A Project of the Office of Justice Programs, U.S. Department of Justice

Long Version
Section 2: An Overview of Sex Offender Treatment for a Non–Clinical Audience
Sex Offender–Specific Treatment Outcome Research
30 minutes

(20 minutes)

What Are the Problems?

People are rightly curious about whether sex offender treatment is effective. This is a topic of fundamental importance, because if treatment reduces recidivism rates in sex offenders, this would be a compelling argument in its favor. Likewise, if sex offender treatment does not reduce recidivism rates, then it would be difficult to justify. In this section, we’ll review several recent sex offender treatment outcome studies to find the most accurate answer to the question “Does sex offender treatment work?”

Use SlideUse Slide #2: Does Treatment Reduce Recidivism in Sex Offenders?

Use SlideUse Slide #3: Problems with Measurement of Treatment Effectiveness

Note: Elicit responses from audience.

As we discussed earlier, the primary purpose of sex offender–specific treatment is to protect the community by reducing the likelihood of future sexual assaults.1 We measure the effectiveness of treatment by comparing sex offenders who have received treatment with those who haven’t. Although that sounds simple enough, let’s look at why answering the question of whether sex offender treatment is effective is not as straightforward as meets the eye. What are some of the problems with performing this type of research?

Ask Questions First, there are variations, including some that are quite extreme, in recidivism rates between different sex offender treatment programs. What do you suppose is the recidivism rate of sex offenders, say, over a four to five year period? That is, if you had a mixed group of a hundred sex offenders who were living in the community over a four to five year period, some of whom had had treatment and some of whom hadn’t, how many of them do you think would have a new charge or conviction for a sex offense during that period?

Judging from your answers, some of you will be surprised that the number is as low as it is. This is consistent with the public’s perception of sex offender recidivism rates. In fact, some people believe that all sex offenders reoffend, it’s only a matter of how long it will take them. While it is true that the longer the period of opportunity to reoffend, the greater the number of sex offenders who do, even sex offender reoffense research studies with the longest follow–up periods rarely find reoffense rates over 50%. The rate of new sex offenses committed by known sex offenders, that is, the recidivism rate of sex offenders as measured by new charges or convictions over a five to six year period, was found to be 13.7%. 2

Of course, not all sex offenses are known to authorities or researchers. Therefore, an offender may be considered a non–recidivist when he had, in fact, reoffended if the sexual assault was never reported or otherwise known. This under–reporting is a vexing problem in knowing the “true rate” of sexual offending. Nonetheless, there is no study that has shown that more than half of large groups of sex offenders reoffend. And for our purposes at the moment, answering the question of whether sex offender treatment is effective, the under–reporting issue is moot. This is so because there is no reason to believe the proportion of unknown sex offenses would be any different in the treatment group than the non–treatment group. That is, if the reported recidivism rate were, say, 10% for the treated group and 20% for the untreated group, but the “true rate” of recidivism—the perfectly accurate number of all new sex offenses that were committed as opposed to the known rate of new charges and convictions—were twice that number, the recidivism rate for the treated group would therefore be 20% and the recidivism rate for the untreated group would be 40%. Both groups would have the same relative percentage of underreporting, and the treatment effect in this example would still have cut the recidivism in half. Stated another way, while the problem of under–reporting of sex crimes is a difficulty in many arenas, it’s not much of a problem when we’re comparing groups of treated and untreated sex offenders.

Use SlideUse Slides #4–5: Hanson & Morton–Bourgon, 2004

Slide #4
[Click to Enlarge]
Slide #5
[Click to Enlarge]

Hanson, R.K. & Morton–Bourgon, K. (2004).

