Fallacy Dispelled: Invalidation Rates of Adjudicated Patents Convey Nothing About Quality of All Issued Patents

About half of the patents adjudicated in district courts and at the Patent Trial and Appeal Board (PTAB) are found invalid. It is shown here, however, that patent invalidity rates found in adjudication will necessarily range around the 50% point regardless of the underlying validity of the stock of all issued patents. Unfortunately, there are those who infer a fallacy from such adjudication statistics: that about half of all issued patents are also invalid.

Alarms should be raised when the U.S. Government Accountability Office (GAO) buys into this fallacy and adopts the rate at which the PTAB finds adjudicated patents invalid as a quality measure of all patents issued by the U.S. Patent and Trademark Office (USPTO). This is alarming particularly because some members of Congress have been misled by the GAO on this issue in recent hearings on the USPTO. Congress and the public have long relied on the GAO as an unimpeachable source of valid, unbiased, and reliable information to ensure the accountability of the federal government. When the GAO discards as unreliable the USPTO empirical quality data and relies instead on such fallacies to render its biased inferences, conclusions, and recommendations, it risks losing its credibility and the appearance of impartiality and objectivity.

Unfortunately, the GAO issued two reports that adopted this fallacious quality measure. These reports are titled “Information on Third-Party Funding of Patent Litigation,” issued December 5, 2024, and “Patent Office Should Strengthen Its Efforts to Address Persistent Examination and Quality Challenges,” issued on May 7, 2025. Therefore, it is now time to dispel this pervasive fallacy, and to that end, I filed with the GAO a formal request for correction of these two GAO reports. In the request, I explain in detail the critical errors of analysis, omission, and biased conclusions that undermine the reports’ objectivity and credibility. In this post, I address only one aspect of the reports’ infirmity — the patent quality fallacy.

Patent Validity is Adjudicated Only in a Special Selection of Disputes for Litigation

A patent holder’s loss rate on validity determinations at the PTAB or the district courts varies around the 50% rate, which has virtually nothing to do with the underlying actual quality of issued patents generally. Nor does this high invalidation rate in adjudication reflect the higher economic value of litigated patents, nor the patent challengers’ disproportionate resources amassed to invalidate patents. Rather, the reasons and factors that govern the outcomes of patent validity adjudications are solely reflective of the litigants’ divergent expectations in “close calls” that result in selection of only those special cases for litigation.

As in other adjudicated legal disputes, patent litigation normally arises and proceeds only when the parties do not settle their dispute based on rational evaluation of the quality of their respective positions. The Priest-Klein economic theory predicts that cases reaching trials cannot be, and are not, a random sample from the full range of possible probabilities of success of any given party. They are rather a narrow selection from those “close call” cases in which the uncertain merits indicate that opposing litigants each have about 50% prior probability of prevailing. Other cases in which both litigants after further investigation come to similar estimate of the likely outcome (on either side) are normally settled — they are not “selected” for litigation.

Priest and Klein show that plaintiff’s success rate in civil trials is invariant under the legal provisions, the deciding tribunal, the standard of the decision, or the underlying facts in dispute. Actual plaintiff success rate over an ensemble of multiple adjudicated cases can deviate up to 25% either way from the 50% point due to asymmetry between the opposing parties in prior information available to them, or their stakes in the outcome. This unique characteristic of cases selected for litigation has been empirically shown for essentially all categories of civil litigation. This is shown in Table 1 of my request for correction, showing that plaintiffs won in 54.8% of about 40,000 civil trials. For example, the empirical data in the table indicate that plaintiffs in labor law cases won in 49.2% of these adjudicated cases. Clearly, one cannot infer from this data that about half of all employment agreements were violated. Such inference would be clearly wrong, as are the inferences the GAO makes from the loss rates of patent holders in adjudicated patent cases.

The empirical results in patent cases follow similar patterns and variations. This conclusion also applies to post-issuance cases selected for litigation at the PTAB, where the challenged patents largely mirror the same patents in cases selected for litigation in District Courts.

About Half of Adjudicated Patents Will Be Consistently Found Invalid Even When Only a Few Percent of All Issued Patents May Actually Be Invalid

The figure below demonstrates a simplification of the Priest-Klein model in the context of patent litigation. For the sake of discussion, assume that 6% of issued patents are likely invalid, as depicted schematically in part (a) of the figure. Here, Q(x) is the ordered distribution of the prior probability that patents in that population percentile x (horizontal axis) are valid. Prior uncertainty of the validity of some patents is shown by the gray zone labeled “close calls,” which is expanded horizontally in the view of part (b).

