In my first two posts, I described how a property system is supposed to work and compared it to the actual performance of the patent system. I concluded that the patent system seems to work reasonable well as property for the pharmaceutical industry, while it fails miserably for other technologies.
So what went wrong? Patent law is a mind-numbingly esoteric subject, and one I'm still learning about myself, so I'm not going to try to attempt a definitive answer in a blog post, but let me make a few general observations.
I largely agree with Bessen and Meurer's description of the general problem: the patent system does an inadequate job of providing notice of patent boundaries. Real property has a variety of mechanisms—fences, no trespassing signs, regular property lines, records on file with the county—that make it easy for someone to figure out when he might be trespassing and with whom he needs to negotiate. Patent law has few, if any, comparable mechanisms. If a smart inventor gets an idea for a new (non-chemical) product and wants to find out whether it's already covered by patents, there's no practical way for him to do that. For any given product, there will be thousands of patents that are potentially relevant—one estimate says that the typical e-commerce site would need to check more than 11,000 patents, for example—and a patent lawyer will charge several hundred dollars per patent to do the necessary checks. Even if an aspiring entrepreneur managed to raise the several million dollars it would take to clear all the patents related to a new product, that wouldn't give him any real assurances because the opinion of any given patent lawyer isn't legally binding. The patent lawyer might tell the inventor that a given patent doesn't infringe, only to get sued and discover that the judge or jury disagrees with the patent lawyer. And of course even an exhaustive clearance effort wouldn't catch submarine patents (which don't surface until after the entrepreneur has started selling his product) or patents in other fields that are interpreted so broadly as to cover inventions far afield from the original patented invention.
A good example of this latter case is Patent 4,528,643, "System for reproducing information in material objects at a point of sale." The invention described by this patent is nominally a mall kiosk that can sell on-demand tapes of music stored at a remote location. "Point of sale" is retail jargon for computerized cash register, and almost everyone assumed that's what the patent covered. However, the owner of the patent, E-Data, began asserting the patent against e-commerce sites that didn't have "point of sale locations" in the conventional sense, and the court bought this broader interpretation. Suddenly, a patent for a mall kiosk became a monopoly over a broad swath of e-commerce technologies. Prior to the court cases regarding the patent, most patent lawyers would have told a potential e-commerce developer that the patent didn't apply to them. But they would have been wrong, and the developer could have faced millions of dollars in royalties.
As a result, in many high-tech fields, especially software, it's taken for granted that any non-trivial product infringes numerous patents, and that finding all the relevant patents is effectively impossible. So startups' standard strategy is to build their product without worrying about what patents they might be infringing, and hope to grow fast enough that they'll be able to hire good patent lawyers when the inevitable patent lawsuits arrive. Once the startups have some free capital, they begin a process of patent stockpiling, attempting to amass enough patents that they'll have some leverage against adversaries in patent litigation. Companies that fail to do this—that spend all their resources on engineers rather than patent lawyers—wind up in the unfortunate situation of Vonage. Vonage, the Internet telephony pioneer, was sued by Verizon, a far less innovative company by almost any measure, because Verizon had spent resources amassing patents while Vonage had focused on actually developing useful products. Vonage ultimately had to pay Verizon $120 million in damages. No one disputes that Vonage developed its technology independently, but independent invention isn't a defense to claims of patent infringement.
Bessen and Meurer's notice theory also explains the relatively better record of chemical patents. Chemical formulas provide a reasonably objective way to distinguish different inventions. Chemical and pharmaceutical patents can be indexed by molecular formula, and it's then reasonably easy to conduct a search to see if a particular chemical has been patented. There's no comparable way of describing inventions in most other fields, so the boundaries tend to be a lot less clear.
So what should be done? There are three broad reforms that I think could make a difference. First, the courts need to be much stricter about allowing patents on abstract concepts. Patents should cover a specific physical device, not a broad category of functionality. Where patent claims are broad or vague, they should either be given the narrowest possible interpretation, or they should be declared invalid altogether.
Second, the patent office should institute reforms to better define the boundaries of patents, and the courts should defer more to these determinations. Currently, the US Court of Appeals for the Federal Circuit, which oversees patent appeals, has a tendency to re-interpret the scope of patent claims without regard for the interpretations given by the lower courts and the Patent Office. Patent examiners should carefully record, in detail, what a patent covers, and then courts should defer to these descriptions. Bessen and Meurer also recommend that the Patent Office offer a service, for a few hundred dollars, of providing opinion letters on whether a given patent infringes a given invention. Unlike an ordinary legal opinion, a Patent Office letter would be giving weight in any future litigation. This seems like a good idea.
Finally, the Federal Circuit should end its unfortunate experiment with patents on software. Supreme Court precedents disallowed patents on software until the 1990s, and the software industry was plenty innovative. A series of decisions by the Federal Circuit eviscerated the rule against patents on software. Bessen and Meurer's statistics indicate that software patents have the lowest value and the highest litigation rates, and I don't think it's a coincidence that the surge in patent litigation corresponded with the legalization of software patents (although those certainly weren't the only bad decisions the Federal Circuit made in the 1990s). The Supreme Court has yet to rule on the Federal Circuit's de facto legalization of patents and software, and in my view they ought to reverse it at the first opportunity.
To return to the original theme of this series, I hope it's clear why I'm uncomfortable with the analogy between patents and traditional property rights. Outside the pharmaceutical industry—and especially in the software industry—the patent system is more like a rent-seeking operation for the benefit of the patent bar than it is a functioning system of property rights. It's possible that the reforms I suggest above (and others Bessen and Meurer propose in their book) could improve things sufficiently that patents will work as property. But until that happens, I think it's a category error to regard patents as a type of property right.