In an update of Hanson and Bussière’s 1998 meta-analytic review of sex offender recidivism studies, Hanson and Morton-Bourgon (2004) analyzed 95 studies containing 31,216 sex offenders. This study examined primarily dynamic (changeable) risk factors, rather than the static (unchangeable) factors reviewed in the 1998 study. The purpose of this review was to update the earlier meta-analysis in light of the ongoing research on sex offender risk assessment and to examine those variables that are significant to risk assessment and that were identified as having a weak or controversial effect in the earlier study. Multiple sources of recidivism measures were utilized, including arrest, treatment records, and self-reports. Results showed that the sexual recidivism rate across all studies was 13.7% over a 5-6 year follow-up period; the general recidivism rate was 36.9%. All types of recidivism were predicted by offenders having an unstable, anti–social lifestyle, or lack of self-control. Those individuals with deviant sexual interests, particularly in children, were most likely to reoffend sexually. Additionally, high rates of sexual interest and activities (sexual preoccupation) were significantly related to all forms of recidivism.

Use SlideUse Slides #6–8: Lösel & Schmucker, 2005

Slide #6
[Click to Enlarge]
Slide #7
[Click to Enlarge]
Slide #8
[Click to Enlarge]

Lösel, F. & Schmucker, M. (2005).

Lösel and Schmucker (2005) conducted the first international meta-analysis of both published and unpublished sex offender biological and psychological treatment outcome studies. In this review, 69 studies with more than 22,000 subjects were analyzed—about one-third published since the year 2000 and one-third published outside North America. In this study, recidivism was operationalized as broadly as possible, ranging from incarceration to lapses in behavior. The average recidivism rate was about 11% in treatment groups and 17.5% for control groups; indicating that overall, treatment provides a 37% reduction in sexual recidivism. The authors found that of the psychosocial types of treatments offered to sex offenders in these studies, cognitive-behavioral treatments had the most significant impact on sexual recidivism. Additionally, hormonal and surgical treatment interventions provided significant reductions in recidivism; however these results were partially confounded by methodological errors. Furthermore, in many instances hormonal treatment was provided in conjunction with cognitive-behavioral treatment for those offenders whose sexual arousal played a role in their offending behavior and thus, these treatment “packages” provided significant effects.

Use SlideUse Slides #9–11: Marques et al., 2005

Slide #9
[Click to Enlarge]
Slide #10
[Click to Enlarge]
Slide #11
[Click to Enlarge]

Marques, J.K., Wiederanders, M., Day, D.M., Nelson, C., & van Ommeren, A. (2005).

In one of the only rigorous scientific studies of treatment with adult sex offenders, Marques and colleagues (2005) employed an experimental design to evaluate the Sex Offender Treatment and Evaluation Project (SOTEP). Stationed at Atascadero State Hospital in California, SOTEP operated from 1985 to June 1995. The study was a randomized clinical trial that compared the rearrest rates of offenders treated in an inpatient relapse prevention program (n=259) with the rates of offenders in two (untreated) prison control groups (225 in the volunteer control and 220 in the non-volunteer control group). Final results from a longitudinal investigation revealed no significant differences among the three groups in their rates of sexual or violent reoffending over an 8-year follow-up period. Overall, 22% of the treatment group committed a subsequent sexual offense and 16.2% had a violent reoffense; 20% of the volunteer group reoffended sexually and 16.3% had a subsequent violent offense; and 19.1% of the non-volunteer control group had a sexual reoffense and 15% had a violent reoffense. However, upon closer examination within these groups the authors found that child molesters who did well in treatment recidivated at lower rates, and those who did not demonstrate as much progress in treatment recidivated at higher rates; and higher risk sex offenders who did not progress as far in treatment recidivated at higher rates that other high–risk sex offenders who progressed further in treatment.

While these findings do not necessarily support the efficacy of treating sex offenders, results must be viewed with caution. In interpreting these findings, it must be recognized—as acknowledged by the authors—that the treatment program was developed approximately 20 years ago. Since that time the field has evolved in a number of ways, and there have been many developments that have led to the refinement and enhancement of sex offender treatment and management practices. For example, there is now a much greater understanding of the specific dynamic variables that, if targeted by treatment and supervision professionals, may result in significant reductions in recidivism. In addition, there is a growing acceptance that the impact of interventions is maximized when tailoring the nature and intensity of interventions to the specific risk and needs of offenders, rather than providing the same services to all offenders regardless of risk and need. The design and implementation of the SOTEP project predated these advances in research and practice and, therefore, the findings, considered within the context of these more recent developments in the field, may be understandably less than optimal.

