Allison, Lemley & Schwartz on Patent Litigation

By Jason Rantanen

In 1998, John Allison and Mark Lemley published a groundbreaking empirical study of patent litigation, Empirical Evidence on the Validity of Litigated Patents.  Allison and Lemley’s focus in that article was on written, final validity decisions by either district courts or the Federal Circuit from 1989 through 1996.  The basic study design philosophy was to look at patent case outcomes; that is, what was the final outcome for patents that were litigated.  That study is still widely cited.

Working with Dave Schwartz, Allison and Lemley recently completed an updated (and much expanded) version of their 1998 study.  The results of that study are being published in several articles, but the one that links most closely with the earlier study is Understanding the Realities of Modern Patent Litigation.

Their undertaking in this project is truly quite impressive.  The authors expanded their scope to all available decisions (not just those that were published in the U.S.P.Q., as in the earlier study) for utility patent infringement suits filed in 2008 and 2009, and personally coded the relevant case information from the docket sheets, district court opinions, briefs, and Federal Circuit decisions for hundreds of cases.  As in the earlier study, the record unit they used were patent cases, with only final decisions for a given patent being counted; in other words, where there was more than one decision in a case, they reported the last final decision on the validity of the patent.  Thus, if there was a final Federal Circuit decision, it superceded a previous district court decision; if there was a remand and subsequent final decision by the district court, it superceded the Federal Circuit ruling, and so on.  Within this set of cases, Schwartz and Lemley coded information about the cases while Allison coded patent-specific information.

Study Findings

The vast majority of patent cases still settle: Out of the roughly 5,000 patent infringement suits filed in 2008 and 2009, only 290 patents went to trial.  All together, less than 10% of the patent suits filed in these years resulted in any merits decision (that is, a summary judgment ruling or trial).  The remainder settled before that point (although, the authors note, there are a small number of cases still pending that may add another 2-3% to the number of suits resulting in a merits decision).

More indefiniteness challenges: One notably change from the 1998 study was the dramatically grater number of indefiniteness challenges.  Out of the 555 summary judgment motions brought on validity issues by either the patent challenger or the patent holder, 176 decided summary judgment motions involved the issue of indefiniteness, a dramatic increase from the 1998 study.  And the authors observed a less substantial, but still noticeable, increase in summary judgment decisions based on patentable subject matter (26), a category that has likely continued to experience growth.  For context, there were 149 decided summary judgment motions on obviousness and 154 on anticipation.

Many arrows in the challenger’s quiver: In terms of the success rates on motions for summary judgment of invalidity, challengers tended to lose on individual issues: with the exception of patentable subject matter challenges, the percentage of successful summary judgment motions of invalidity was 20% or below for each individual issue (102, 103, indefiniteness, enablement, written description).  But, the overall rate of invalidation on summary judgment was 30% and the overall rate at which patent challengers successfully won on invalidity was 42%.  This is an example of what Lemley has called the “fractioning” of patent law; that is, the idea that a patentee must win every issue in a case while the accused infringer must prevail on only one issue; a particular challenge with patent law and its multiple grounds for invalidation.

Overall, patent challengers tend to win: Overall, patent holders tended not to win in cases that went to a definite merits resolution.  Overall, patentees won only about 26% of the time (164/636 definitive merits rulings).  In addition to invalidity, patent holders lost on frequently-brought motions for summary judgment of noninfringement 54% of the time (257/473) and infrequently obtained summary judgment on more rarely-brought motions for summary judgment of infringement (41/128).  Thus, although patentees had a fairly high success rate at trial (winning on 59% of patents when juries made the decision and 64% when the bench did), the 1-2 punch of summary judgment followed by trial meant that most patent cases that went to judicial resolution were resolved in favor of the patent challenger.

I’ve only covered some of the data presented in the paper and barely scratched the surface of the authors’ analysis.  If you’re interested in reading more, the article can be downloaded directly from the Texas Law Review’s website or via ssrn: 

39 thoughts on “Allison, Lemley & Schwartz on Patent Litigation

  1. 8

    In that same period nearly a million applications were filed for patent in the U.S. There’s some kind of perspective there.

