Few hearts bleed for Emory Michau Jr. Now 60, Michau was jailed in 1993 for molesting an 8-year-old boy. After serving his sentence, he solicited sex from a 17-year-old youth and was jailed once more. By 2003, after a further two years inside, he was scheduled for release. Yet Michau remains incarcerated in Charleston county jail, South Carolina. Not for his previous crimes, but because he has refused to undergo psychiatric assessment to determine whether he poses a danger to the public. If the assessment went against him, Michau could be locked up indefinitely in a secure mental facility.
His case highlights a growing controversy surrounding the US justice system. Thousands of sexual offenders who have already served their jail terms are kept incarcerated in mental institutions, and some psychiatrists and legal commentators say there are major inadequacies in the assessment methods that put them there. On top of its implications for the civil liberties of the people who would otherwise have been released, the practice is estimated to cost more than a quarter of a billion dollars a year to implement and does little to reduce levels of sexual abuse or rape in society.
In many countries, public revulsion towards sex offenders – particularly those who abuse children – has led to harsher sentencing. Since April 2005, judges in England and Wales have been able to impose indeterminate sentences on dangerous convicted criminals, including some sex offenders. Similar provisions exist under Canadian law. In the Australian state of Queensland, sex offenders can be imprisoned indefinitely, subject to an assessment made after they have served some time in jail.
Advertisement
The US is different in that some sex offenders are first dealt with as criminals, but after serving their time in prison are turned over for indefinite “civil commitment” on the grounds that they are suffering from a mental illness that makes them dangerous to the public. In January, New Hampshire became the 18th US state to implement civil commitment for sex offenders who are deemed to be dangerous, and others are debating whether they should follow suit.
By May 2006, 3646 individuals were being held in the US under these laws, according to the most recent survey, conducted by Adam Deming of the Indiana Sex Offender Management and Monitoring Program in Indianapolis. Of these, 2627 had been committed as dangerous sexual predators, while the other 1019 were waiting for their evaluations to be completed (see Map). Few can expect to be released any time soon. Just 427 of 3493 offenders detained since 1990 had been released by 2004, according to a survey by Roxanne Lieb of the Washington State Institute for Public Policy in Olympia.
While the details of civil commitment laws vary from state to state, the US Supreme Court has ruled that civil commitment can be applied only to convicted sex criminals who have a mental disorder that makes them likely to commit further sexually violent acts. This is where the problems begin, as many sex offenders do not easily fit any of the categories defined in the psychiatrists’ bible, the Diagnostic and Statistical Manual of Mental Disorders or DSM, published by the American Psychiatric Association.
The DSM lists a range of what are known as paraphilias: intense sexually arousing fantasies, sexual urges or behaviours that recur over a period of at least six months. Sexual sadism and paedophilia fall under this heading, yet only a small minority of rapists have sadistic fantasies, and even persistent child abusers may not be sufficiently fixated on sex with children to qualify as paedophiles, according to a strict interpretation of the DSM.
Some observers, including Lieb, argue the fault lies with the DSM for providing too narrow a definition of paraphilia. Critics of civil commitment argue that some offenders are being shoehorned into diagnoses that do not apply, to satisfy the Supreme Court’s requirement to allow them to be detained.
“Some offenders are being shoehorned into diagnoses that do not apply”
Psychiatrists stress that being a sexual offender does not necessarily make you mentally ill according to any recognised criteria. Michael First, a psychiatrist at Columbia University in New York and an adviser on the DSM, argues that diagnoses made for the purpose of civil commitment proceedings often confuse persistent criminal behaviour with mental illness. “The behaviour itself is not enough to make the diagnosis,” he says.
Of 2082 diagnoses recorded by Deming in his survey, 1135 detainees had been marked down as paedophiles, while 692 were evaluated as belonging to the catch-all category of “paraphilia (not otherwise specified)”. Six states failed to provide Deming with the particular paraphilia diagnoses used to detain offenders.
Eric Janus of the William Mitchell College of Law in St Paul, Minnesota, claims that diagnoses are open to manipulation. He has studied men detained under Minnesota’s civil commitment laws. Just 25 per cent of those committed in 1993 were diagnosed with a paraphilia, but by 1996 the figure for newly committed men was more than 90 per cent. By 2001, 97 per cent of all men held under civil commitment in Minnesota were judged to be paraphilic, including a substantial proportion who were not given this diagnosis when they were first detained.
These are not the only questions being raised about the consistency of diagnoses. Jill Levenson, who studies criminal justice policy at Lynn University in Boca Raton, Florida, compared the diagnoses made by different professional psychiatric evaluators for nearly 300 offenders assessed under that state’s civil commitment law. In general, she was encouraged by the level of agreement. In 85 per cent of cases, two evaluators agreed on whether or not an offender was a paedophile. When it came to a diagnosis of “paraphilia (not otherwise specified)”, however, her survey revealed the level of agreement fell to 68 per cent.
“No one is quite sure what counts as a mental disorder,” Janus argues. If no one can be sure of that, how can courts rely on psychiatric diagnoses to incarcerate offenders who have served their sentence?
