University of New South Wales Law Journal
|
|
Home
| Databases
| WorldLII
| Search
| Feedback
University of New South Wales Law Journal |
|
STEPHEN SMALLBONE AND JANET RANSLEY[*]
Sexual offenders have increasingly become the target of special criminal justice policies, including preventive incapacitation policies. These policies rely on assumptions that sexual offenders are an especially high-risk offender population and that within this population, it is possible to identify those individuals who present the highest risk. Recent sexual offender recidivism and risk prediction research indicates that:
• average sexual recidivism rates are lower than is popularly assumed;
• sexual offenders are a very heterogenous population, within which wide variations in recidivism patterns have been observed; and
• current risk prediction methods are associated with high rates of false-positive and false-negative predictions.
We conclude that actuarial risk prediction scales may contribute important information to psychological and psychiatric risk assessments, but that their limitations need to be better understood and more openly communicated to the courts.
A broad range of criminal justice policies directed exclusively at sexual offenders have emerged internationally in recent years. For example in the United States, special policies directed towards sexual offenders include community notification statutes, mandatory sexual offender registration statutes, requirements for specialised sexual offender treatment, and involuntary commitment to secure psychiatric hospitals for offenders assessed as ‘sexually violent predators’.[1] The increasing awareness and intolerance of sexual crime, and particularly of child sexual abuse, in the US is clearly reflected in the increasing numbers of sexual offenders entering US prisons. Between 1980 and 1994, while the US prisoner population increased by 200 per cent, the number of imprisoned sexual offenders increased by 330 per cent.[2] Despite strong epidemiological evidence that the incidence of child sexual abuse has declined steadily in the US since 1993,[3] the number of imprisoned sexual offenders has continued to grow, from 94 000 in 1998[4] to 118 500 in 2002,[5] a 26 per cent increase.
Similar developments have been observed in Australia. Public policy developments here have included:
• public campaigns designed to increase awareness and reporting of child sexual abuse;
• the formation of special police taskforces and changes to policing practices designed to target active child-sex offenders;
• legislative reviews concerning penalties, sentences, offender registration, community notification and employment screening;
• the establishment of formal inquiries and commissions charged with examining past and current child sexual abuse; and
• the development and implementation of specialised intervention programs for convicted sexual offenders.[6]
The number of sexual offenders in Australian prisons has also grown, although not to the same extent as in the United States. The Australian prisoner population increased by 102 per cent between 1982 and 1998, from 89.8 prisoners per 100 000 members of the general adult population to 139.2 per 100 000.[7] As in the US, the imprisonment of sexual offenders has outpaced the general growth in imprisonment rates. Between 1988 and 1998, the proportion of prisoners serving sentences for sexual offences grew from 10 per cent to about 14 per cent. Although median sentence lengths decreased over that time nationally, in Queensland the median sentence length for sexual offenders increased by 20 per cent from approximately 75 months in 1988 to almost 90 months in 1998.[8] Thus, more sexual offenders are being imprisoned, and in some states sexual offenders are serving longer sentences than in the past.
Although many recent legislative and policy developments have been justified by claims that sexual offenders (and especially sexual offenders against children) are an especially high-risk population of offenders, the empirical basis for this justification is tenuous. In their meta-analysis of sexual offender recidivism studies, involving more than 22 000 imprisoned sexual offenders from five countries (United States, Canada, United Kingdom, Australia and New Zealand), Hanson and Bussière[9] reported an average sexual recidivism rate of 13 per cent within five years at risk. Even over the longer term (up to 20 years at risk), average sexual recidivism rates have been estimated to be about 35 per cent.[10] Although the Diagnostic and Statistical Manual of Mental Disorders (‘DSM-IVTR’)[11] advises that pedophilia is a chronic, life-long disorder, empirical analyses show that the risk of sexual recidivism declines substantially with age.[12]
Sexual offenders are a very heterogenous group and average sexual recidivism rates are likely to conceal wide variations among offender subgroups. For example, Hanson and Bussière’s[13] meta-analysis showed that non-familial offenders (those who had offended against children outside family settings) were more than twice as likely to be reconvicted for new sexual offences (19 per cent over five years at risk) than were familial offenders (8 per cent over five years). Of course, as with nonsexual offences, many sexual offences are never reported to police and official recidivism data will therefore underestimate true recidivism rates. It is unlikely, though, that the extent to which official recidivism data underestimate true recidivism is simply reflected in the discrepancies between official and unofficial victimisation. Rather, a relatively small group of persistent offenders may be disproportionately responsible for sexual offence recidivism. Smallbone and Wortley[14] asked 180 Queensland child-sex offenders about the unofficial or ‘true’ extent of their sexual offending. Almost half said they had only ever offended against one child. Ten per cent said that they had offended against more than 10 children, including a small number (fewer than 1 per cent) who reported more than 100 child victims. These data illustrate how the selective incapacitation of a relatively small group of persistent, serial sexual offenders should, at least hypothetically, produce a substantial reduction in sexual victimisation.
