Risk assessment in juvenile sex offenders is compromised by the fact that little conclusive evidence exists on risk factors that predict sexual recidivism. It has been difficult to find such evidence because, given the low sexual recidivism rates, very large samples are required. In addition, extant studies have employed varying definitions and follow-up periods and over-relied on clinical samples. This study identifies factors associated with desistance, general (non-sexual) recidivism, and sexual recidivism in a group of 1,525 juvenile sex offenders at moderate risk of reoffending who were followed for an average of almost 10 years.

At 7.7 per cent, the sexual recidivism rate is low, and the general recidivism rate (68 per cent) is much higher. Most sexual reoffending occurs before age 20 and entails mainly rape and sexual assault. Sexual recidivism is associated with low educational performance, conduct disorder, and aggression regulation problems. No associations are found with offender typology or with factors associated with sexual deviance. The main points of this study are as follows:

• • Juvenile sex offenders mainly commit their offences (all offending) in youth.

• • About 8 per cent of juvenile sex offenders reoffend sexually.

• • This continuation of sexual offending (in adulthood) is part of an active and versatile criminal career.

• • Only psychological deficits and educational level were found to predict sexual reoffending, but effects were not very strong.

• • Different subgroups may be associated with different risk factors for sexual recidivism.

(p. 221) This essay is organized as following. After an introduction in section I, section II discusses empirical and theoretical accounts of the criminal career parameters found for juvenile sex offenders. Risk factors associated with sexual reoffending are discussed in section III. Section IV looks at heterogeneity within the group of juvenile sex offenders. In section V the methodology of the study reviewed in this essay is described, followed by the results section (section VI). The final section consists of the conclusion and discussion of findings.

I. Researching Juvenile Sex Offending

Criminological research has shown that some of those who offend in their youth continue to do so in adulthood (Piquero, Farrington, and Blumstein 2003). Specifically, juvenile sexual offending is often regarded as a precursor of continued sexual offending in adulthood. This widely held belief ‘once a sex offender, always a sex offender’ is also reflected in the strict criminal justice policies for sexual offenders implemented in various countries. Most of these policies apply to both juveniles and adults, reflecting this continuity hypothesis, which implies that today’s juvenile sex offenders will be tomorrow’s adult sex offenders. Recently, however, an increasing number of studies have challenged this continuity hypothesis (Nisbet et al. 2004; Vandiver 2006; Zimring, Piquero, and Jennings 2007; Zimring et al. 2009; Van den Berg, Bijleveld, and Hendriks 2011; Lussier et al. 2012; Lussier and Blokland 2014). These mostly prospective longitudinal studies showed that actually relatively few juvenile sex offenders reoffend sexually in adulthood. This would indicate that sexual offending is more of a ‘passing phase’ rather than an announcement of chronicity.

However, it appears that some juvenile sex offenders do become chronic sex offenders, in line with theories that postulate heterogeneity in juvenile sex offenders (Groth 1977). And while various classification schemes for juvenile sex offenders have been proposed, little is known about this small group of offenders who continue sexual offending into adulthood. Part of this knowledge gap may be due to the fact that the subset of offenders who do continue to offend appears to be only a very small proportion of all juvenile sex offenders; therefore, very large samples may be required to investigate them in sufficient detail. Also, many studies investigating juvenile sex offenders had follow-up periods that were too short (Caldwell 2010).

The current study aims to increase our knowledge of the criminal career development of juvenile sex offenders, and specifically to investigate those who persist in sexual offending. Using a large sample of 1,525 juvenile sex offenders, this study will first present a comprehensive view of the entire criminal career (onset, frequency, variety in offending and desistance) to establish the extent to which juvenile sexual offending is a precursor of continued sexual offending in adulthood. Second, the sexual recidivists in the sample will be identified, and we will investigate risk factors that are associated with continued sexual offending. Lastly, risk factors identified by risk assessment instruments (p. 222) will be compared to risk factors for the current sample of juvenile sex offenders. This comparison is vital because a valid assessment of the risk of sexual recidivism for a juvenile sex offender contributes not only to protecting the public but also to ensuring that the rights and welfare of juvenile sex offenders are protected. Therefore, this study holds crucial information for policy, risk assessment, and theory development regarding juvenile sex offenders.