The second major flaw in the system relates to the methods used to assess the likelihood that a particular individual will reoffend if released. Here, evaluators rely heavily on actuarial risk assessments. Conceptually, these are similar to the statistical tools used to calculate a person’s car insurance premiums given the crime rate in their neighbourhood, the extent of their driving experience, and so on. The tools used to assess sex offenders consider factors such as the extent of their prior offending, the sex of their victims, and whether these victims were strangers. They are derived using information about a sample of known sex offenders, and subsequently validated by analysing their predictive value when applied to other offenders after their release.
Research into the value of these tools involves some arcane statistics (see “Powers of prediction”). However, the findings are simple enough: the tools are better than expert clinicians at predicting whether a sexual offender will reoffend.
That doesn’t mean they always get it right. According to Karl Hanson of Public Safety and Emergency Preparedness Canada in Ottawa, who is a co-author of Static-99, the assessment tool most commonly used in US civil commitment proceedings, the best that risk-assessment tools can achieve is to identify a high-risk group – encompassing between 10 and 15 per cent of all sex offenders – who have about a 60 per cent chance of being reconvicted within 15 years of release. Using such tools to commit these “high-risk” offenders would therefore mean that 40 per cent of those kept locked up would not have been reconvicted within 15 years had they been released instead.
In practice, civil commitment proceedings are even less likely to accurately predict reoffending, according to Raymond Knight, a psychologist at Brandeis University in Waltham, Massachusetts. In the hearings, risk assessments are typically adjusted to account for other factors deemed relevant by professional evaluators, including aspects of an offender’s psychiatric profile. “Clinicians are very bad at weighting new information and adding it to an actuarial assessment,” Knight says.
Even allowing for the fact that reconviction rates underestimate offending – because many sex crimes go unreported or their perpetrators aren’t found – the evidence suggests that a substantial minority of those held under civil commitment would not reoffend if released. “Are assessment tools accurate enough to take away people’s liberty prospectively? My answer would be no,” Janus concludes.
Most states’ civil commitment laws have been worded to sidestep such objections, and require only that those detained are more likely to reoffend than not. Some set the bar even lower. “These laws have been developed with more concern for public protection than civil liberties,” says Cynthia Calkins Mercado, a clinical psychologist at the John Jay School of Criminal Justice in New York.
“These laws have been developed with more concern for public protection than civil liberties”
So do these laws protect the public? Statistics on sex crimes suggest they do not to any great extent.
The US Bureau of Justice Statistics has studied offending patterns for a cohort of more than 270,000 prisoners released in 1994 across 15 states. Given that most states had not then enacted civil commitment laws, the sex offenders in that group are likely to include individuals who today would be detained as dangerous sexual predators.
Sex offenders accounted for about 4 per cent of those released, and over three years of follow-up these individuals were on average about four times as likely to be arrested for a subsequent sex offence as those previously jailed for other crimes. Yet because there were fewer of them, the sex offenders still accounted for only a minority of the sex crimes committed by the group as a whole: of the ex-cons subsequently arrested for sexual offences, 87 per cent had previously been imprisoned for some other type of crime.
What’s more, analyses by the Bureau of Justice Statistics of felonies in large urban counties across the US shows that most sex crimes are committed by people who have never been convicted of any crime. The latest available figures, from 2002, reveal that 79 per cent of those charged with rape had no prior felony convictions. Evidence from various jurisdictions similarly suggests that most convicted child abusers had also not previously been in trouble with the law.
To critics of current policies, the mismatch between public perceptions and crime statistics is at the root of the problem. People are appalled by sex offending, but do not like to acknowledge that it is widespread throughout society. So the public and media demonise convicted offenders, and politicians devise laws such as civil commitment in response. “It’s a way of articulating society’s condemnation of sexual violence without doing anything fundamental about it,” says Janus.
Powers of prediction
To judge the tools used to assess the danger posed by sex offenders, statisticians rely on what is known as the receiver operating characteristic (ROC) curve – most famously used to investigate the performance of US radar following the Japanese attack on Pearl Harbor.
ROC curves plot the proportion of false positives (in this case, offenders identified as high risk who do not offend again) on the horizontal axis against the proportion of true positives (similar “high-risk” offenders who do reoffend) on the vertical axis.
Just guessing should give a straight line, while progressively better prediction tools move the curve towards the top left (see Graphs). These plots can be summarised by calculating the proportion of the area of the graph under the curve: 0.5 for random guesses; 1.0 for a perfect prediction, which would give a single point in the top left corner.FIG-mg25924202.jpg
In a new unpublished study conducted for the US National Institute of Justice, psychologists Raymond Knight of Brandeis University in Waltham, Massachusetts, and David Thornton of the Sand Ridge Secure Treatment Center in Mauston, Wisconsin, have tested the leading actuarial risk assessments using information on a sample of sex offenders treated at a secure centre in Massachusetts between 1959 and 1984. Looking at sex offences committed in the first three years after release, the areas under the curves varied between 0.67 and 0.70 – apparently not a bad result.
But if you consider what’s called the predictive area – the total area between a straight line representing a prediction based on chance, and perfect prediction based on a single point in the top left corner – then things look less good.
“We’re explaining less than 40 per cent of the predictive area,” says Knight. “When I explain that, most people are horrified.”
