For the past three years, a conventional wisdom formed that software patents are (1) numerous, (2) poorly examined, and (3) low quality. This "weak-patent" viewpoint has been the argumentum ad populum for anti-software-patent activists in previous years, and has now become the catchphrase at most every software patent symposium being held around the U.S. and abroad.
Many of the commentators perpetuate the notion that software patents are somehow of a "lesser quality" when compared to other technologies. Despite the USPTO's recent hard-line approach, some maintain that patent quality is getting worse when it comes to software:
"Novelty is supposed to be judged relative to 'prior art,' but it is almost impossible for a single human being to root out all relevant prior art. One result has been a steady reduction in patent quality, with patents of dubious novelty being granted routinely" (Hal Varian, professor of business, economics and information management at the University of California, Berkeley).In fact, the primary motivation behind the Community Patent Review Project is to curb "excessive numbers of low-quality patents" that are presumed to be predominate in the software arts.
"The current patent system has given rise to too many low quality patents being issued, and a growing pattern of assertions of weak patents that threaten to damage productive companies and stifle innovation" (congressional testimony from the Business Software Alliance).
But are software patents, as a class, inferior in quality? Some new research being conducted by John Allison and Ronald Mann from the University of Texas, suggests the answer is "no."
The preliminary research conducted by Allison and Mann empirically analyzed patents held by software firms. Specifically, they examined roughly 34,000 patents held by leading firms that were issued during the five-year period from 1998-2002.
The research addresses the threshold question of exactly what should count as a "software patent." Unlike prior researchers, Allison and Mann examined individual patents to determine which ones involve software inventions. Furthermore, the research makes objective distinctions between the "quality" and "value" of patents. Building on prior research, items such as (1) the number of claims in the patents, (2) the number of prior art references in the patents, (3) the number of forward citations, and (4) the breadth of patent families were analyzed in determining the overall quality (and inherent value) of software patents.
With regard to prior art, Allison and Mann found that the average software patent has 18.5 references cited, while non-software tech patents cite 14.1 references. For non-patent prior art (NPPA), non-software patents cite less than 25% of the NPPA cited in software patents. Moreover, 63.5% of non-software patents cite no NPPA at all, while only 34.5% of software patents fail to cite any NPPA.
With regard to claims, software patents, on average, have more claims than non-software tech patents (22.9 versus 17.8). Software patents also have a slight edge in forward citations as well (1.1 versus 1.0). Also, no discernible difference was determined between patents issued to large entities versus small ones.
In short, the preliminary work of Allison and Mann suggests that the current "concerns" over the quality of software patents, as a class, may be overstated:
In our view, the data are important for two separate reasons. First, the data substantially undermine the traditional story that large firms in the software industry are plagued by a large number of low-quality patents obtained by the smaller firms in the industry. On the contrary, by objective standards, the software patents as a class compare quite favorably to patents that the same firms are obtaining at the same time on non-software inventions. Similarly, the patents obtained by small firms are no worse than the patents of the large firms.
Those findings have in our view twin implications for patent policy. The first is the simplest: they undercut the common suggestions that software patents should be prohibited entirely or should face special hurdles for examination designed to stem the alleged flood of low-quality patents. If in fact there is no flood of low-quality patents, then there is little reason to take aggressive action to respond. The second implication is more speculative, but rests on the idea that the effort of the drafter does not depend substantially on the size or patenting experience of the drafter. To the extent that this is true, our findings undermine the concern that small firms will suffer disproportionately from reforms that raise the bar for patent grants, such as increased examination fees, special procedures for "gold-plated" patents, or additional opportunities for pre-grant opposition. If patent drafting is a routine exercise in which firms of all sizes do a better (or worse) job based on the incentives that the PTO’s procedures present, then this presents a reasonable case for reforms designed to reward applicants that put more effort into their application or who are willing to provide more credible support for their application (as evidenced by a willingness to submit their application to a more onerous process).
Download a draft copy of "The Disputed Quality of Software Patents" here