II. The Criminal Career of Juvenile Sex Offenders

The criminal career approach is aimed at describing and explaining continuity of offending over the different stages across the life course. For the general offender population, criminal career studies showed that early-onset offending may be related to persistent and serious offending over the life course (Piquero, Farrington, and Blumstein 2003). Therefore, it is not surprising that juvenile sex offending is often regarded as a precursor to adult sex offending. The criminal career approach has only recently been applied in the study of juvenile sex offenders, describing and explaining persistence and versatility in offending and desistance from (sexual) offending (e.g., Laws and Ward 2011; Lussier and Blokland 2014). Most empirical evidence so far has found that discontinuity in sexual offending is the norm for juvenile sex offenders (Lussier and Blokland 2014). Only a small group of juvenile sex offenders reoffend sexually after the initial sexual offence (e.g., Fortune and Lambie 2006; Caldwell 2010). On average, sexual reoffending rates were found to average 10 per cent (e.g., Fortune and Lambie 2006; Vandiver 2006; Zimring, Piquero, and Jennings 2007; Zimring et al. 2009; Caldwell 2010; Lussier et al. 2012), although they vary considerably across studies. For example, in a review of six studies on recidivism rates, Fortune and Lambie (2006) found sexual reoffending rates to range from 0 per cent to 42 per cent. In a review and meta-analysis of 63 datasets including 11,219 juvenile sex offenders, Caldwell (2010) found the sexual recidivism rates to vary between 0 per cent and 18 per cent. This sizeable variation in sexual reoffending rates may be due to different follow-up periods, differences in the measurement of the sexual reoffending, and differences in the characteristics of study samples (Worling and Långström 2006).

The follow-up period is of interest here as it informs us whether sexual reoffending continues into adulthood or whether any sexual reoffending is also part of an adolescence-limited criminal career (see Moffitt 1993). Caldwell (2010) separated the overall average sexual recidivism rate of 7.08 per cent into adult sexual reoffending and juvenile sexual reoffending. For adult reoffending Caldwell (2010) found 20 studies that had an average sexual recidivism rate of 6.5 per cent over a follow-up of 73.8 months (SD = 46.5). Within studies that focused on juvenile sexual reoffending only, Caldwell (2010) found 15 studies with an average follow-up of 30.5 months (SD = 10.2) (p. 223) (or approximately 2.5 years) and a sexual recidivism rate of 9.9 per cent. The follow-up period as well as the reoffending rates differed significantly between the adult and juvenile reoffending group. For the adult group, longer follow-up periods were associated with lower sexual reoffending rates. Caldwell (2010) explained this difference by assuming that changes in the developmental stages (from adolescence to young adulthood and adulthood) influence sexual behaviour. In a large birth cohort study, Zimring, Piquero, and Jennings (2007) found the prevalence for adult sex offending to be 3.2 per cent (n = 99). Of the 99 male adult sex offenders in this study, 8.5 per cent had been adjudicated for a juvenile sex offence. Therefore, Zimring, Piquero, and Jennings (2007) concluded that, in general, a juvenile sex offence is not a precursor of adult sexual offending. The best predictor of adult sexual offending was in fact found to be the number of all police contacts (i.e., for any crime) as a juvenile. In a more recent study, Lussier and Blokland (2014) focused on whether juvenile sex offenders become adult sex offenders. Again, a small subgroup of about 5 per cent of the juvenile sex offenders were found to continue sexual offending in young adulthood. The individuals in this group were characterized by the highest number of arrests for both sexual and non-sexual offending during the follow-up period. Lussier and Blokland (2014) noted that this finding underlines versatility in the criminal career of this small group.

These studies show, all in all, that juvenile sexual offending appears to only rarely extend into adulthood. Those who do reoffend sexually appear to be characterized by frequent overall offending and thus an overall serious criminal career (see Zimring, Piquero, and Jennings 2007; Lussier and Blokland 2014). As such, juvenile sex offending and specifically continuation of sexual offending in adulthood might be, rather than part of a specialized sexual offending career, associated with general anti-social tendencies (Van der Put et al. 2013). As such tendencies are shared with many other offenders, most of them non-sexual, this raises the question of what other characteristics are predictive of repeated sexual offending that may extend into adulthood.

III. Risk Factors Associated with Sexual Reoffending

Identifying the risk factors associated with sexual reoffending is crucial for judicial and clinical practice: assessing the likelihood of a juvenile’s risk of recidivism is important when deciding on an appropriate punishment or treatment. Many experts have, however, stressed the difficulty of assessing the risk of sexual reoffending among juveniles due to the different developmental stages these youths may be in (e.g., Prentky and Righthand 2003). Within the empirical literature it is emphasized that different developmental stages in life are associated with different risk factors (e.g., Worling and Långström 2006). Juveniles move through particularly turbulent and critical stages in life (adolescence), characterized by rapid and substantial psychological, biological, and (p. 224) social changes. These changes cause the risk factors for sexual reoffending to change in a similarly rapid pace, complicating the assessment of the risk of sexual reoffending.