United States civil commitment laws for sexual offenders were introduced in the 1880s and later fell into decline, but were re-popularised in the 1990s.[15] Various Australian jurisdictions have now introduced similar measures aimed at the preventive incapacitation of sexual or serious violent offenders.[16] These laws generally displace the presumption of proportionality, permitting longer-thanusual or indefinite detention where the offender is thought to pose a serious danger to the community.[17] For example, Queensland legislation provides for the indefinite, preventive detention of:
• convicted sexual offenders against children under 16, who are found on the basis of medical evidence to be incapable of exercising proper control over their sexual instincts, or whose mental condition is subnormal – in the Criminal Law Amendment Act 1945 (Qld) (the ‘CLAA’);[18]
• people convicted of violent or sexual offences attracting possible life sentences where the court is satisfied that the person does not merit reference to the Mental Health Court for examination, and is a serious danger to the community because of their antecedents, character, age, health or mental condition, the severity of the offence and any special circumstances – in the Penalties and Sentences Act 1992 (Qld) (the ‘PSA’);[19]
• prisoners in custody for serious sexual offences who would be a serious
• danger to the community if released, based on two psychiatrists’
• assessments of the level of risk of the person committing another serious
• sexual offence if released, or if released without a supervision order – in the
• Dangerous Prisoners (Sexual Ofenders) Act 2003 (Qld) (‘DPSOA’).
An important difference between these three Acts is their volitional elements. While the CLAA requires that a person has committed certain offences and is incapable of controlling their sexual instincts,[20] the PLA and the DPSOA require only that a person has committed certain offences and is a serious danger to the community. Their ability to control their actions is not paramount, although it may be relevant to the finding of dangerousness.
The second major difference between the Acts is their risk elements. Orders under the CLAA rest on a finding of fact that the person is incapable of exercising proper control over their sexual instincts, or is subnormal.[21] While the first of these findings requires expert advice, there is no risk element involved – the issue is simply whether the person lacks control, not whether they also are dangerous or a risk to the community. There is an element of risk in the orders relating to the second finding, subnormal offenders – they must be subnormal and require care, supervision and control in their own interests or for the protection of others. But this element still falls well short of the role of risk in the PLA and the DPSOA, where the main question for the court is the assessment of dangerousness and the risk of re-offending. Indeed, under the DPSOA the psychiatric reports are called risk assessment reports.[22]
Thus, while the legislation from 60 years ago was primarily directed at assessing people’s condition, the focus in the newer legislation has shifted to assessing their risk potential – what they might do in the future. In his Second Reading Speech on the DPSOA, the Attorney-General described the tests in the CLAA as archaic and out of touch with community standards,[23] saying that they were not in accordance with modern understandings of pedophilia. Furthermore, the CLAA provided no protection from offenders capable of controlling their actions but choosing not to do so. In other words, although the new legislation was described by the Attorney-General as providing a scheme ‘akin to the detention authorised under mental health laws’,[24] this detention is not founded on any condition suffered by a person, but on expert assessment of their future conduct.
In Fardon v Attorney-General for the State of Queensland[25] (‘Fardon’), the High Court upheld the constitutional validity of the DPSOA. This Act was challenged on the basis that it conferred on the Supreme Court of Queensland functions incompatible with its judicial role, based on the decision in Kable v Director of Public Prosecutions for New South Wales.[26] The majority found no incompatibility between the judicial function and the imposition of preventive sentences based purely on predictions of dangerousness, provided the detention is preventive as opposed to punitive, and that procedural safeguards are met, in particular: there is real judicial discretion; the onus of proof is on the Attorney-General; the rules of evidence apply; and clear criteria exist.[27] Justice Kirby’s dissenting judgment was based on his opinion that the Act imposes punishments for future crimes, and on the unreliability of predictions of dangerousness.[28]
In light of the wide variations in persistence observed among sexual offenders, selective incapacitation policies may be more justifiable than are incapacitation policies designed to apply uniformly to sexual offenders, or even to certain subgroups of sexual offenders. However, the practical application of selective policies remains problematic. Key concerns, albeit widely understated in the professional and research literature, focus on the limited accuracy of current risk prediction methods and associated problems of using risk prediction in applied settings.
