The Washington PostDemocracy Dies in Darkness

Opinion How a dubious statistic convinced U.S. courts to approve of indefinite detention


In the 2002 case McKune v. Lile, the Supreme Court upheld a Kansas law that imposed harsher sentences on sex offenders who declined to participate in a prison rehab program. The substance of the Kansas law the court upheld isn’t as important as the language the court used to uphold it. In his opinion, Justice Anthony Kennedy reasoned that they pose “such a frightening and high risk of recidivism” which he wrote “has been estimated to be as high as 80%.” Five year earlier, in Kansas v. Hendricks, the court allowed the states to detain sex offenders found to have a “mental abnormality” can be continued to be held indefinitely under civil commitment laws, even after they’ve served their sentences. The majority justified its decision by explaining that commitment hearings are administrative, not criminal, and the intent of such laws is treatment, not punishment.

The year after the McKune decision, the court then upheld an Alaska law that puts sex offenders on a public registry, even those who were convicted before the law was passed, which would seem to be a violation of the Constitution’s prohibition of ex post facto laws. Here too, Justice Kennedy noted that “a conviction for a sex offense provides evidence of substantial risk of recidivism. The legislature’s findings are consistent with grave concerns over the high rate of recidivism among convicted sex offenders and their dangerousness as a class. The risk of recidivism posed by sex offenders is ‘frightening and high.'”

In a forthcoming article in Constitutional Commentary, Ira Mark Ellman and Tara Ellman note that Kennedy’s magic words about the recidivism rate of sex offenders — frightening and high — have been cited 91 times by courts around the country, most in the course of upholding state laws allowing for severe ex post facto punishments that can last from years, to decades, to a lifetime. They include registration requirements for which compliance can range from burdensome to impossible; residency restrictions that effectively restrict ex-offenders to living in the shadows (or under a bridge); and restrictions that can make it nearly impossible to find a job, forge meaningful relationships, worship, or generally participate in civilized life.

The scary thing is, as the Ellmans explain, there’s no empirical data to support Kennedy’s oft-cited phrase, and the statistic Kennedy himself cited is paper thin.

McKune provides a single citation to support its statement “that the recidivism rate of untreated offenders has been estimated to be as high as 80%”: the U.S. Dept. of Justice, Nat. Institute of Corrections, A Practitioner’s Guide to Treating the Incarcerated Male Sex Offender xiii (1988). Justice Kennedy likely found that reference in the amicus brief supporting Kansas filed by the Solicitor General, then Ted Olson, as the SG’s brief also cites it for the claim that sex offenders have this astonishingly high recidivism rate. This Practitioner’s Guide11 itself provides but one source for the claim, an article published in 1986 in Psychology Today, a mass market magazine aimed at a lay audience. That article has this sentence: “Most untreated sex offenders released from prison go on to commit more offenses– indeed, as many as 80% do.” But the sentence is a bare assertion: the article contains no supporting reference for it.

But perhaps the author was merely offering an estimate based on his training and expertise. The problem there is that he had little of either.

He is a counselor, not a scholar of sex crimes or re-offense rates, and the cited article is not about recidivism statistics. It’s about a counseling program for sex offenders he then ran in an Oregon prison. His unsupported assertion about the recidivism rate for untreated sex offenders was offered to contrast with his equally unsupported assertion about the lower recidivism rate for those who complete his program.
So the evidence for McKune’s claim that offenders have high re-offense rates (and the effectiveness of counseling programs in reducing it) was just the unsupported assertion of someone without research expertise who made his living selling such counseling programs to prisons.

The Ellmans write that another quote from the solicitor general’s brief in McKune has been widely cited in supporters of these laws and by the courts that have upheld them. Here’s that quote, with its source:

Sex offenders exact a uniquely severe and unremitting toll on the Nation and its citizens for three basic reasons: “[t]hey are the least likely to be cured”; “[t]hey are the most likely to reoffend”; and “[t]hey prey on the most innocent members of our society.” United States Dep’t of Justice, Bureau of Justice Statistics (BJS), National Conf. on Sex Offender Registries (National Conf.) 93 (Apr. 1998).