Sexual reoffending risk is still frequently appraised through clinical assessment. However, such clinical assessments are generally regarded as lacking transparency and replicability (Worling and Långström 2006). Standardized risk assessment instruments have supplemented clinical assessment, as these are regarded as much more ‘objective’ and transparent. For the prediction of juvenile sexual reoffending risk these instruments have not performed well: their predictive validity is low, possible because most have been built from instruments for adults that focus on risk factors relevant in adulthood instead of in adolescence (Viljoen et al. 2008; Knight, Ronis, and Zakireh 2009). Viljoen and colleagues (2008) examined and compared the performance of the Juvenile Sexual Offense Recidivism Risk Assessment Tool II (J-SORRAT-II), the Structured Assessment of Violence Risk in Youth (SAVRY), and the Juvenile Sex Offender Assessment Protocol-II (J-SOAP-II) in predicting violent and sexual reoffending risk in youths convicted of and treated for a sexual offence. All three instruments employ different factors to estimate the recidivism risk. The J-SORRAT-II is focused on the static factors of past victimization, offence, treatment, and school history, while the SAVRY and the J-SOAP-II employ static as well as dynamic items. Predicting the risk of general violent reoffending is the aim of the SAVRY, so that the scales focus on general historical factors (e.g., offending) and environmental and individual characteristics (e.g., impulsiveness, parental monitoring). The J-SOAP-II aims to predict sexual reoffending and is therefore more focused on sex drive and preoccupations and on the manner in which this drive is negotiated. In addition, the J-SOAP-II employs measures of impulsive anti-social behaviour and the so-called measure of intervention (e.g., are there feelings of guilt?). This diversity in the items and scales used, across and within instruments, illustrates how much is still unclear about sexual offending risk assessment for juveniles.

Viljoen and colleagues (2008) found that none of the tools significantly predicted sexual reoffending, while the J-SOAP-II and the SAVRY predict non-sexual violent reoffending. The most important finding was that age at the sexual offence seemed to influence the predictive ability of the instruments. This underlines once more that age and the associated developmental stage are characterized by different risk factors. A risk assessment instrument that is possibly less affected by the influence of age is the Estimate of Risk of Adolescent Sexual Offence Recidivism (ERASOR), which estimates the short-term risk of reoffending (maximum of 1 year risk assessment) (Worling and Curwen 2000). This instrument was built from an adult risk assessment instrument and employs five scales: sexual interests and attitudes, historical sexual assaults, psychosocial functioning, environmental variables, and treatment. In a literature review, Hempel and colleagues (2013) found the ERASOR and the J-SOAP-II to be the most accurate in predicting short-term sexual reoffending compared to the SAVRY, J-SORRAT-II, the Juvenile Risk Assessment Scale (JRAS), and the Hare Psychopathy Checklist: Youth Version (PCL:YV). However the predictive validity of the J-SOAP-II and ERASOR for long-term sexual recidivism was still found to be insufficient (Hempel et al. 2013).

(p. 225) In a bid to shed more light on the risk factors for sexual reoffending, Worling and Långström (2006) reviewed empirical literature published between 1980 and 2002 concerning risk factors for sexual reoffending among juvenile sex offenders. Factors were classified into four categories: supported, promising, probable, and unlikely risk factors. The risk factors for sexual reoffending with strongest empirical support included deviant sexual interests, prior sanctions for sexual offending, sexual offending against multiple victims, social isolation, uncompleted offence-specific treatment, and sexual offending against a stranger victim. Markedly, most of these factors are offence related (e.g., prior sanctions for sexual offending, sexual offending against multiple victims, or a stranger victim). While these five risk factors have empirical support, Worling and Långström (2006) also identified promising risk factors in need of better empirical grounding: problematic parent–adolescent relationships and attitudes supportive of sexual offending. A more recent systematic review by McCann and Lussier (2008) of risk factors associated with sexual reoffending by juvenile sex offenders reported on ‘consistently supported’ factors, such as a history of sexual offending, deviant sexual interests, and sexual offending against a stranger victim. McCann and Lussier (2008) also found a younger age at the sexual offence to be associated with an increased likelihood of sexual reoffending. Contrary to the findings of Worling and Långström (2006), McCann and Lussier (2008) found previous non-sexual offending, male and child victims, and the use of threats to predict sexual reoffending. Again, these are offence-related characteristics.

