Technological Innovation and Intellectual Property

[This post is the third in a series of posts based on empirical research in a new book, Patent Failure, by James Bessen and Michael Meurer.]

Patents on computer programs, financial processes and business methods have been controversial at least since the 1960s. Surveys regularly find that computer programmers are opposed to patents on software by a wide margin. In what other field is the class of inventors so opposed to patents?

Is there a problem?
Some people contend that there is nothing particularly wrong with patents on software, arguing that “patent thickets” are not preventing innovators from entering software markets. But the latest evidence suggests that patent thickets do, in fact, inhibit startups in software. More important, patent thickets might not be the only or even the most important problem with software patents.

Indeed, our evidence suggests an important problem of another sort: software patents are four times more likely to be litigated than are chemical patents; business methods patents are twelve times more likely to be litigated; finance patents are 49 times more likely. Moreover, the evidence also suggests that these patents have lower values than chemical patents, so these patents are not being litigated more because they are more valuable.

Other people admit that there are problems with software patents, but they suggest that this is only temporary: once judges and patent examiners understand this technology better, once they have become familiar with the prior art, etc., then the uncertainty about these patents will abate and litigation rates will go down. But the evidence shows that after a decade of issuing software patents in large numbers (over 200,000), the probability that a newly-issued software patent will be litigated is continuing to rise.

So it does seem that patents on software and related technologies at least have a particular problem with litigiousness. And this problem is central to the poor performance of the patent system generally. In the previous post we highlighted how litigation costs have substantially outgrown the profits that public firms receive from patents outside of the pharmaceutical and chemical industries. In 1999, 38% of the cost of litigation among public firms arose from lawsuits involving software patents; preliminary data suggest that this share has increased since then. Litigation over software patents is clearly a major factor in the poor performance of the patent system. So in a very real way, the overall performance of the patent system cannot be fixed unless the particular problems of software patents are also fixed.

Why are there so many lawsuits over software patents?
A variety of evidence leads us to conclude that software patents are involved in relatively more litigation because they are more likely to have “fuzzy boundaries.” Statistical evidence shows, for example, that software and business method patents are much more likely than other patents to have their claim construction appealed to the Federal Circuit. Part of this tendency arises because of the nature of the technology and part arises because of the way the courts have treated this technology.

Our reading of the case law convinces us that patent law tolerates too many software claims untethered to any real invention or structure; in such a world clear boundaries are unattainable. When patent claims relate to actual devices or chemical structures, then their meanings can be interpreted by reference to those physical or chemical entities. However, when the words refer to abstract ideas, they are often subject to multiple interpretations and are therefore more ambiguous. For example, many people thought that “point of sale location” (in the famous E-Data patent) was computer industry jargon for that place in a retail store where transactions take place, formerly occupied by a cash register. When the Federal Circuit interpreted this claim, they decided that it referred to any location where an e-commerce transaction might take place, although it is highly unlikely that this is what the inventor had contemplated 17 years earlier. Thus a broadly worded invention for a kiosk in a retail store was read to cover a broad swath of e-commerce. Not surprisingly, this patent generated quite a few lawsuits.

Patent doctrines that might serve to prevent such fuzzy claims have been undermined. For example, the enablement doctrine has historically been used to keep patents from claiming much more than what was actually invented. Unfortunately, as a result of Federal Circuit decisions on software patents during the 1990s, these patents no longer need to provide computer code, a flowchart, nor any detailed description of specific operation in order to be enabled.

Fixing the problem
A lot of people have very strong opinions about how patent policy should or should not change regarding software patents. We wish we had such clarity, but we do not. We are convinced that the current treatment of software patents creates significant problems and that these are getting worse. But the problem is complex and fixing it will likely involve multiple changes in law and institutions.

Certainly, KSR, the recent decision on obviousness, should help and so would a stronger indefiniteness requirement. Additionally, it might help to restore a substantial enablement requirement for software patents so that these patents are restricted to claiming more or less what was actually invented and disclosed.

Possibly, a subject matter test might help. We confess we do not have a rule that cleanly distinguishes inventions using software that should be patentable from abstract processes that should not be patentable. Some people argue that any attempt to proscribe subject matter will only increase uncertainty and encourage avoidance through clever claim drafting. But the evidence suggests that the subject matter tests used following Benson and Flook did not, in fact, encourage excessive litigation during the 1980s, even though there was some evasive claim drafting. Litigation rates for software patents then were about the same as those for all patents. On the other hand, we doubt that a subject matter test by itself would be sufficient to fix the problems of software patents.

A few weeks ago, I wrote about software patent advocates who argue that software patents have done little harm because there is no evidence of patent “thickets” in software. I, too, was skeptical that patent thickets have so far discouraged innovation much in software because most software firms have only recently begun acquiring patents. Instead, I pointed to significant evidence of harm arising elsewhere, especially from litigation, and that the thicket effects appeared to be in other industries.

It turns out I was wrong and so are the people I was criticizing. There is now solid evidence that software patents make innovation harder for startups in software.