A thorough discussion of the relative merits and limitations of clinical, actuarial and combined clinical/actuarial prediction methods is beyond the scope of the current article. It may be sufficient to note here that actuarial (empirically-based) prediction methods have generally been associated with the strongest evidence for predictive accuracy.[29] The accuracy of clinical prediction, which relies solely on professional judgement, has at best proved to be only slightly better than chance.[30] However, while modern actuarial methods are more reliable and accurate, they have rarely been shown to account for more than about 10 per cent of the variance in recidivism. That is to say, as much as 90 per cent of the variance in observed recidivism outcomes remains unexplained by actuarial methods. Although proponents of actuarial prediction methods argue that the inclusion of additional clinical information only serves to further reduce reliability and accuracy, in practice few psychiatrists and psychologists would rely solely on actuarial methods. Rather, actuarial assessment is generally used to establish a base-line probability, with adjustments made according to the perceived relevance of other available information.
The Static-99[31] is one of the best validated and most widely-used actuarial sexual offender risk prediction scales. It is also among the most accurate. The Static-99 is a 10 item scale that produces scores for individual sexual offenders ranging from zero (the lowest risk) to 12 (the highest risk). A standard statistical measure of the predictive accuracy of scales such as the Static-99 is the so-called Area Under the Curve of the Receiver Operating Characteristic statistic (the AUC, for short). The AUC value reported for the Static-99 is typically around 0.65 to 0.75 (an AUC value of 0.50 would indicate prediction no better than chance; a value of 1.0 would indicate perfect prediction). This means that there is a 65 per cent to 75 per cent probability that a randomly selected recidivist will have a higher Static-99 score than will a randomly selected non-recidivist, a degree of accuracy that could be described as modest at best.
The AUC statistic also conceals problems with the sensitivity (accuracy in classifying recidivists) and specificity (accuracy in classifying non-recidivists) of actuarial scales. The sensitivity and specificity of the 12-point Static-99 scale were found by Sjostedt and Langstrom[32] to be maximized with cut-off scores of >1 and <6, respectively. As Campbell[33] has pointed out, with maximum sensitivity (identifying as likely recidivists all those who score two or higher) almost all (92 per cent) of the sexual recidivists would be correctly identified, but more than half (55 per cent) of the non-recidivists would be incorrectly classified as recidivists. With maximum specificity (identifying as likely non-recidivists all those who score five or lower), 93 per cent of the non-recidivists would be correctly classified, but 65 per cent of the recidivists would be incorrectly classified as non-recidivists. In statistical terms, predictive accuracy is always a trade-off between sensitivity and specificity. In practical terms, the threshold of tolerance for false-positive or false-negative errors should depend on the consequences of those errors. The threshold of tolerance for both false-positive and false-negative errors should be considered very seriously by courts.
The limited accuracy of current risk prediction methods highlights longstanding problems about how psychological and psychiatric expertise might contribute to legal judgments. These problems are brought into particular focus in the administration of Queensland’s DPSOA. This Act requires the court to consider two psychiatrists’ assessments ‘of the level of risk that the prisoner will commit another serious sexual offence’,[34] and to be satisfied ‘by acceptable, cogent evidence; and to a high degree of probability’[35] that ‘a prisoner is a serious danger to the community’.[36] Clearly the setting of such a high threshold of probability is intended to protect against the arbitrary or punitive imposition of indefinite, post-sentence detention or supervision. However, by setting such a high threshold, the DPSOA may inadvertently rely on the ability of professionals to provide advice that is currently beyond their level of knowledge and expertise. Actuarial risk prediction scales may contribute important information to psychological and psychiatric risk assessments, but their limitations must be better understood by courts, and more openly acknowledged.
[*] Stephen Smallbone PhD is a psychologist and Janet Ransley PhD is a lawyer, in the School of Criminology and Criminal Justice at Griffith University.
[1] Leonore M J Simon and Kristen Zgoba, ‘Prevention of Sex Crimes Against Children: Legislation, Prevention and Investigation’ in Richard Wortley and Stephen Smallbone (eds), Situational Perspectives on Sexual Ofences Against Children (forthcoming, 2006).
[2] Lawrence A Greenfeld, ‘Sex Offenses and Offenders: An Analysis of Data on Rape and Sexual Assault’ (US Department of Justice, Bureau of Justice Statistics, NCJ 163392, 1997).
[3] Lisa Jones and David Finkelhor, ‘The Decline in Child Sexual Abuse Cases’ (Office of Juvenile Justice and Delinquency Prevention, Juvenile Justice Bulletin, NCJ 184741, 2001).
[4] Allen J Beck and Christopher J Mumola, ‘Prisoners in 1998’ (US Department of Justice, Bureau of Justice Statistics, NCJ 175687, 1999).