In summary, we see that the literature and risk assessment practice have proposed a broad array of factors that may be associated with sexual reoffending. Very little consistent support has been found for factors that may be regarded as part and parcel of the aetiology of sexual offending. Only factors that may be regarded as a marker for sexual deviance and offending-related characteristics have been found to play a role. The fact that so little empirical progress has been made in this regard is not surprising given the methodological limitations of most studies, such as short follow-up periods and small research samples (Spice et al. 2013). The study discussed in this essay is less burdened by such issues, as the study sample consists of 1,525 juvenile sex offenders followed over a period of an average of almost 10 years, well into adulthood. This study assessed the extent to which a large number of characteristics of juvenile sex offenders measured in adolescence predict sexual reoffending.

IV. Heterogeneity in Juvenile Sex Offenders

Juvenile sex offenders were found to constitute a heterogeneous group, consisting of a number of distinct groups with different aetiologies and risk profiles that lead to different criminal careers (e.g., Becker 1998). Various typologies for juvenile sex offenders (p. 226) have been proposed in the empirical literature. Some of these typologies have been based on underlying motivation for the sex offence (e.g., Knight and Prentky 1993), offending history (e.g., Butler and Seto 2002), personality profiles (e.g., Worling 2001), or offending characteristics (e.g., Långström and Grann 2000; Bijleveld and Hendriks 2003; Hunter et al. 2003). The classification based on offending characteristics has received most attention in empirical research, specifically the typology based on victim age (child abuser and peer abuser) and that based on the distinction between those who committed the offence alone or with at least one co-offender (solo offenders and group offenders).

Juvenile sex offenders who abused a child (child abusers) and those who abused a peer or adult (peer abuser) were found to differ not only on offending characteristics but also on personality profiles. For instance, child abusers were found to lack social skills and to have low levels of self-esteem compared to peer abusers (Hunter et al. 2003). Similar results were found by Gunby and Woodhams (2010), as the child abusers in their sample experienced deficits in self-esteem and greater social isolation. Also, it has been reported that the prevalence of disorders (e.g., pervasive developmental disorders, neuroticism and psychopathology scores) is higher for those who abused a child than for those who abused a peer (e.g., Van Wijk 1999; Van Wijk et al. 2005; Hendriks 2006). The groups of peer and child abusers were also found to differ in recidivism risk. Hagan and colleagues (2001) compared 50 peer abusers with 50 child abusers and 50 persons convicted of offences other than sexual ones and found child abusers to display significantly higher rates of sexual recidivism. Hendriks (2006) also found that reoffending rates were associated with the juvenile sex offender type. In summary, child and peer abusers were found to differ on risk factors such as offending characteristics and personality profiles, and child abusers were found to have significantly higher sexual recidivism rates. Therefore, the continuity of sexual offending might differ between these groups, implying that distinct treatment and prevention programmes might be suitable for these subtypes (Hunter et al. 2003; Hendriks 2006).

The second offending-based classification is derived from the finding that a substantial number of all juvenile offences are committed in a group. For juvenile sex offenders several studies looked into the difference between juveniles who committed the sexual offence alone (solo offenders) and those who committed the offence with at least one co-offender (group offenders). Overall, group offenders were found to have relatively normative personality profiles (e.g., Bijleveld and Hendriks 2003; Bijleveld et al. 2007), although intelligence and school achievements were often below average (Bijleveld et al. 2007; ‘t Hart-Kerkhoffs et al. 2011). Group offenders had less sexual offending in their criminal history than solo offenders did (Bijleveld et al. 2007).

All in all, the most comprehensive typology with the most empirical support appears to be the typology developed by Hendriks (2006). In this typology Hendriks (2006) distinguished between three groups of juvenile sex offenders: those who sexually abused a child with no co-offenders (solo child abusers), those who abused a peer without a co-offender (solo peer abusers), and those who committed a sexual offence (p. 227) with at least one co-offender (group offenders). Empirically, these three groups are distinct in terms of offending characteristics, personality profiles, and recidivism rates (Hendriks 2006).

