Tomorrow, the Supreme Court will hear oral argument in Microsoft v. i4i Limited Partnership, which examines whether the invalidity defense of 35 U.S.C. § 282 must always be proved by clear and convincing evidence. Microsoft has argued that the standard should be a preponderance of the evidence, especially when the prior art used to contest validity is "unconsidered evidence," or evidence not considered by the Patent and Trademark Office (PTO).
Amicus briefs from the legal academic community generally support Microsoft. A brief of 37 law, business, and economics professors (with counsel Mark Lemley) argues that given the constraints on PTO review (e.g., that examiners only have 16-17 hours per patent), the clear and convincing evidence standard is appropriate only when the PTO considered an issue on the record. This echos a 2007 Stanford Law Review article by Lemley and Doug Lichtman, and Lichtman makes the same point in a (tritely titled) New York Times op-ed. Microsoft is also supported by a brief from the William Mitchell Intellectual Property Institute, arguing that the clear and convincing evidence standard is "historically anomalous" and "conflicts with general legal principles," and a brief from digital IP scholar Lee Hollaar, arguing that the heightened standard raises particular problems for software patents. But does the standard of patent validity actually matter?
Perhaps not, argue Hughes Hubbard patent practitioners Etan Chatlynne, Stephen Kenny, and Lucas Watkins in Investigating Patent Law's Presumption of Validity, Part II: An Empirical Analysis of How Unconsidered Evidence and Evidentiary Standards Affect Jury Verdicts. This new paper looks at two data sets of obviousness and novelty challenges: (1) from Graham (1966) to the first Federal Circuit opinion (1982) (when invalidity could be proved by a preponderance of the evidence if you have unconsidered evidence); and (2) from KSR (2008) until October 2010 (under the Federal Circuit's current clear and convincing evidence standard).* In the pre-1982 period, validity challengers using unconsidered evidence succeeded 32% of the time, while those using only considered evidence succeeded only 13% of the time. In the post-2008 period, validity challengers using unconsidered evidence still succeeded 34% of the time. The authors argue that their data "suggest that a shift to a preponderance of the evidence standard will not significantly affect a validity challenger's ability to establish the factual bases of invalidity to a jury"—in other words, that the outcome of Microsoft v. i4i doesn't really matter.
The authors acknowledge that the over 25-year gap between the datasets "surely implicates additional factors such as changes in applicable practice, procedure, search technology and much of the surrounding body of legal rules." Lichtman and Lemley also addressed this point in their Stanford Law Review article: "The lack of empirical evidence on [whether presumptions matter] may be endemic . . . changes in substantive legal rules also change the behavior of parties in deciding which cases to litigate, so it is far from a simple matter to predict how changes in a legal presumption would change actual case outcomes."
It seems unlikely that one could find a clean natural experiment to determine whether presumptions matter—the cases immediately before and after a doctrinal change in the burden of proof might be somewhat helpful, but there are probably not enough such cases to draw any statistically significant conclusions. A controlled experiment, however, could be more illuminating. For example, Dan Kahan showed that legal definitions of rape have relatively little impact on judgments of guilt based on identical sets of facts. A similar result in the patent context would provide additional support to those arguing that the clear and convincing / preponderance distinction does not matter for showing patent invalidity. (If you know of other empirical work about the impact of burdens of proof or other jury instructions, feel free to add links in the comments!)
* The paper doesn't explain how the cases for the datasets were selected, but an inquiry to one of the authors revealed that the pre-1982 cases were "any cases published in a federal reporter between 1966 and 1982 that mentioned that a jury had an opportunity to render a verdict on a patent's validity," and the post-2008 cases were all Federal Circuit cases that were "a final judgment on the merits of invalidity from an appeal of a judgment based on a jury verdict," which is the same dataset from Part I of the study, which showed that at least 74% of those cases "were evidentiary-standard independent."