[5] Paige M Harrison and Allen J Beck, ‘Prisoners in 2002’ (US Department of Justice, Bureau of Justice Statistics, NCJ 200248, 2003).
[6] Stephen Smallbone and Richard Wortley, Child Sexual Abuse: Ofender Characteristics and Modus Operandi, Trends and Issues in Crime and Criminal Justice No 193, Australian Institute of Criminology (2001).
[7] Carlos Carcach and Anna Grant, Imprisonment in Australia: Trends in Prison Populations and Imprisonment Rates 1982-1998, Trends and Issues in Crime and Criminal Justice No 130, Australian Institute of Criminology (1999).
[8] Carlos Carcach and John Chisholm, Imprisonment in Australia: Sentenced Populations, Trends and Issues in Crime and Criminal Justice No 174, Australian Institute of Criminology (2000).
[9] R Karl Hanson and Monique T Bussière, ‘Predicting Relapse: A Meta-Analysis of Sexual Offender Recidivism Studies’ (1998) 66 Journal of Consulting and Clinical Psychology 348–62.
[10] Terence W Campbell, ‘Sex Offenders and Actuarial Risk Assessments: Ethical Considerations’ (2003) 21 Behavioural Sciences and the Law 269–79.
[11] American Psychiatric Association, Diagnostic and Statistical Manual of Mental Disorders (4th ed, 2000).
[12] R Karl Hanson, ‘Recidivism and Age: Follow-up Data from 4673 Sexual Offenders’ (2002) 17 Journal of Interpersonal Violence 1046.
[13] R Karl Hanson and Monique T Bussière, above n 9.
[14] Stephen Smallbone and Richard Wortley, above n 6.
[15] Peter C Pfaffenroth, ‘The Need for Coherence: States’ Civil Commitment of Sexual Offenders in the Wake of Kansas v Crane’ (2003) 55 Stanford Law Review 2229.
[16] Sentencing Act 1991 (Vic) pt 3 div 2 (1A); Criminal Law (Sentencing) Act 1988 (SA) pt 2 div 3; Criminal Code (WA) s 662(a); Sentencing Act 1995 (WA) s 98. See also the Powers of Criminal Courts (Sentencing) Act 2000 (UK) s 80(2)(b).
[17] See Ralph Henham, ‘The Policy and Practice of Protective Sentencing’ (2003) 3(1) Criminal Justice 57; Elizabeth Richardson and Arie Freiberg, ‘Protecting Dangerous Offenders from the Community’ (2004) 4(1) Criminal Justice 81.
[18] Sections 18(3)–(6).
[19] Sections 162, 163.
[20] Criminal Law Amendment Act 1945 (Qld) s 18(1).
[21] Criminal Law Amendment Act 1945 (Qld) s 18(1).
[22] Dangerous Prisoners (Sexual Ofenders) Act 2003 (Qld) ss 8, 11.
[23] Queensland, Parliamentary Debates, Legislative Assembly, 3 June 2003 (Rod Welford, Attorney-General of Qld).
[24] Ibid.
[25] [2004] HCA 46; (2004) 210 ALR 50.
[26] [1996] HCA 24; (1996) 189 CLR 51.
[27] Fardon [2004] HCA 46; (2004) 210 ALR 50, 19 (Gleeson CJ).
[28] Ibid 124–6 (Kirby J).
[29] Joel A Dvoskin and Kirk Heilbrun, ‘Editorial: Risk Assessment and Release Decision-Making: Toward Resolving the Great Debate’ (2001) 29 Journal of the American Academy of Psychiatry and Law 6; Thomas R Litwack, ‘Actuarial versus Clinical Assessments of Dangerousness’ (2001) 7 Psychology, Public Policy and Law 409.
[30] R Karl Hanson and Monique T Bussière, above n 9.
[31] R Karl Hanson and David Thornton, Static 99: Improving Actuarial Risk Assessments for Sex Ofenders, User Report 99–02, Department of the Solicitor General of Canada (1999).
[32] Gabrielle Sjostedt and Niklas Langstrom, ‘Actuarial Assessment of Sex Offender Recidivism Risk: A Cross-Validation of the RRASOR and the Static-99 in Sweden’ (2001) 25 Law and Human Behavior 629.
[33] Above n 10.
[34] Dangerous Prisoners (Sexual Ofenders) Act 2003 (Qld) s 11(1).
[35] Dangerous Prisoners (Sexual Ofenders) Act 2003 (Qld) s 13(3).
[36] Dangerous Prisoners (Sexual Ofenders) Act 2003 (Qld) s 13(4).
AustLII:
Copyright Policy
|
Disclaimers
|
Privacy Policy
|
Feedback
URL: http://www.austlii.edu.au/au/journals/UNSWLawJl/2005/19.html