Saturday, March 19, 2011

Allison, Lemley & Walker: Repeat Patent Litigants

Do repeat patent plaintiffs have stronger patents? Are they more likely to settle to prevent these patents from being invalidated? John Allison (UT Austin Business), Mark Lemley (Stanford Law), and Joshua Walker (Lex Machina, Inc.) empirically address these questions in Patent Quality and Settlement Among Repeat Patent Litigants, which was just published in the March issue of the Georgetown Law Journal.

The study was based on the Stanford IP Litigation Clearinghouse, which links all patent suits since 2000 with their corresponding patents (and is "free to academicians, public interest researchers, judges, policymakers, and the media" and run by Lex Machina). The authors compare the 106 patents that were litigated 8 or more times with 343 patents that were litigated only once. Among their findings:
  • The most-litigated patents do worse in litigation than once-litigated patents (winning 10.7% vs. 47.3% of cases), and are more likely to settle (90.5% vs. 84.0%).
  • Plaintiffs for patents initially owned by small entities are equally likely to settle and much less likely to win than large patent plaintiffs (winning 12.3% vs. 53.1% of cases on the merits).
  • Non-practicing entities (NPEs) (some of which might be considered "trolls") are more likely to own the most-litigated patents, slightly more likely to settle, and much less likely to win than product-producing entities (winning 8% vs. 40% of cases on the merits).
  • Software patents are overrepresented among the most-litigated patents, and owners of software patents are slightly more likely to settle and much less likely to win (winning 12.9% vs. 37.1% of cases on the merits).
Part III explores "possible reasons why the most-litigated patents turn out to be weaker than other-litigated patents, as well as some reasons why the owners of those patents nonetheless seem willing to let the cases go to judgment more often than they should." Some of these possible reasons:
  • The most-litigated patents tend to have more defendants, and more defendants is correlated with fewer plaintiff wins (results are mixed on whether more defendants means less settlement). "But this just pushes the puzzle back a level": why aren't repeat plaintiffs suing fewer defendants or dropping cases against defendants who won't settle?
  • Hidden license agreements with nonparties or higher potential value for the most-litigated patents might affect the stakes, but this cannot be tested empirically.
  • Multiple lawsuits may be evidence of obviousness, as the need for multiple lawsuits may result from "widespread, near-simultaneous invention." But the authors "do not believe this can explain [their] results" because (1) the most-litigated patents tended to be older (predating industry adoption) and (2) patentee loses were rarely based on obviousness.
  • The most-litigated patents may be overclaiming, a hypothesis supported by the higher number of continuation applications filed for these patents (because continuations may be used to change claims over time to cover later-developed technologies).
  • The problem might be software patents or NPE patents, which "are both taking disproportionate resources in patent litigation [while] social benefit from those cases appears to be slight."
The authors conclude by noting that these results challenge "the evidence economists have long relied upon to demonstrate patent value" as well as models of litigation.

For those interested in more details, the papers is clearly organized and well written. I'm generally a fan of empirical work, so I enjoyed this data-driven paper, and I look forward to seeing what others have done with this dataset.


  1. One thing I find interesting in a recent FTC report is the FTC's distinction between "good" NPEs and patent trolls (which it refers to as "PAEs"). Many have long noted that there is a need to differentiate between NPEs such as universities and those other entities who abuse the system through arguably-excessive patent litigation. Distinguishing between those bad actors and other NPEs may be helpful in narrowing the focus and the terms of the debate over patent trolls.

  2. Thanks for the comment! Yes, I agree that lumping all NPEs together is problematic. Allison et al. have a table of "Entity Status Classes" on p. 684, and they note this issue: "Rather than take a position on what, if any, NPEs should be considered trolls, we classify each patent owner and let the reader decide. We do, however, report the results for practicing versus nonpracticing entities . . . . Notably, virtually all of the NPEs in our data set fall into Classes 1 (companies in the business of acquiring and asserting patents from others) and 5 (inventor-owned companies). So our data do not depend at all on things like whether universities, university spin-offs, or intellectual property (IP) subsidiaries count as patent trolls."

  3. With respect to software patents we know that there are several problems. First the courts and the PTOs anti-software stance for years resulted in fewer patents filed in this area which hurts the quality of the patent prior art. Second, the anti-software bias may be clouding the judgment of juries and judges. Third, the failure of many areas of software to develop standardized terms has made it more difficult to examine software patents. Fourth, the PTO’s insistence that examiners of software patents have the same production quota as examiner’s in the mechanical arts has not provided these examiner’s sufficient time to review these cases.
    Note that NPEs are the beginning of a secondary market for research and development that cannot be exploited by the inventor. This will result in more resources being devoted to invention. Since all per capita increases in wealth are the result of increases in our level of technology, this is a social good. Is this social good accounted for by the authors?

  4. Thanks for your questions, Dale! For more on software patents, you might look at pp. 707-09 of the article (linked above). The authors note that even though software patents are "taking disproportionate resources in patent litigation . . . . [i]t is less clear what conclusions we should draw from this fact." They say that their results could be evidence of antisoftware bias, but whatever the reason courts are invalidating software patents, "[s]ociety is spending a large chunk of its patent law resources dealing with what are—for whatever reason—the weakest cases. And patent plaintiffs are pursuing those cases despite the overwhelming odds against them." Although this "should give substantial ammunition to those who argue against software patents," "one could read this evidence as proof that the system is working," and "[t]he truth probably lies somewhere in between."

    Regarding your second question, the authors do not discuss the benefits of NPEs creating a secondary market for R&D because this paper is not an evaluation of the costs and benefits of NPEs—it is an empirical study of the most-litigated patents ("the very patents that economists consider the most valuable"), which finds that they are "astonishingly weak."