The Ellmans explain why that source too is bogus:

The “report” is merely a collection of speeches given at a 1998 conference of advocates for sex offender registries. The collection’s cover sheet disavows any Justice Department endorsement of its contents. The “least likely” phrase is taken from a speech in this collection given by a politician from Plano, Texas who never claimed any scientific basis for it. Indeed, she did not even claim it was true. What she actually said was that it is a statement she likes to make. The Solicitor General’s representation of this statement as a Justice Department conclusion about the nature of sex offenders was at best irresponsible.

The real rate of sex offender recidivism is complicated and difficult to calculate, but it’s nowhere near 80 percent. In fact, people convicted of sex offenses re-offend at a lower rate than any other class of crimes. The Ellmans point to a meta-study of 8,000 sex offenders which found that even among high-risk offenders, 20 percent had committed another sex crime within five years, and 32 percent had within 15. Of the high-risk offenders who had gone 16 years without re-offending (about two thirds of them), the study found no example of one of them committing another crime. Among low risk offenders, the risk is minimal. “About 97.5% of the low-risk offenders were offense-free after five years; about 95% were still offense-free after 15 years,” the Ellmans write.

Such nuances were lost on Kennedy, and they’re absent in these laws. Most states require sex offenders to register for decades — some for life. Many states make no distinction between high and low-risk offenders. About a fourth of those on sex offender registries committed their crimes as juveniles. Others are on registries for sexual contact as adults with juveniles who consented and were sexually mature, but who were unable to consent under the law, such a 19-year-old having sex with a 15-year-old.

And despite the court’s assurances in Hendricks that indefinite detainment is about treatment, not punishment, that doesn’t seem to be how it has worked out in the real world. The St. Louis Post-Dispatch reported last year that the alleged treatment center in Missouri for offenders detained after serving their sentences looks a lot like a prison, and that no one who has entered the program since 1999 has ever completed treatment. In June, U.S. District Judge Donovan Frank criticized a similar law in Minnesota, noting that there too, no one had successfully completed treatment since the state started sending offenders there in the mid-1990s.

To fully appreciate what happened here, let’s remove the sex offender element, which can distort our perspective. First, a pop science magazine published a poorly-sourced statistic without double checking its accuracy. The U.S. Office of the Solicitor General then carelessly a subsequent source’s careless citation of that statistic, again without double checking its accuracy. Next, a Supreme Court justice relied on the solicitor general’s brief’s use of that statistic in a plurality opinion (and then again in a majority opinion), again without doing basic research into the accuracy of that statistic. Finally, dozens of lower courts around the country have since run with the language of that opinion, again without verifying its accuracy — often when it isn’t even the controlling case law — in upholding laws that suspend the constitutional rights of thousands of people, in some cases allowing for possible detention for the rest of their lives.

There are a lot of issues to unpack here. Here are several:

Balancing tests invite bad data

This is the main problems with subjecting basic freedoms to balancing tests. Once you’re prepared to say you’re willing to abridge a fundamental liberty to some overriding government interest, you have to start assessing when that interest becomes important enough to start eroding rights. That sort of analysis usually means citing statistics. And we all know what Mark Twain said about statistics.

Judges can be lazy

It would be problematic enough if the courts merely subjected basic rights to balancing tests with opinions that carefully considered the best data available. But that isn’t what happened here. The 80 percent figure in Kennedy’s opinion could easily have been fact-checked. But Kennedy didn’t do that, and neither did his clerks or any of the dozens of courts that later cited his opinion.

Nor is this the only time this has happened. I’ve written before here at The Watch about several other decisions in which the Supreme Court has cited sloppy statistics or based its decisions on popular myths. The court’s decision to allow sobriety checkpoints to check for drunk drivers, for example, balanced drivers’ Fourth Amendment rights against a public safety argument that was based on an inflated statistic about alcohol-related highway fatalities. Many of the court’s Fourth Amendment decisions with respect to roadside searches are based on a popular but mistaken belief that police officers are regularly ambushed during traffic stops. Many of the court’s “use of force” decisions are based on the fallacious assumption that cops are frequently forced to make split-second, life-or-death decisions. The court has also long accepted government claims about the accuracy of eyewitness testimony that science has known for decades is mistaken, including as recently as 2012.