Around the time that a patent from this large population is asserted or challenged, the holder of the patent and its challenger will each conduct a detailed investigation of the merits of their respective cases to refine and form their own respective perceived probability of their success. The divergence of their respective expectations of the outcome would typically arise only in uncertain cases in the narrow “close calls” range.

Part (b) in the figure schematically plots two additional probability curves, showing the probability of success in litigation as perceived by the patent challenger and by the patent holder respectively. These are shown as smooth curves, but in reality, these curves are fuzzy, as for each hypothetical patent in dispute along the x axis, the perceived probabilities may slightly vary based on differing specific information available to the parties. However, the plots capture the overall general trend with respect to the magnitude of the prior probability of validity Q(x). As shown in part (b), cases with disputed patents for which the patent holder and its challenger similarly perceive the likelihood of validity will be settled and not litigated. That is, when patents are likely valid, the probability of success as perceived by the patent holder is high, and low as perceived by the challenger. When patents are likely invalid, the probability of success as perceived by the patent holder is low, and high as perceived by the patent challenger. In both scenarios, there would be no appreciable divergence of expectations of the parties as to the outcome of the litigation and they will settle. In the former scenario, when patents are likely valid, settlement will perhaps involve the patent challenger paying restitution and/or taking a license; and in the latter scenario, when patents are likely invalid, the patent holder will perhaps disclaim certain patent claims or agree to withdraw the assertion of the patent. Therefore, the validity of the patents in both scenarios and their numbers will never be reflected in adjudications.

The result of this selection is shown in part (b) of the figure: disputes are selected for litigation only in “close call” cases where the parties’ internal assessment after further investigation diverge on the likely outcome — where there is substantial overlap of relatively high probability of success that each opposing party perceives. As I have shown elsewhere, that overlap has been for extremely small fraction of patent cases for over a century, during which less than two patents per 1,000 patents in force were selected for litigation. Upon commencement of litigation in those rare cases, subsequent discovery and briefing will narrow down the disagreement gap on the likely outcome, where it has been shown that only about 5.5% of litigated patent cases are ultimately adjudicated on the merits. This means that only about 1 out of 10,000 patents in force were adjudicated.

It follows that such narrow selection within these “close call” cases means that the validity of adjudicated patents has the highest prior uncertainty — near 50% prior probability of validity, as shown by the value of Q(x) in part (b) of the figure. Because such patent disputes also involve other legal and factual factors such as questions of infringement, asymmetries in the information available to the parties, or their stakes in the outcome, patent invalidation results may deviate from the 50% validity rate results but nevertheless will remain substantially high.

In conclusion, the patent invalidation rate in adjudication will be high (around the 50% point). However, that rate essentially conveys no information about the relative fraction of all issued patent percentiles below the “close calls” gray zone that are likely invalid, compared to those above this zone that are likely valid. This example shows that despite the fact that only 6% of issued patents are assumed invalid, about 50% of adjudicated patents will be found invalid. Simply stated, patents selected for adjudication are specifically from the transition zone between two distinct patent quality populations of unknown relative size. The GAO erroneously inferred their relative size from adjudication outcomes for patents in the uncertainty transition zone. This is a fatal error that dooms these conclusions to the dustbin.

Because a patent claim is a legal instrument defining legal boundaries of exclusivity based on the written words, it is essentially equivalent to a (private) statute. Just as disputes on statutory interpretation are central in other litigation under law, so do the majority of litigated patent disputes turn on claim interpretation, which is a question of law. Indeed, research shows that only 11% of litigated patent cases pass beyond the preliminary Markman claim construction adjudication phase to continue on to adjudication on the merits. Therefore, patent litigation often centers on matters on which USPTO examination could have had no effect. This includes questions whether claim terms are to be interpreted as covering after-arising technology, which no “thorough and rigorous” examination at the USPTO could have divined at the time of examination.

Any suggestion that USPTO patent examination quality is a controlling factor in the high rate of patent invalidation in the district courts and at the PTAB is a fallacy that should be rejected and excised with prejudice from all discourse on USPTO examination quality.

See the detailed explanation of this and other GAO errors in my formal request for correction.