    1. 8.1

      ..or more properly put: a lack of one…

      (not withstanding my efforts to ask people to critically think of the rate of litigation using a denominator of all active granted patents)

      Certain posters kept on needling non-posters about “litigation rate” and when the litigation rate (not even in proper perspective taking into account all active patents) is shown to be DROPPING, the discussion quite evidently became “inconvenient.”

  2. 7

    This is an example of what Lemley has called the “fractioning” of patent law; that is, the idea that a patentee must win every issue in a case while the accused infringer must prevail on only one issue;

    Because nowhere else in law do you have to prove all of the elements of a crime or tort.

    Overall, patentees won only about 26% of the time (164/636 definitive merits rulings). In addition to invalidity, patent holders lost on frequently-brought motions for summary judgment of noninfringement 54%

    This seems pretty easy to understand – the office uses BRI, the patentee threatens the public with that scope, and then once the million dollar litigation commences they get to gamble on the actual scope of rights.

    the overall rate at which patent challengers successfully won on invalidity was 42%

    So 42% of the patents that issued are determined invalid after being given a presumption of validity. That’s one hell of a Type II error rate from the office. Weren’t we just talking about this?

    1. 7.1

      Of course, OPQA will tell you that there are Type II errors in only about 4% or so of the allowed cases they review.

      Heckuva job OPQA.

  3. 6

    Re the interesting thread below about estimating patent suit outcomes, and why some patent owners persist all the way to final judicial outcomes that they reportedly lose 3/4 of the time.
    Of course having one’s major product line being seriously threatened by infringing competitors, and seeking an injunction, is a factor in some cases.

    But as to making accurate patent litigation outcomes in advance, the subject article itself notes as to “..differences in patent-litigation outcomes. The characteristics of individual lawyers, clients, and judges seem to matter quite a bit.” [That is not easily quantified.]
    I would also add that seeking an opinion on litigating or continuing litigation from only the law firm standing to gain or lose millions in billing fees, while common, does not make good sense. A second and independent opinion from another good firm makes sense for the same reason that insurance companies often require a second opinion before major surgery.

    1. 6.1

      Of course none of these factors favor suing on a patent more than this study’s confirmation of a more than 90% chance of getting a cash pre-trial litigation settlement from most defendants, which will thereby not not even risk an invalidity or non-infringement decision. As long as one avoids suing one of the relatively few companies that quite regularly refuses to settle.

  4. 5


    Please also consider this article link to as it relates to composition of matter patents and challenges from generic companies. From the article…”Of the 48% of cases litigated to completion (not settlement), the branded firm nearly always wins a suit asserting an active ingredient patent (92%), but usually loses assering secondary patents (32% wins).”

  5. 4

    Let’s just remember that just because Lemley is counting the beans that it doesn’t give him any special insight.

  6. 3

    “This is an example of what Lemley has called the “fractioning” of patent law; that is, the idea that a patentee must win every issue in a case while the accused infringer must prevail on only one issue; a particular challenge with patent law and its multiple grounds for invalidation.”

    Interesting as a theory, but come on, irl all the patentee has to do is poop out something on a page after having invented something and scream “I’m entitled unless you can show otherwise!”. “Winning” isn’t really required. Just not “losing” is sufficient.

  7. 2

    Max, wouldn’t the patentee and the potential infringer both calculate the odds of both invalidity and validity and factor both into their decision to settle? Thus under ‘perfect results’ we would expect 50% wins, not 25% wins.

    In any event I’m not sure that the perfect results is what we would expect. Having litigated on behalf of clients in several courts I find the underlying presumptions in a 50% split somewhat dubious, in that litigation is rarely 1) perfectly predictable and 2) litigants often fail to act rationally and predictable. 3) extraneous factors are often determinative rather than the merits of the matter when it comes to settlement.

    1. 2.1

      That is not taking into consideration the fact that patent litigation is far more expensive for defendants than for patent owners, and they have a much higher evidentiary burden, hence they have far more financial incentive to settle even if they have a good chance of ultimately winning on an invalidity or non-infringement issue.

    2. 2.2

      I think there are other reasons why we shouldn’t expect the “perfect” 50% result here, James. An underlying assumption to that 50% rule is that the cause of action is independent of others. For patent owners, that often isn’t true – losing on invalidity in one case means that the patent (or at least the claims at issue) is worthless against other defendants. As a result, you would expect patent owners to discount their settlement values, to encourage settlement. This in turn means that you’d expect the defendant win rate at trial to go up, since it becomes more advantageous for the defendant to settle the flip-a-coin cases.

