Home Up  

[Doc. List E7]     [Newsletter E6] 


A meta-analysis, reviewed by Dr Frans Gieles



'Once a thief, always a thief' is a Dutch proverb. It's not yet a legend, but the belief that 'a sex offender will always keep being a sex offender' is widely spread. At least, convicted sex offenders are compulsory bound to undergo some kind of treatment. At worst, people want to keep them in lifelong custody. These measures are based on the belief mentioned.

The Dutch professor Frenken has frequently been quoted in the newspapers and other media. He always gives a recidivism rate of 90% or more. Thus, he expects that 90% or more of convicted sex offenders will relapse. For this reason, compulsory treatment is seen as necessary for every convicted offender. The courts and the treatment center believe him - he's a professor and thus an expert. But is it true that 90% reoffend?

In this article I will review an article about a meta-analyse of 61 recidivism studies. The authors conclude to a mean recidivism percentage of 13.4% for sexual offenders, which is lower than the mean overall recidivism percentage for all offenders. Certainly much lower than Frenken's statement.

Let's first have a look at the title, the authors and the summary the authors made of the article.

Predicting Relapse: A meta-Analysis of Sexual Offender Recidivism Studies,
R. Karl Hanson and Monique T. Bussière,
Department of the Solicitor General of Canada.
In: Journal of Consulting and Clinical Psychology,
Edited by the American Psychological Association,
1998, Vol. 66, No. 2, pp 348-362


"Evidence from 61 follow-up studies was examined to identify the factors most strongly related to recidivism among sexual offenders. On average, the sexual offense recidivism rate was low (13.4%; n = 23.393). There were, however, subgroups of offenders who recidivated at high rates. Sexual offense recidivism was best predicted by measures of sexual deviancy (e.g., deviant sexual preferences, prior sexual offenses) and, to a lesser extent, by general criminological factors (e.g., age, total prior offenses). Those offenders who failed to complete treatment were at higher risk for reoffending than those who completed treatment. The predictors of nonsexual violent recidivism and general (any) recidivism were similar to those predictors found among nonsexual criminals (e.g., prior violent offenses, age, juvenile delinquency). Our results suggest that applied risk assessments of sexual offenders should consider separately the offender's risk for sexual and nonsexual recidivism."


The study provides a quantitative review of the sexual offender recidivism literature. The method of review is a meta-analysis, a generally accepted method of research review that can overview many studies and thus large samples, which yield high statistical power. All participants were sexual offenders. Three types of recidivism have been examined: sexual, nonsexual violent, and general (any).

61 Studies from several countries between 1943 and 1995 have been examined. The reported follow-up periods ranged from 6 months to 23 years, with a mean of 66 months. All studies used the matched, longitudinal follow-up design, which is known as the best available design. All studies were found to be reliable, indicating overall acceptable levels of diligence in identifying recidivists. If de differences between the studies were great, they have been made comparable by several statistical techniques. The 61 studies provided information about 28.972 sexual offenders.


A short explanation in advance

To conclude to a recidivism rate, one can count the number of reoffenders among the participants. The rate is a figure between 0% and 100%.

To know which factors influence the reoffending, comparisons have been made between one factor, reoffending, and many other measured factors. A factor is a force that can have influence (e.g., insight). A factor has to be measured by some method. The outcome of the measurement is a variable (e.g., an intelligence quotient) If such a variable correlates high with reoffending, it is seen as a predictor.

Correlation is the coherence of two variables (e.g., intelligence quotient and school success). The level of correlation is reflected in r, a figure between +1.00 (if it rains, the streets are always wet) and - 1.00 (if it rains, the streets are never dry). The significance of this figure depends on the amount of observations or participants. The more observations, the more significance. Therefore, the number of participants is usually given after the r with the letter n.

To give some more explanation before, phallometric assessment is a method to measure the swelling of the penis at the moment the person sees certain pictures. This method intrudes in the most private parts of the person: not only his penis, but also his feelings. It's a quite 'hard' method. In the library of this web site, you can find research that find out that 26% of the male participants reacted to 'pedophilic stimuli' with an erection.


In this review, I follow the language of the authors ('deviant interests', 'child molesters'). Keep in mind that, in the English speaking world, a 'child molester' can be a person, who violently has raped a little girl, but also a person who possessed one picture of a nude child and also a teenager who kissed a girl's breasts at the school yard.

Average recidivism rates

On average, the sex offense recidivism rate was 13.4% (n = 23.393). The average rate for rapists was 18.9% (n = 1.839), for child molesters it was 12.7% (n = 9.603).

For nonsexual violence, the average recidivism rate was 12.2% (n = 7.155); for child molesters it was 9.9% (n = 1.774), but for rapists it was 22.1% (n = 782).

When recidivism was defined as any reoffence, the rates were higher: 36.3% overall (n = 19.374), for child molesters 36,9% (n = 3.363) and for rapists 46.2% (n = 4017).

Predictors of sexual offense recidivism

The article gives a long table of 71 measured factors. Only some of these correlate with sexual reoffending. Contrary to the popular belief, being sexually abused as a child was not associated with increased risk (r = -0.01, which is not significant).

The strongest predictors of sexual offence recidivism were measures of sexual deviance (as the authors name it); sexual interest in children (measured by phallometric assessment) was the strongest (r = 0.32). Lower predictors were 'antisocial personality disorder', 'number of prior offenses' and 'failure to complete treatment (r = respectively 0.14, 0.13 and 0.17). The sole developmental history variable related to sexual offense recidivism was a negative relationship with the mother (r = 0.16).

