FORUM on Corrections Research

Academic Contributions

Sex offender research

Detention with sexual offenders: Can we do better?¹

Andrew M. Haag²
Counselling Psychology, University of Calgary

Justifiably, there is a great deal of public concern with regard to sexual offenders (Brown, 1999; Banister & Pordham, 1994; Hogue, 1993). Moreover, the question as to the most appropriate response strategy with sexual offenders can become politically and emotionally charged. The fact that an issue is charged, however, does not provide licence for ignoring empirical data on a topic.

In particular, it is important to, as accurately as possible, determine the level of dangerousness that any one offender actually presents at any one given time. This implies that careful attention should be paid to the predictive accuracy of our decision making with offenders.

This article seeks to explore the predictive accuracy of one of the more invasive forms of decision making in Canada: detention.

Background

In Canada, detention refers to an intentional decision on the part of the National Parole Board of Canada (NPB) to confine a federally sentenced offender (i.e., someone sentenced to two years or more) past his or her regular statutory release date (Department of Justice Canada, 1992a and 1992b). An offender reaches the statutory release date at the two-thirds point in the sentence. The decision to detain an offender is made by the NPB if it is believed that the offender is likely to commit a violent or sexual offence or a serious drug offence prior to the expiry of his or her court-imposed sentence. In Canada, the date at which an offender's sentence ends/expires is also referred to as the warrant expiry date (WED). All sexual offenders with a determinate federal sentence in Canada are potentially eligible for detention.

In making the decision to detain, according to section 132 of the Corrections and Conditional Release Act, the NPB is obligated to consider any factor that they deem relevant to determining the likelihood of a future serious offence, including: 1) patterns of persistent violence or sexual behaviour involving children; 2) the seriousness of the current offence; 3) reliable information suggesting that the offender has difficulty controlling violent/sexual impulses; 4) the use of a weapon in any offence; 5) explicit threats of violence; 6) behaviour of a brutal nature in any offence; 7) a substantial degree of indifference; 8) medical, psychiatric or psychological evidence; 9) reliable information concerning offence planning; 10) the availability of appropriate supervision programs; 11) the number of sexual offences involving a child; 12) behaviour of a sexual nature associated with the commission of any offence; and 13) reliable information about the offender's sexual preferences indicating that the offender is likely to commit a sexual offence involving a child.

Missing from this list is any legal obligation for the NPB to consider actuarial data. Hopefully, however, this information would be included in potential psychiatric or psychological reports submitted to the NPB.

The topic of detention has been the subject of prior research. In particular, Johnson (2002) sought to determine if there is a bias in the criminal justice system against sexual offenders. Specifically, Johnson (2002) found that sexual offenders were (1) given more restrictive types of release than other offenders and (2) released later in their sentence than other offenders. It was interesting to note that these same sexual offenders had lower rates of reoffending than the comparison group considered.

In a different review by Grant (1996), it was found that detained offenders were more likely to have been convicted of a sexual offence than those who were not detained. This was interpreted by Grant to suggest that detention decisions disproportionately affect sexual offenders.

The aim of this study was to further explore detention among sexual offenders. In particular, this study sought to examine the predictive validity of the NPB decision-making process with sexual offenders.

Method

Participants

The participants of this study consisted of every sexual offender under federal custody in Canada who had a WED in 1995. A sexual offence was defined as an offence meeting the Static-99 definition for a sexual offence (Hanson & Thornton, 1999; Harris, Phenix, Hanson & Thornton, 2003). In other words, this study examined a cohort of federally sentenced sexual offenders across Canada. As this methodological approach largely lacked a selection bias, it was thought to increase the generalizability of the findings to all federally sentenced sexual offenders in Canada.

The total number of potential participants for this project was 775 male sexual offenders who reached warrant expiry in 1995. All participants were 18 years of age or older. This study did use exclusion criteria, however. Specifically, this study excluded offenders who: 1) were deported or it was documented that they voluntarily left the country (26 participants); 2) died prior to the conclusion of the follow-up period (19 participants); 3) received a pardon during the follow-up period (15 participants); 4) successfully appealed their convictions or sentence lengths (15 participants); or 5) did not meet the Static-99 definition of a sexual offence (26 participants) (Hanson & Thornton, 1999; Harris, Phenix, Hanson & Thornton, 2003).

After the above exclusions, the project was left with a cohort of 674 male sexual offenders who had reached warrant expiry in 1995.

Variables considered

For this study, the independent variable considered was the decision by the National Parole Board of Canada with regard to the detention of the sexual offenders in the participant pool.

The dependent variable of concern was sexual recidivism, which was defined as convictions for sexual offences within seven years after warrant expiry. Sexual recidivism data was obtained from the Canadian Police Information Centre (CPIC) records (i.e., Canadian criminal records checks - a national record of all convictions for any one offender as maintained by the courts and the Royal Canadian Mounted Police).

Results

Descriptive statistics

The average (mean) age and sentence length of the sexual offenders in this cohort are reported in Table 1. Overall, participants were, on average, about 41 years old at warrant expiry and had a sentence of just over four years.

With regard to conditional release, about one in five offenders (21.5% or 145) were detained until their warrant expiry date. About two in five offenders (40.7% or 274) were released on their statutory release date (i.e., at the two-thirds point in their sentence). Another 2.2% (15) were released on statutory release but subsequently had their statutory release revoked; the revocation then led to these offenders only being released at their warrant expiry date. About one in four offenders (26.4% or 178) were released on day parole, and 8.2% (55) were released on full parole. Seven offenders voluntarily chose to wave their statutory release date and were released on their warrant expiry date.