      Somebody should write a paper…

      1. 2.2.1

        Good point. This has also encouraged IPR settlements. It can make sense for the patent owner to give the IPR petitioner a cheap license if they can thereby continue to assert their claims against all other defendants. [Assuming the other defendants failed to request joinder in the IPR]. More so than in the parallel litigation itself, because the IPR is much faster that the litigation, and thus decreases the odds of other defendants paying just to get out of litigation discovery and other costs.


          Paul, this only works if the IPR filer is paid off before filing. If they file, there petition sits there on the record. Others can simply copy it and refile themselves. Not so?

  8. 1


    “Overall, patentees won only about 26% of the time (164/636 definitive merits rulings).”

    Coincidence, or what? Suppose that the patent litigation system is working perfectly, that is to say, the only cases that go to full determination on the merits are those that both sides think they ought to win (because all the rest settle). With a 50:50 split on infringement Y/N and on validity Y/N, patent owner ought to win 25% of the time.

    So 26% suggests that the system delivers perfect results. Amazing!

      1. 1.1.1

        Not really. You would expect that both stages would be factored in. So, the 26% probably reflects other factors such as the different burdens on the d and p, and perhaps the changing climate towards some being heavily anti-patent.


          I think you’re right that it only makes sense for both stages to be factored into the decision to go to trial. Assuming rational decision makers (on both sides), assuming that both sides are reasonably able to estimate the probability of a favorable/unfavorable outcome at trial (that’s a weak assumption, I know), and ignoring the fact that this is a repeat game for some plaintiffs (which means that the cost of an adverse outcome goes beyond the present case), then the outcome percentage here should be driven by the ratio between settlement value and the net value of a judgment for the plaintiff. The 26% figure suggests to me that, on average, the parties to these cases are estimating that the net value of a plaintiff judgment in these cases is about 4 times the settlement value, more or less. The cases that go to trial will tend to be those where the pricing of the settlement by the plaintiff is the most aggressive, of course.

    1. 1.2

      Max, another way to say what you said is that Lemley has only confirmed the obvious.

      Now, here’s the twist: if the prevailing party were to always get their attorneys fees as in England, one would not likely sue if the the win rate was only 25%. One would want to shift the overall win rate to better than 50%. Right?

      Now, with that in mind, which system is better for patents, the English system or the US system?

      Perhaps we see a kernel of a argument in Lemley’s article to move to the English system for awarding attorneys fees in order to discourage patent litigation.

      1. 1.2.1


        You act as if litigation per se is bad.

        That is an unfounded assumption – note the very premise that MaxDrei uses: each side believes they ought to win.

        You are not reading close enough here.


          anon, not really suggesting that litigation is bad. But if patentees want the win rate to be above 50%, the win rate on validity and infringement each must be above 75%. I assume that in the US today that if patentees had such a strong case that infringers would ordinarily settle. But if they could push to trial and win on either validity or infringement and then could collect their attorneys fees, they just might not settle if the overall calculus remained in their favor.

          This would seem to argue that the English system is not a good system to the extent it does not promote settlements.


            In the bad old days – an adjudicated infringer was deemed a ‘wrongdoer.’ And then you could pray to a court of equity for equitable relief – after your trial at law. Meh, guess it’s all relative now. Because google is gooood and any speed bump to google is Bad.

            It would be fascinating to see this study repeated for the golden age of patents – see the ‘centennial of the patent system’ – you know Edison, Westinghouse, Tesla, etc. and compare results. Hint: injunctions, disgorgement (all profits) and unjust enrichment – did not stop innovation – electricity and electricity applications marched on into commercial applications. Inventors were celebrated and entrepreneurs and of course Morgan got rich. But the rest of us got electricity, lights, refrigeration, heat, hydropower etc etc. And many if not all these great strides were on patent – and powerful patents indeed. Does anyone stop to think that maybe we have more enforcement actions because – it has historically never been a more financially rewarding time to willfully infringe a patent? I mean, at best, you are going to have to pay a running royalty, even after getting caught willfully infringing. With the downside being loss of marketshare for not infringing – which will be a much greater loss in enterprise valuation.