Iain Cockburn and Megan MacGarvie have written a paper (“Patents, Thickets and the Financing of Early-Stage Firms: Evidence from the Software Industry”) that does a very thorough and careful job analyzing the effect of patents on new firms in software markets. This is a surprisingly difficult thing to do because so many things might affect a decision to enter a market. For example, one needs to look at specific software markets (e.g., encryption software) rather than at the software industry as a whole because patents are likely to have their strongest effect within a narrow market segment. Although several earlier studies are flawed, Cockburn and MacGarvie do a masterly job, looking at detailed market segments and controlling for lots of possible confounding factors.

The bottom line: They find that the more patents held in a market, the less likely new firms are to enter and the greater the delays those firms face in getting financing. On the other hand, if a prospective entrant holds patents of its own, that increases their odds of entry—that is, defensive patenting improves the odds.

Nevertheless, the result is surprising especially because the level of patenting was quite low during the sample period: only 16% of the firms in the sample ever filed for a patent. Given the recent explosive growth in patenting by software firms, this certainly raises some troubling concerns.

Some researchers have argued recently that worries about software patents are overblown. For example, Martin Campbell-Kelly asserts his thesis that “The anxieties expressed in the early 1990s about the effect patents would have on the software industry have not been realized.” What were these “anxieties”? Campbell-Kelly tells us, “Opponents of software patents argue that patent ‘thickets’ will necessarily impede the flow of new software products.”

Similarly, Rob Merges states boldly that “the predictions of the software patent doubters in the early 1990s have been effectively refuted so far.” He, too, tells us that the patent doubters were concerned about the effect of patent thickets on entry into the software industry. So, too, do Ronald Mann and Michael Noel and Mark Schankerman.

Beating a straw man?

I wonder whether these proponents of software patents are looking for evidence where they are unlikely to find it, at least not today. They seem to look for a surprisingly limited range of effects among a very limited range of software innovators. Then, failing to find evidence of these effects, declare that the critics of software patents are refuted.

First, concerns about software patents have involved much more than worry about software industry entry and these concerns have been held more broadly and much longer than these authors seem to suggest. Major computer companies opposed patents on software in their input to a report by a presidential commission in 1966 and in amici briefs to the Supreme Court in Gottschalk v. Benson in 1972. Major software firms opposed software patents through the mid-1990s (for example in USPTO hearings in 1994). Surveys of software developers in 1992 and 1996 reported that most were opposed to patents (Oz 1998).

Moreover, this opposition has been concerned about more than just patent thickets, including concerns about the lack of prior art, the difficulty of defining the boundaries of software patents, the fear of excessive litigation and other costs of patents. Yet these researchers ignore these issues despite growing evidence of problems in these areas (see the chapter on software patents in my forthcoming book, Patent Failure, with Mike Meurer).

Second, firms outside the software publishing industry obtain most software patents—95% of them (see Bessen and Hunt 2004). Computer and electronics firms acquire the largest shares and these industries are known to have substantial patent thickets. Clearly, researchers cannot draw representative conclusions by examining only a small group of patentees who happen to be in the software publishing industry.

Conversely, software firms acquire relatively few patents at all. Only 20% of venture-backed software startups apply for patents within four years of receiving financing [Mann 2005]. Through the 1990s, the majority of public software firms still did not have patents. Today, although many have begun acquiring patents, many large firms, such as Google, still have very few patents. Clearly, one is going to have a hard time finding adverse affects of software patent ownership among a group of firms who have very few of them.

This may be changing, of course. However, that means that adverse effects of software patents within the software industry, including from patent thickets, may arise in the future, if and when there is a patent thicket in this industry. But a search for effects of a patent thicket in this small group of patentees today is sure to find a misleading null result.

Whatever one thinks of software patents a priori, surely we can do better than to take a myopic view of the potential problems of software patents and of where they are likely to be found. This research seems a lot like arguing that cigarette smoking isn’t bad because it doesn’t cause cirrhosis of the liver or because it helps fight obesity.

Campbell-Kelly, Martin. 2005. “Not All Bad: An Historical Perspective on Software Patents.” Michigan Telecommunications and Technology Law Review, 11(2): 191-248.
Mann, Ronald J. 2005. “Do Patents Facilitate Financing in the Software Industry?” Texas Law Review, 83(4): 961-1030.
Merges, Robert P. 2006. “Patents, Entry and Growth in the Software Industry.” working paper.
Oz, Effy. 1998. “Acceptable protection of software intellectual property: a survey of software developers and lawyers.” Information & Management, 34(3): 161-173.

Charles Babbage is famous as the inventor of the programmable computer. Less well-known, he was also a fierce critic of the British patent system. He wrote that patent law creates “factitious privilages of little value,” where “the most exalted officers of the State in the position of a legalised banditto . . . stab the inventor through the folds of an Act of Parliament and rifle him in the presence of the Lord Chief Justice of England” (Quarterly Review 43, 1830, 333).

One abuse of the patent system is the practice of “submarine patenting”—keeping a patent application secret for a long time and then springing it on an industry that has already invested heavily in the technology. Stuart Graham and David Mowery examine the role of submarine patents in software patenting.
Read the full story »

Evidence of strategic patenting is found with software patents by James Bessen and Robert M. Hunt. Firms in some industries have acquired large numbers of software patents and the investment these firms make in R&D appears to have declined, relatively, rather than to have increased.
Read the summary