Use SlideUse Slides #12–14: McGrath et al., 2003

Slide #12
[Click to Enlarge]
Slide #13
[Click to Enlarge]
Slide #14
[Click to Enlarge]

McGrath, R.J., Cumming, G., Livingston, J.A., & Hoke, S.E. (2003).

McGrath et al. (2003) recently conducted a study that evaluated a prison-based cognitive-behavioral, relapse prevention treatment program for adult sex offenders in the State of Vermont, which included a community aftercare component. The purposes of this study were to identify the characteristics of men who completed treatment and compare them with those who refused or dropped out of treatment, and then to compare the reoffense rates among these three groups. McGrath et al. identified 195 participants—most (n=90) did not receive any sex offender treatment (no treatment group); 56 participants entered and completed treatment (completed treatment group); and 49 participants entered treatment but dropped our or were terminated from the program (some treatment group). Overall, participants were at risk in the community for an average of about six years. Recidivism data were obtained for all new charges for sexual, violent, or other offenses. Results showed that almost one quarter of the total sample (23%) were found to have committed a new sex offense during the follow-up period. Sex offenders in the completed treatment group had a significantly lower sexual recidivism rate (5%) that both the some treatment group (31%) and the no treatment group (30%). The completed treatment group also had a significantly lower rate of violent recidivism than the no treatment group.

Use SlideUse Slides #15–17: Seager et al., 2004

Slide #15
[Click to Enlarge]
Slide #16
[Click to Enlarge]
Slide #17
[Click to Enlarge]

Seager, J.A., Jellicoe, D., & Dhaliwal, G.K. (2004).

Seager et al. (2004) examined 177 men over the age of 18 who were convicted of a sex offense and were offered the opportunity to participate in a manualized treatment program, in which offender progress was clinically evaluated. Seager et al. hypothesized that positive evaluations of treatment progress and completion would correlate with lower rates of recidivism. Additionally, it was anticipated that those offenders who successfully completed treatment would recidivate at lower rates than those who refused treatment or dropped out. Convictions and charges for violent or sex offenses were used as estimates of recidivism. A total of 81 offenders successfully completed the treatment program, 28 were unsuccessful, 17 offenders dropped out, and 19 refused to participate. Overall, 12% of the offenders in this study were reconvicted for a sex or violent offense and 23% were charged with a new sex or violent offense. Only 4% of successful treatment completers and 7% of unsuccessful treatment completers were convicted for a new sexual or violent offense. Of those who dropped out, refused, or were terminated from treatment, 32% incurred a new conviction and 49% had new charges. Seager et al. found that dropping out, refusing, or being terminated from treatment was related to higher risk for sexual and/or violent offending.

Now let’s go back to the low reported sexual recidivism rates of sex offenders. While the news of low recidivism may be encouraging from a public safety perspective, from a statistical standpoint in comparing treated and untreated groups of sex offenders, low recidivism rates make it nearly impossible to find significant differences between groups. This is because, from a statistical perspective, when the base rates for the study are low as they are with sex offenders (typically between 10 and 25%) and the number of study subjects is also relatively small (say fewer than 200 offenders, which is typical of these studies), the treatment effect has to be very large in order to achieve statistical significance. In fact, the treatment effect has to be greater than 50% with these low base rates and small study samples. Even the best sex offender treatment programs are not so effective that they can reduce sex offender recidivism by 50%. Therefore, statistically their reduction in recidivism is not considered to reach significance. In short, both treated and untreated groups tend to have low sexual recidivism rates. The way around this problem is to increase the study sample size, because as the sample size increases, the difference in recidivism rates between the treated and untreated groups can diminish and still achieve statistical significance. This method—increasing study sample sizes—currently is being done in a sex offender treatment outcome study by combining several samples of offenders into one study. We will be talking about that study shortly.