V. Method

VI. Results

VII. Conclusion and Discussion

Using a large sample of juvenile sex offenders who had been prospectively followed well into adulthood, this study aimed to expand the existing knowledge of the criminal career and recidivism risk of juvenile sex offenders. Objective registered information on offending and personal characteristics was obtained. Our findings show that only a small subgroup of juvenile sex offenders reoffended sexually. This corresponds with previous empirical findings (e.g., Fortune and Lambie 2006; Vandiver 2006; Zimring, Piquero, and Jennings 2007; (p. 236) Zimring et al. 2009; Caldwell 2010; Lussier et al. 2012). A majority of juvenile sex offenders were criminally active only in youth. In previous studies Van den Berg, Bijleveld, and Hendriks (2011) and Lussier and colleagues (2012) (on a more problematic part of the sample investigated here) already showed evidence that most juvenile sex offenders are adolescent-limited offenders, only partly fitting the model developed by Moffitt (1993). In this much larger study we again found evidence for an adolescence-limited criminal career pattern for juvenile sex offenders.

The sexual recidivists, however, did not follow an adolescence-limited criminal career pattern. This small subgroup (about 8 per cent of the juvenile sex offenders) were found to continue sexual offending in adulthood. The frequency with which they did so decreased somewhat with age, whereas the frequency with which they committed other offence types appeared to increase in adulthood. Overall, the sexual recidivists were characterized by the highest number of (all) offences in youth as well as adulthood. This leads us to believe that juvenile sex offending and continuation of sexual offending in adulthood is part of an active and versatile criminal career. Other empirical accounts also showed that sexual recidivists have a frequent and serious criminal career (Zimring, Piquero, and Jennings 2007; Lussier and Blokland 2014). Specifically, Van der Put and colleagues (2013) found similar results in their study and postulated that sexual reoffending is due to general anti-social tendencies rather than sexual deviance. Our study lends empirical support to this hypothesis, as we found that for sexual recidivists conduct disorder and problems with aggression are more common than in non-sexual recidivists and desisters. These psychological deficits are associated with anti-social tendencies, possibly supporting the statement by Van der Put and colleagues (2013). Another hypothesis is that this small group of sexual recidivists might be life-course-persistent offenders, as described in the dual taxonomy by Moffitt (1993). These offenders are hypothesized to follow a long, versatile, and serious criminal career, and indications of such a career for the sexual recidivists were found in this study. To support this hypothesis, studies with even longer follow-up periods than ours would have to be conducted.

The second aim of the study was to assess the validity of commonly used risk factors for sexual recidivism. We selected offending and personal characteristics based on empirical, clinical, and theoretical evidence. However, we uncovered only very weak associations between a select group of variables and sexual recidivism: only psychological deficits and educational level were found to predict sexual reoffending, and again the effects were not very strong. One explanation for this in a sense unexpected result may be that our sample comprised a diverse group of juvenile sex offenders who had been referred mostly through the criminal justice system, while previous empirical investigations used mostly small clinical samples. Thus, it might be that different factors predict sexual recidivism in clinical and criminal justice samples. Regardless, the lack of corroboration for risk factors does show that risk assessment instruments for juvenile sex offenders likely do not have broad applicability. This lack may also be regarded as yet another manifestation of heterogeneity: different subgroups may be associated with different risk factors for sexual recidivism and may need different treatment. In the current (p. 237) study we investigated heterogeneity using the typology by Hendriks (2006) in our comparisons, but we found no evidence for an association between offender type (child abusers, peer abusers, and group offenders) and sexual recidivism. Some association was found between offender type and non-sexual recidivists and desisters. Desisters were found to be more often child abusers—but this seems counterintuitive, since child abusers are regarded as having the highest risk of sexual reoffending.

This study has important strengths as well as limitations. First, we had a prospective, rich, and unique dataset consisting of long-term objective and detailed information on offending and personal characteristics. Sexual recidivism in juvenile sex offenders is hard to study because reoffending is relatively rare. The large number of sample members, and thus the fairly large number of sexual recidivists, makes this dataset uniquely suited to explore the criminal career of juvenile sex offenders more elaborately over the life course. The fact that we had a good variety, from severely problematic offenders (deemed to need lengthy inpatient treatment) to offenders who were not deemed to need any tailored intervention, adds to the generalizability of our findings.

Several limitations to the study are also noteworthy. First, the data on risk factors were limited, as not every treatment file contained similar information. This led to missing values, adding up to such an extent that we could not perform multivariate analysis. Second, we were able to investigate only those offences and risk factors that made it to official records and as such were part of the official, registered, side of the juvenile sex offenders’ lives. It is likely that more sexual reoffending took place. Much more in-depth research is needed to understand the mechanisms responsible for the association between juvenile sexual offending and continuation of sexual offending in adulthood. Therefore, future studies and analyses should attempt to include measures of self-reported (sexual) offending. Given the difficulty of gathering such data on sex offending, however, innovative methodologies may be warranted.