Burdens of Proof for Sexual Misconduct Claims in Senate Confirmations and on College Campuses
How confident must we be that allegations of sexual misconduct are true before we impose consequences on the accused? This question is central to the debate over the confirmation of Supreme Court nominee Brett Kavanaugh; it’s also highly relevant to the Education Department’s effort to write new rules under Title IX for handling sexual misconduct claims on college campuses. The answer to the question will no doubt depend on the severity of the contemplated consequences. Nonetheless, a few observations apply across the different contexts in which similar questions arise.
For anyone who is thinking deeply about this issue, the best place to start is (I think) Louis Kaplow’s 2012 Yale Law Journal article “Burden of Proof.” Kaplow’s article doesn’t directly address sexual misconduct claims, but his analytical framework applies — with a few modifications — to the Kavanaugh confirmation and the Title IX rules. So start there before reading on. (The article is 122 pages, so this probably requires signing off Twitter and closing a few browser windows first.)
One way to approach the question is to consider the error costs on each side. What is the cost of a false negative (i.e., failing to impose consequences when the allegation is in fact true)? What is the cost of a false positive (i.e., imposing consequences when the accused is in fact innocent)? If we think that false positives are a lot worse than false negatives (as in the famous Blackstone formulation, “Better that ten guilty persons escape than that one innocent suffer”), then a higher threshold will be warranted. Hence, the choice in criminal law of a “beyond a reasonable doubt” threshold, sometimes equated to a 95% confidence threshold.
When the consequences are noncriminal, we tend to choose a lower threshold. The “preponderance of the evidence” standard in civil cases roughly translates to a confidence threshold of 50.1% (or really, >50%). Imposing civil liability (e.g., monetary damages) on an innocent defendant doesn’t seem as bad to most of us as putting an innocent person in prison.
In the Kavanaugh case, the cost of a false positive is that a person who otherwise would be on the Supreme Court instead remains a D.C. Circuit judge. Another highly qualified jurist with similarly conservative views would likely take his place. If Senate Republicans can’t confirm a replacement before the midterms or in a lame duck session and then lose their majority in the upper chamber, the result might be a deadlock between Senate Democrats and President Trump, in which case we could have an eight-member Supreme Court for some time. We have had an even number of justices on the Supreme Court at several earlier points in U.S. history, and there’s a decent argument that the optimal number of justices is even rather than odd. Another possibility is that President Trump and Senate Democrats would come together on a consensus pick, which might go some way toward bolstering public confidence in the Court. (Paging Brian Sandoval?)
Two other costs of a false positive merit mention. The first is the possibility that Kavanaugh’s replacement might be much more conservative than Justice Kavanaugh would have been. Obviously, whether one thinks this is an “error cost” or “error benefit” depends upon one’s own ideological views. The smart money is betting on Seventh Circuit judge Amy Coney Barrett if Kavanaugh’s confirmation fails. And a Justice Barrett might be more likely than a Justice Kavanaugh to vote to overturn Roe v. Wade — which, from a progressive perspective, would be a real cost.
Should the risk to Roe matter to our assessment of the Kavanaugh confirmation? One might hope for a confidence threshold in confirmation hearings that stands apart from partisan and ideological considerations, in which case the answer would be no. It’s also not clear how much daylight separates Kavanaugh and Barrett on abortion, especially considering Kavanaugh’s refusal to say anything positive about Roe v. Wade during confirmation hearings beyond acknowledging its existence. And the swing votes in the Senate who profess to be pro-choice — specifically, Republicans Susan Collins of Maine and Lisa Murkowski of Alaska — would be swing votes on Barrett too (at least as long as the composition of the Senate remains as it is now). If they don’t want to see Justice Barrett on the Court, the two of them can probably stop that from happening on their own.
The other relevant error cost is the cost to Brett Kavanaugh himself if he is denied a spot on the Supreme Court based on something that he did not do. Of course, no one has a right to a Supreme Court seat, and hundreds of millions of Americans manage to live happy lives without one. But being denied a Supreme Court seat is different from being denied a Supreme Court seat because the Senate concluded that you committed sexual assault. The reputational consequences of the latter outcome are severe indeed.
Still, I don’t think the reputational-cost factor does all that much to alter the confidence-threshold calculus. First, Susan Collins’s and Lisa Murkowski’s votes on confirmation cannot absolve Judge Kavanaugh in the court of public opinion. Second, reputational costs are endogenous to confidence thresholds. The reputational costs of being found liable by an adjudicator applying a 40% confidence threshold are less severe than the reputational costs of being found liable by an adjudicator applying a 95% confidence threshold. In the former case, the fact that the adjudicator found you liable doesn’t tell me that you actually did it. In the latter case, the adjudicator’s finding probably carries much more reputational weight.