Table 1

Mean Age and Sentence Lengths

Variable	Mean	Standard deviation
Age at warrant expiry	40.82 years	11.47 years
Sentence length	1,570 days	1,013 days
	(4.3 years)

Sexual recidivism and detention were found to be significantly correlated overall . . . .

It should be noted that the above figures represent the point at which an offender initiated a continual release until their warrant expiry date. In other words, if an offender received a conditional release which was then revoked for some reason, and the offender was then subsequently released on another conditional release at a future date, the figure above would document the second (or last) conditional release as the start of their time in the community prior to warrant expiry.

The global recidivism rate for the cohort of sexual offenders was 14.4%, meaning 97 offenders were convicted of another sexual offence within the follow-up period of seven years after warrant expiry.

Sexual recidivism and detention were found to be significantly correlated overall (r = .20, p < .01). Moreover, when the ROC curve analysis was performed, there was noted to be a moderate predictive relationship between detention and sexual recidivism (AUC = .615, Std. Error = .03, 95% CI = .55 to .68). The relationship between detention decisions and sexual recidivism remained consistent regardless of statistical controls for time in the community prior to warrant expiry (r = .14, p < .01).

. . . the ability of the National Parole Board to accurately detain sexual offenders was only statistically significant with programming completers.

Table 2 documents the correlation between detention and sexual recidivism based on treatment status. Treatment for this data set was defined as any intervention noted to CSC files that was specifically designed to address sexual offending behaviour. Treatment status was categorized by indicating whether the offender had: 1) completed programming; 2) declined to accept programming (programming non-starter); or 3) started programming and then left for any reason (programming drop-out). There were then separate predictive analyses performed for detention and recidivism based on treatment completion status. This data provides evidence that the ability of the National Parole Board to accurately detain sexual offenders was only statistically significant with programming completers. This level of statistical significance was not observed with programming drop-outs or programming non-starters.

Table 2

Correlation between Detention and Sexual Recidivism based on Treatment Status

Programming status*	n	r	p
Completed	461	.25	.01
Non-Starter	171	.08	.32
Drop-Out	41	.28	.07

The details of a survival analysis comparing detained and non-detained offenders are noted in Figure 1. A survival analysis is a graphical method that analyzes the time to the occurrence of an event. In this case, the event in question was sexual recidivism. The graph illustrates the percentage of offenders in different groups (in this case detained versus not detained) who recidivated over various lengths of time. The results indicate that detained offenders did not last as long in the community before committing a new sexual offence as compared to the offenders who were not detained.

Discussion

The results of the current study suggest that, on the whole, the NPB appears to be appropriately detaining sexual offenders who go on to commit future sexual crimes with some degree of accuracy. Certainly, this finding in and of itself was encouraging.

When further analysis was performed, however, it was found that the NPB's ability to appropriately detain offenders decreased to non-significance among those who had either dropped out of programming or never started programming. From the current data, it is not possible to know why the predictive accuracy decreased for these other groups. Certainly, further analysis of this would be a fruitful area for future research.

It is suggested ... that the accuracy of NPB detention decisions could be improved by more strongly considering actuarial measures in their decision making with regard to detention.

As a global comment, despite statistical significance, the NPB's detention decision was still only somewhat better than chance overall. In other words, there is a great deal of room for improvement. The predictive power of the NPB's decision making is still considerably less than the levels of accuracy reported by actuarial measures such as the Static-99 or Static-2002 with regard to sexual recidivism (Hanson & Thornton, 1999; Hanson & Thornton, 2003). Given the literature outlining the superiority of actuarial methods globally to that of clinical decision making, this latter comparison is not surprising (Grove & Meehl, 1996). It is suggested, however, that the accuracy of NPB detention decisions could be improved by more strongly considering actuarial measures in their decision making with regard to detention.  

---

References:

Banister, P., & Pordham, S. (1994). Public perceptions of crime seriousness. Issues in Criminological & Legal Psychology, 21, 4-10.

Brown, S. (1999). Public attitudes toward the treatment of sex offenders. Legal & Criminological Psychology, 4(2), 239-252.

Department of Justice Canada. (1992a). Corrections and Conditional Release Act.

Department of Justice Canada. (1992b). Corrections and Conditional Release Regulations.

Grant, B.A. (1996). Inmates referred for detention (1989-90 to 1993-94): A comparative analysis. Ottawa, ON: Correctional Service Canada.

Grove, W.M., & Meehl, P.E. (1996). Comparative efficiency of informal (subjective, impressionistic) and formal (mechanical, algorithmic) prediction procedures: The clinical-statistical controversy. Psychology, Public Policy & Law, 2(2), 293-323.

Hanson, R., & Thornton, D. (1999). Static 99: Improving actuarial risk assessments for sex offenders. Ottawa, ON: Department of the Solicitor General and Her Majesty's Prison Service.

Hanson, R., & Thornton, D. (2003). Notes on the development of the Static-2002. Ottawa, ON: Public Works and Government Services Canada.

Harris, A., Phenix, A., Hanson, R.K., & Thornton, D. (2003). Static-99 coding rules revised - 2003. Ottawa, ON: Corrections Directorate, Solicitor General Canada.

Hogue, T.E. (1993). Attitudes towards prisoners and sexual offenders. Issues in Criminological & Legal Psychology, 19, 27-32.

Johnson, S.L. (2002). An examination of release decisions for sexual offenders: Do biases exist? Forum on Corrections Research, 14(1), 38-40.

---

Come visit us in cyberspace…

You can access every issue of Forum on Corrections Research on the Internet, along with the myriad other research publications we produce. To do so, visit the Correctional Service of Canada website at http://www.csc-scc.gc.ca

The website will allow you to download individual articles or entire issues of FORUM, or even add your name to our mailing list.