Predictors of nonsexual violent and general (any) recidivism

These predictors were the same risk factors common to general criminal populations. These recidivists tended to be young, unmarried, and of a minority race. They also engaged in diverse criminal behavior, the abuse of alcohol and were likely to have antisocial or psychopathic personality disorders.

Sexual criminal history was only weekly related to general recidivism. Overall, the clinical presentation and the treatment history variables showed small to moderate correlation with general recidivism. Here also, a negative relationship with the mother was a risk factor.

Predictors or risk factors

Sexual recidivism was best predicted by measures of sexual deviancy. In contrast, other forms of recidivism were best predicted by criminal history. Psychological symptoms were, on average, unrelated to any form of recidivism. Negative clinical representation (e.g., low motivation for treatment or denial of any problem) was related to general, but not to sexual recidivism. Finally, failure to complete treatment appeared to be a consistent risk marker for both sexual and general recidivism.

Predicting relapse by combined risk scales

To predict relapse, one should not measure only one variable, but combinations of variables that mark the risk. No variable was sufficiently related to justify its use in isolation. It appeared that clinical assessment (interviews, tests) failed as a valid predictor (the r was about 0.10). In contrast, statistical risk prediction scales appeared to be a better predictor (r about 0.45). One of these prediction scales, the SIR scale (Statistical Information on Recidivism) could predict general recidivism better than sexual recidivism (r = 0.41 and 0.09 respectively). Another scale, the VRAG (Violence Risk Appraisal Guide), predicted general violent recidivism with an r = 0.47, but sexual recidivism with an r = 0.20 (n = 159). In other words: sexual recidivism is difficult to predict.


The findings contradict the popular view that sexual offenders inevitably reoffend. Only a minority of the total sample (13.4% of 23.393) was known to have committed a new sexual offense within the average 4.5-year follow up period.

This 13.4% is a mean; so some of the 61 studies concluded to a lower rate, other studies (e.g., with a longer follow up period) to a higher rate, but the recidivism rates never exceeded 40%.

Sexual reoffence is difficult to predict, but the strongest predictors were deviant sexual interests, prior sexual offenses and deviant victim choices (boys outside of the family). Most of the psychological variables failed to predict reoffence, except severe personality disorders. A low clinical presentation (e.g., low remorse and denial) failed to predict sexual recidivism. Failure to complete treatment, however, was a significant predictor.

The results suggest that sexual offenders may differ from other criminals. For nonsexual offending, sexual and nonsexual criminals seem much the same, but separate processes appear to contribute to sexual offending.

In this study, measures of subjective distress had no relationship with any type of recidivism; the average correlation was near zero with no significant variability. No measure of a transient state like subjective distress can predict any recidivism years later. This holds on the inter-subjective level. Within the subject, however, subjective distress can trigger a sexual offence cycle.

A remarkable finding is that offenders who failed to complete treatment were at increased risk for both sexual and general recidivism. Reduced risk could be due to treatment effectiveness; alternatively, high-risk offenders may be those most likely to quit, or be terminated, from treatment.

Almost all the predictors of sexual offence appeared to be historical and stable variables. Historical factors cannot change and sexual preferences are difficult to change. Consequently, such variables cannot be used to assess treatment outcome or monitor risk to the community. The most dynamic factor was treatment attendance.

The authors end their article with a list of risk factors that not yet have been adequately researched (because of follow up research requires many years), but which are generally believed as being risk factors. They advise to research these factors in future studies. Here is this list:

- the use of sex as a coping mechanism,
- associations with other sexual offenders,
- attitudes tolerant of sexual crimes,
- heterosocial perception deficits and
- unfulfilled intimacy needs.


We know that these believed risk factors are used in the current treatment practice. The 'therapists' try to diminish the risk by combating, for example, the 'associations with other (potential) sexual offenders' - read: to meet the members of the Dutch ped workgroups - and the 'attitudes tolerant of sexual crimes' - translated: reading Brongersma's book. Would future research conclude that these factors are beliefs only, or facts?

Prof. Frenken is wrong with his believed 90% recidivism rate; it is only a belief, not a result of research. I have never seen any reference to any research report that supports his 90%. Careful meta-analytic research results in an average recidivism rate of 13.4% generally, and 12.7% for 'child molesters'.

Thus, the automatic referral to a compulsory treatment has to be changed by deciding each case on an individual basis. The automatic lifelong custody and the automatic notifying of the community after the release of a convicted sexual offender needs to be changed as well. This is all based merely on a belief rather than on knowledge of the facts. Careful research has given us a look at the facts. One of the facts is that sexual reoffending is difficult to predict.

Can we predict whether the teenage boy at the schoolyard will kiss a girl's breasts again? We can expect that he would do this again; but we hope he would do it with the girl's consent and pleasure. So is human nature. Can we predict whether a man would possess another picture in the future? Supposedly, he will and so add a new 'criminal act' to the statistics. Can we predict that the violent rape of a young girl shall happen twice? We cannot but we hope it will not repeat itself.

The weakness of this careful study is this: there is no difference made between one type of 'child molester' and another. To make the difference was not possible in this meta-analyse, because of the lack of difference in the 61 reviewed studies.

The 'therapists' at the current treatment centers are extremely determined to change the beliefs of their patients. Shouldn’t they also be strong enough to have a critical look at their own beliefs?


[Doc. List E7]     [Newsletter E6] 

Home Up