Use SlideUse Slide #18: Problems with Measurement of Treatment Effectiveness (cont.)

So what are some of the other problems with researching the question of whether sex offender treatment reduces recidivism?

Most sex offense laws are state laws, not federal laws. This means that each state and local jurisdiction manages sex offense crimes in its own way. Different legal jurisdictions have different sex offense laws as well as policies about enforcement, arrest, conviction, plea–bargaining, and so forth. Each state in the U.S. has its own laws about sex offending, and practices can vary widely from locale to locale, especially as they relate to prosecution of some types of sex crimes such as sex between younger and older adolescents. This also makes it difficult to track offenders who move from one state to another. Since we don’t have a federal tracking system in place to ensure that sex offenders who reoffend will become known to officials in their previous locale, sex offenders who appear not to have had subsequent arrests or convictions may have reoffended in another state, unknown to officials and researchers in their previous state of residence.

Another problem with sex offender treatment outcome research is that there can be wide variations in the composition of various treatment groups. For example, if one study is comprised of sex offenders who are maximum security prison inmates and another study involves incest offenders who are on probation in the community, our expectation is that these groups would have very different reoffense rates with or without treatment. The prison sample would likely include offenders who had committed more severe offenses, had perpetrated against more victims, had a greater number of non–sexual offenses, and so forth. Comparisons between these two groups would be like comparing the proverbial apples and oranges.

Likewise, as already noted, the duration of follow–up has a profound effect on recidivism rates. That is, how long the offenders in a study sample are tracked in the community during which they had an opportunity to reoffend. The shorter the follow–up, the fewer reoffenses will have occurred. One study showed that some sex offenders go 20 years before they are rearrested for an additional sex offense (Prentky, Lee, Knight, & Cerce, 1997). For these individuals, they would be correctly classified as non–recidivists for the first 20 years, then they would become recidivists. Therefore, if one study has a follow–up period of five years and another study has a follow–up period of 20 years, it would be difficult to compare them.

Since we are not simply interested in knowing whether the individuals reoffend over a brief period, but ideally over their lifetimes, the best studies have a lengthy follow–up period. Treatment effects that last a lifetime are significantly more valuable than effects that have much shorter duration. However, the longer the follow–up period, the more study sample attrition will occur. In other words, people are lost to follow–up because they move away or for other reasons cannot be located, and this study sample shrinkage makes it difficult to know whether subjects who could not be tracked had similar reoffense rates to those still accessible to researchers.

In addition to these factors, there are problems with definitions. There has not been uniformity among researchers in what constitutes “reoffense” in sex offender treatment outcome studies. The more restrictive the definition of reoffense, the fewer will be the recidivists in the study. At the extremes, for example, if one study only included as recidivists those who were re–imprisoned, while another study included as recidivists everyone who had been charged with a new offense, the latter study would end up with many more recidivists.

Finally, there is little uniformity between sex offender treatment programs in their intensity, duration, location (whether in a prison, a psychiatric facility or in the community), and the treatment methods utilized. Again, this makes studies difficult to compare to one another because they are often addressing quite different programs.

What Do We Do and What Do We Know?

So what are we to do? In spite of all of these problems, it still makes sense to attempt to study sex offender treatment effectiveness. To date, there have been several studies where one sample of sex offenders was provided treatment, and another matched group of very similar sex offenders was provided no treatment, and they were followed for some period of time to find out about any differences in reoffense between the two groups. However, even more useful information would be provided by a study where subjects were randomly assigned to different treatment approaches, but that research has yet to be done.

Although no study has been conducted that compares different treatment approaches within the same study using random assignment of offenders to treatment conditions, there have been some treatment/no–treatment comparisons using matched samples.

Use SlideUse Slide #19–21: Barbaree & Marshall, 1988

Slide #19
[Click to Enlarge]
Slide #20
[Click to Enlarge]
Slide #21
[Click to Enlarge]

Barbaree & Marshall, 1988

Now that we understand the problems with studying sex offender treatment effects, let’s review some of the best studies that have been done to date.