Now let’s look at the other side of the balance: the error costs of not imposing consequences if the allegations are true (i.e., the costs of a false negative). There is the personal cost of Christine Blasey Ford, who now must live with the fact that her assailant is on the highest court in the land — a prospect that apparently led her to consider moving her family to New Zealand. (There is also a reputational cost to Dr. Ford of the Senate disbelieving her, though again, that cost is endogenous to the confidence threshold that the Senate chooses.) There is the cost to the American legal system of elevating a justice who has committed a serious crime and lied about it to the nation. Theoretically there is the risk of recidivism — with the costs to individuals in the justice’s vicinity — though in Judge Kavanaugh’s case, we have no reason to believe that he poses a personal risk to anyone (and as I’ve written before, based on my own observations, Judge Kavanaugh is unusually generous and gracious in his treatment of colleagues, clerks, and staff).
Weighing the error costs on each side, we might conclude that the costs of a false positive are light relative to the costs of a false negative (as, indeed, I think they are). And that might suggest a confidence threshold even lower than 50%. Note that confidence thresholds below 50% are not foreign to the law. For example, the standard for judicial review of the factual findings of an administrative agency is “substantial evidence,” i.e., “such relevant evidence as a reasonable mind might accept as adequate to support a conclusion.” The question is not whether the judge believes that a purported fact is more than 50% likely to be true; the question is whether the judge believes that a reasonable person could believe that a purported fact is more than 50% likely to be true. And, of course, confidence thresholds below 50% are quite common in daily life: how confident do you need to be that the discoloration on your cheese is a malignant mold before you choose a different block?
Yet the analysis is still incomplete. The confidence thresholds that we apply in sexual misconduct cases affect not only the distribution of error costs ex post, but also the likelihood of sexual misconduct ex ante. For example, a 99.9% confidence threshold would likely lead to more cases of sexual assault and harassment because potential wrongdoers would know that it will be very difficult to ever hold them liable. Even a 50.1% confidence threshold might do too little to deter if potential wrongdoers expect there will be little evidence to corroborate a victim’s account.
Lowering the confidence threshold does not, however, always enhance deterrence. That is because the deterrence effect depends on the difference between the consequences of wrongdoing and the consequences of doing right (or doing nothing). Take the example (and I’m borrowing this from Kaplow with a small tweak) of a police department that gives tickets to every car parked in a metered spot without checking whether there is money in the meter (i.e., a confidence threshold of 0%). No one would have any incentive to put money in the meter, because the probability of a ticket doesn’t depend on it.
What behavioral effects might we expect from a below-50% confidence threshold for sexual misconduct claims in the confirmation process (e.g., something like a “substantial evidence” test)? Individuals with even remote aspirations of winning Senate confirmation would probably take additional precautions to ensure that they are not accused of sexual misconduct. They might, for example, choose to adhere to Antioch rules (which, in my view, would be a good thing). They also might curtail heavy drinking, knowing that sexual misconduct claims are more likely to arise in settings where lots of alcohol is consumed. Fear of being wrongfully accused might deter some entirely innocent people from pursuing Senate-confirmed positions, but at the same time, individuals with sexual misconduct in their pasts might self-screen themselves out of those positions as well. We would probably see a shift toward more female nominees, which also seems like a good thing (though, of course, not all cases of sexual misconduct are male-on-female). There is, to be sure, the risk of a boomerang effect when confidence thresholds go too low, but that risk seems rather remote here given the other legal consequences that will continue to attach to sexual assault and harassment.
The analysis is somewhat different in the Title IX context. The costs of false negatives are still quite serious, and likely even more so when the costs are borne by a college student: consider the harm to an 18-year-old sexual assault victim of seeing her or his attacker every day on the way to class or the dining hall. The risk of recidivism is even more salient. But the costs of false positives are grave as well: getting kicked out of college — while not a death sentence — is a fate much worse than remaining a D.C. Circuit judge for life. Symmetry of error costs supplies an argument for the preponderance-of-the-evidence standard, though — curiously — the Trump administration reportedly wants to give campuses the option to set a higher (i.e., more defendant-friendly) “clear and convincing evidence” threshold. One hopes that schools will carefully consider the balance of error costs and behavioral effects before they accept that invitation.