In my series on bite mark evidence, I referenced a law review article that explained how a 1974 opinion by a California appeals court admitting bite mark evidence into a criminal case in a very limited context has since been cited by judges all over the country in subsequent opinions. The analysis in that opinion was lacking, but few courts bothered to question it. In fact, it seems clear that few bothered even to read it. In some cases, courts claimed the opinion was much broader than it actually was, an incorrect interpretation that still other courts cited. Courts and prosecutors have also cited the opinions of other courts about the admissibility of forensic evidence in cases in which the forensic expert claimed that the defendant and only the defendant could have committed the crime, but failed to note that the defendant in the case they’re citing was later proven innocent by DNA. Which brings to me to the next point . . .

The courts are bad at science

The courts have been an utter failure at keeping bad science out of the criminal justice system, from the bite mark evidence already noted, to hair and fiber analysis, to ballistics chemistry, and so on. We’ve covered that extensively here, so I won’t elaborate.

More recently, ProPublica and other outlets pointed out the dubious credentials of Oklahoma’s expert witness in Glossip v. Gross, where death-row inmates argued that a drug Oklahoma planned to use in their executions “would lead to an unconstitutional level of suffering.” But the opinion from Justice Samuel Alito — who has no medical or scientific training — simply pronounced that expert credible. And because Alito was writing for the majority, an opinion informed by an “expert” widely discredited in the medical community will now control every challenge to lethal injection drugs in state or federal court.

Government actors sometimes mislead the courts, and there’s no penalty when they do

As the Ellmans point out, the U.S. solicitor general brief’s use of the bogus statistic in McKune is “at best irresponsible.” Justice Kennedy is guilty of not verifying that stat. But former solicitor general Ted Olson, as well as whichever subordinate wrote that brief, is guilty of misleading the court. At the very least, they’re guilty of not thoroughly researching the issue to at least see if there were other studies that put the figure much lower. It seems more likely that they did come across those other studies and chose not to include them because they didn’t support the government’s position.

Of course, prosecutors aren’t typically held accountable for even really egregious misconduct. So none of these current and former government officials will suffer any professional sanction for simply misleading the court, despite that their transgressions came in critical cases involving life, death, and imprisonment.

The courts are too attached to finality

Despite the fact that it was never supported by any real science, and that what science has been done now shows it to be fraudulent, to date no court in America has ruled against the admissibility of bite mark evidence. Despite the fact that it was relatively new and hadn’t been rigorously studied, the courts were quick to admit testimony from proponents of the Shaken Baby Syndrome diagnosis. But now that the consensus has shifted against the diagnosis as it was widely used, the courts have been slow to revisit those old convictions. Some have, and the record here is at least better than with bite mark analysis, but the criminal justice system certainly hasn’t responded with the urgency you’d expect given the possibility that so many people have been imprisoned over faulty expert testimony.

So I suppose it isn’t surprising that despite the numerous studies done on sex offender recidivism debunking Kennedy’s claim in McKune, 13 years later, the Supreme Court hasn’t endeavored to correct its mistake. Instead, the mistake has been perpetuated in court after court, resulting in a destructive policy that’s imprisoning people who don’t deserve it, and doing nothing to make these communities safer.

Over and over, the courts have been quick to allow bad data and bad science into criminal trials, but are then reluctant to reconsider convictions when the bad data or bad science are exposed. This is because after a conviction, the courts (with help from Congress) have put a premium on the idea of finality. Ironically, this is ostensibly to protect the integrity of the criminal justice system. In reality of course, it protects only the appearance of integrity.

In the end, the courts are making profoundly important decisions with little to no effort to ensure the validity, accuracy, and credibility of the statistics, experts, and data that inform those decisions. The really depressing thing is that there’s very little we can do about it.