Barbaree and Marshall studied a community sample totaling 126 child molesters, half of whom were treated with cognitive–behavioral methods and half of whom were similar to the treated group, but did not receive treatment. They included both official and unofficial measures of recidivism (including reconviction, new charge, or unofficial record). With a four–year follow–up, they found the treated group had quite different recidivism rates from the untreated group. For the child molesters whose victims were girls outside the family, 43% of the untreated group reoffended compared to 18% of the treated group. For the child molesters who targeted boys outside the family, 43% of those individuals who did not obtain treatment reoffended, whereas 13% of the treated group recidivated. For the incest offenders whose victims were female, 22% of the untreated group recidivated versus 8% of the treated group. Although this was a small study, the differences in recidivism between the treated and untreated groups are quite dramatic.

Use SlideUse Slide #22: Rice, Quinsey, and Harris, 1991

Rice, Quinsey, and Harris, 1991

Let’s look at other research findings. In a 1991 study, Rice, Quinsey, and Harris evaluated the reconviction rates of 136 extrafamilial child molesters who had been housed in a maximum security psychiatric hospital, and over an average of a six year follow–up period, found essentially no differences in recidivism between the untreated and the treated groups.

Hall, 1995

In 1995, Hall performed a meta–analysis of 12 sex offender treatment outcome studies. A meta–analysis is a study of studies; that is, rather than examining one study sample, a number of independent samples are evaluated. There are many advantages to meta–analysis, principal among them is that it creates much larger sample sizes, which aids statistical analysis by avoiding the small sample size problems discussed earlier. Another advantage to meta–analysis is that it increases confidence in the findings of the study because it analyzes several different studies, thereby reducing the possibility of idiosyncratic findings.

Hall examined 12 studies with a total of 1,313 sex offenders. In this study, recidivism was defined as a formal legal charge for a new sex offense after completion of treatment for the treatment group. The mean length of treatment was 18.5 months and the mean follow–up period across the 12 studies comprising the meta–analysis was 6.85 years. Hall found a small but significant overall reduction in recidivism in treated versus untreated offenders. The overall recidivism rate for treated sex offenders was 19%, versus a recidivism rate for the untreated offenders of 27%. There was quite a bit of variability in the outcomes of the different studies in the meta–analysis, however. Generally, the studies that showed the greatest treatment effect were those that had longer follow–up periods, had higher overall rates of recidivism, were outpatient rather than inpatient or prison–based treatments, and used cognitive–behavioral and/or anti–androgen hormonal medication.

Use SlideUse Slide #27: General Accounting Office, 1996

U.S. General Accounting Office, 1996

In 1996, the U.S. Government General Accounting Office summarized 22 reviews of research on sex offender treatment covering 550 studies between 1977 and 1996. The findings of the report were that the results of sex offender treatment are promising, but inconclusive.

Alexander, 1999

In 1999, Margaret Alexander reviewed 79 sex offender treatment outcome studies. Consistent with the growing awareness that attempting to determine if sex offender treatment reduces recidivism is too broad a question, she examined recidivism rates for treated and untreated sex offenders in various sub–groups, including offender age, victim age, offense type, type of treatment, location of treatment, the years in which treatment occurred, and length of follow–up. Recidivism typically was defined as a rearrest for a new sex offense. There were a total of 10,988 study subjects in the 79 studies she examined, but of course there were many fewer subjects in each of the sub–groups.

Alexander found that the subset of offenders who participated in relapse prevention treatment programs had a rearrest rate of 7% compared to 18% for untreated offenders. She also found that 20% of the treated rapists were rearrested, compared to 24% of the untreated rapists.

Of the subset of child molesters in Alexander’s study, she found 14% of the treated group were rearrested compared to 26% of the untreated child molesters. Interestingly, treatment effectiveness was especially strong for exhibitionists. Of the subset of exhibitionists in her study, 20% of the treated group were rearrested compared to 57% of the untreated exhibitionists.

