimated patent rents for all US patents based on renewal rates (an approach pioneered by Pakes & Schankerman); in this study, the authors use the stock-market value of patent-holding companies to estimate rents for public firms. Both approaches have potential drawbacks, so having an additional estimate to compare to other studies is useful.
Estimating the costs to firms due to patent litigation is also both difficult and valuable. Turner et al. use an event-study approach, looking at the stock market response to the initiation of litigation. As they note, the "cumulative abnormal return" they measure "should reflect expected litigation costs, which are pure social losses, and damages, which are transfers." Ideally, we would like to separate the two. Losses to firms in patent litigation may also reflect transfers to consumers due to the weakened market power, which are also not social losses. Event studies are a common economics and finance technique for measuring costs that are otherwise difficult to capture (though their use has been criticized by Glynn Lunney (Hein subscription required), an efficient-capital-markets-hypothesis skeptic who thinks the technique always leads to overestimation of costs). I'm not going to delve into the methodology here—my point is simply that I think it is worth having more discussion about these results on their merits, with an understanding of what Turner et al. are attempting to measure.
Finally, it is worth noting that the numbers in my hypothetical were created to starkly illustrate the distinction between Turner et al.'s costs and benefits and social costs and benefits. If you think patents do allow inventors to appropriate most of the social benefit of their inventions, then the social benefit won't be much more than patent rents. And if the costs due to litigation are mostly waste (e.g., legal fees) rather than transfers (e.g., damages to the patentee or gains in consumer welfare), then the rapid rise in costs is more alarming. Turner et al.'s results imply that overall, public firms do not have positive incentives due to patents (even though individual public firms or private firms might have significant net benefits). This isn't necessarily a problem, but it's a conclusion that may surprise some readers. In any case, I appreciate the effort to help quantify certain aspects of the patent system, and I'd love to see more work along these lines!