            Of historical note for you history buffs – under the Taft administration study of potential patent system improvement – something like a reexamination procedure was proposed. Congress and the committee recoiled as to the abject separation of powers issues – and rejected it. One proposal of the Taft commission, however, was to make PTO an independent agency – to divorce her from the political branch and potential political favors inherent therein. Guess we got the worst of both under the AIA – an executive branch court of review to collaterally attack the claims underlying the judgments of the Courts – probably in direct contravention of the 7th amendment to boot. What a world.


              I’m interested in learning more about the Taft study and proposal. Where would be a good place to start?

              Also, do you have any good books or articles to learn more about the “golden” age of patents?



              Nice overview, but your assumption that anyone cares about things like separation of powers or 7th amendment issues today is just absurd.



                How sad is it to compare the subject matter of the great Marbury case (a mere justice of the peace commission) with the apparent lack of care underlying the separation of powers struggle for writing innovation law.

                I believe that the collective legal society should feel great shame for this apathy.


                Yea, that Constitution and Bill of Rights is so like … yesterday. No question current SCOTUS views the patent clause as the red headed step child.



                The Supreme Court cares. Read, e.g., Granfinanciera, SA v. Nordberg, 492 U.S. 33, 109 S. Ct. 2782, 106 L. Ed. 2d 26 (1989). link to

                There is no way that IPRs are constitutional under controlling law.


              iwasthere, do you know who proposed reexaminations, and later IPRs? Yes, the AIPLA and the IPO. Don Dunner, then president of the AIPLA and Judge Newman, then a board member of the IPO .

              What where they thinking?

              They finally got the kinks out so that we almost now have the German system where validity is exclusively determined by the PTAB who, by statute, must be competent scientifically. There is no requirement however that they be competent in the science of the patent. So they too, just as much as judges and juries, must be educated by experts.

              Now, this in theory might be a better system than using lay juries, but using preponderance of evidence and BRI? Huh? Now that combination is lethal.

              Has anybody thought of the possibility that we require juries to have technical degrees when sitting in patent cases? That might be a better approach.


                Well Judge Newman has seen the error of her ways – and is now Vociferously against post judgment review by USPTO. So that’s good. The sick joke is that they left the old policy court statutes on the books for the AIA ‘court.’ The Commissioner herself can sit as judge and appoint the two other panel members!!! Talk about your Hayburn’s Case.

    2. 1.3

      I think your logic is flawed. In the simplistic analysis that cases only go to trial when each side believes they can win, they don’t look separately at the issues of infringement and validity. Rather, they look at the overall case–i.e., will they win. What if the patent owner is 99% they will win on infringement, but thinks there’s only a 25% chance they will prevail on validity? That is effectively the same thing as a one variable outcome, with a 25% chance of winning. The patent owner does not think they ought to win.

      1. 1.3.1

        Right – I discuss this in depth in a working paper here: link to

        Bottom line: there’s no basis for assuming that the Priest-Klein hypothesis on the settlement of disputes applies to individual issues in disputes. (P&K is the basis for the 50% figure that’s thrown around). Also, while the asymmetric stakes move is a viable modification of Priest-Klein, it again applies to disputes, not individual issues.

        (All of this assumes that P-K is actually right and that it applies to patent appeals. Which I’m highly skeptical given that around half of district court merits decisions are appealed).


          Your paper appears to be right on, Jason.
          As you note in your paper, there are a lot of assumptions baked into P&K’s 50% rule, including the assumption of symmetric stakes. You may be familiar with this paper on that topic:
          link to

          As I noted in an earlier comment, I surmise that a major reason for the apparent deviation from the 50% level in patent cases is the asymmetry in outcomes – for the defendant it’s a one-time game, and a pretty straightforward calculation, i.e., projected probability of success at trial times the damage exposure, versus the offered settlement cost. There are usually not a lot of collateral costs for the defendant, such as the reputational damage costs you might have in a malpractice or product liability case. The plaintiff in a patent infringement case has to factor in the potential for lost future benefits from the patent, however, since invalidity (and, to a lesser extent, an adverse claim construction) carries forward to the next negotiation.

Comments are closed.