Overall, when all 10,988 subjects were combined, Alexander found that 13% of the treated group were rearrested compared with 18% of the untreated sex offenders.

Hanson et al., 2002

In perhaps the most comprehensive study to date, the Association for the Treatment of Sexual Abusers has established a Collaborative Data Research Project with the goals of defining standards for research on treatment, summarizing existing research, and promoting high quality evaluations. As part of this project, researchers are conducting a meta–analysis that includes 43 treatment studies and more than 9,000 offenders. The first report from this project provides findings that indicate an overall positive effect of treatment, in terms of both sexual recidivism (12% of the treatment subjects compared to 17% of the control group subjects) and general recidivism (28% of the treatment subjects compared to 39% of the control group subjects).

Use SlideUse Slide #34–35: Hanson et al., 2002

Slide #34
[Click to Enlarge]
Slide #35
[Click to Enlarge]

Seto and Barbaree, 1999

The effects of sex offender treatment are variable according to many factors having to do with offender age, offense type, treatment type, and a host of other variables (including offender characteristics, such as impulsivity and psychopathy among rapists, and degree of sexual preoccupation with children among child molesters). We’ve talked as if sex offender treatment can either reduce recidivism or have no effect. But can sex offender treatment ever increase the likelihood of sexual recidivism?

Among the factors that might affect recidivism that has not been extensively researched is treatment behavior. That is, we might logically assume that sex offenders who behave cooperatively in treatment sessions, submit high quality work assignments, and are rated by treatment providers as highly motivated and having made positive changes would reoffend at lower rates than sex offenders who do not exhibit these apparently positive characteristics.

We have long known that there is a particular type of criminal offender, known as a psychopath, who typically commits a large number of crimes, that are often violent in nature. Among the hallmarks of psychopathy are an absence of conscience, remorse, and guilt or empathy; and the presence of criminal versatility, grandiosity, superficial charm in relationships with others, an excessive need for stimulation, pathological lying, conning and manipulation, impulsivity, irresponsibility, and promiscuous sexual behavior.

A very interesting study was conducted by Seto and Barbaree (1999), which examined the effects of sex offender treatment according to the presence or absence of traits of psychopathy. In their study of 283 sex offenders, they evaluated the relationship of clinical ratings of treatment behavior, as indicated by treatment session behavior, homework quality, motivation, and change achieved. Paradoxically, they found that good treatment behavior was unrelated to general recidivism but was associated with higher sexual and/or violent recidivism. Furthermore, they found that men who scored higher in psychopathy and were rated higher in treatment behavior were the most likely to reoffend. Their study subjects were divided into four groups: low in psychopathy and good in treatment behavior; high in psychopathy and good in treatment behavior; low in psychopathy and poor in treatment behavior; and high in psychopathy and poor in treatment behavior. The offenders who were high in psychopathy and had been rated as good in treatment behavior were more likely to commit a new offense of some kind, and much more likely to commit a new sexual and or violent offense, than the offenders in any of the other three groups in this study.

Use SlideUse Slide #36: Seto and Barbaree, 1999

Use SlideUse Slide #37: Serious Sex Offender Recidivism Related to Treatment Behavior and Psychopathy

One possible explanation for this counter–intuitive finding is that this is a group of psychopathic sex offenders who are socially skillful and good at manipulation—including manipulation of treatment providers—and they appear to be making excellent treatment progress. However, when they are in the community, they utilize these same skills to gain access to victims. A more alarming explanation is that some sex offenders may learn manipulative or other skills in treatment that actually increase their risk for recidivism. It has certainly been suggested that some sex offenders hone their sex offending skills during treatment by learning the modus operandi of others. Whatever the explanation, treatment apparently can have the opposite of its intended effects, and this appears to especially be the case for psychopaths. It is generally agreed that cognitive–behavioral treatment methods that are effective with non–psychopaths have the potential to make psychopaths more likely to reoffend.