Recidivism in criminal behavior—A 10-year follow-up study of 45 adolescent boys identified as sex offenders

Abstract

In this study, we investigated the recidivism rate in a group of 45 men who committed their first sexual offence when they were adolescents. When committing the initial sexual offences during 2002–2005, the participants had a mean age of 16.2 years. We used data from the Swedish National Council for Crime Prevention about criminal offences committed during a 10-year follow-up period. The recidivism rate was 64.4%. No participant had committed sexual or physical offences (i.e., violence against others) exclusively but 15.5% had reoffended in both sexual offences and other types of criminal offences. The group that had committed new criminal offences had to a higher degree experienced parental separation, had an immigrant background, a higher degree of truancy, a higher degree of alcohol or drug abuse, and had at an early age been involved in other criminal offences than sexual ones. Problems such as truancy, alcohol and drug abuse are signals visible outside the family-system, and the need for early interventions seems obvious.

Keywords:

• adolescent sex offenders
• background variables
• recidivism
• self-rating scales
• ten-year follow-up

Studies that focus on adolescents who sexually offend started to appear in the mid-1940s, and by the 1980s research and clinical literature had rapidly increased (Malin, Saleh & Grudzinskas, 2014). It has been suggested that there is a great risk of recidivism in sex crimes among adolescents who have sexually offended (e.g., Becker, Cunningham-Rathner & Kaplan, 1986), and an array of studies have aimed to study potential risk factors for recidivism (e.g., Fortune & Lambie, 2006; Gerhold, Browne & Beckett, 2007).

However, most studies show that the majority of those who sexually offended in adolescence did not continue to sexually offend as adults (Lussier & Blokland, 2014; Nisbet, Wilson & Smallbone, 2004). Among adolescents who have sexually offended, sexual reoffending rates vary between 0% and 42% in different studies. For adolescents who have received treatment, reoffending rates are approximately 10% (Fortune & Lambie, 2006). Caldwell (2010) conducted a meta-analysis of 63 data sets (including 11,219 juvenile sex offenders with a weighted mean follow-up period of 59.4 months) and found a mean sex recidivism rate of 7.1% and a recidivism rate of general criminality of 43.4%. In an 8-year follow-up study based on 351 male adolescent sex offenders (mean age 15.8 years at the first offence), 45% were charged for a new criminal offence, whereof 30% were charged with physical violence, and 10% with sexual abuse (Carpentier & Proulx, 2011). A comparison of recidivism between 40 juvenile sex offenders and 133 nonsex offenders 2 years after release showed that the recidivism rate among the juvenile sex offenders was 3% compared to a 23% reoffending rate in the other group (Calleja, 2015). Hargreaves and Francis (2014) conducted follow-up studies 5 and 35 years after the first sex offence and found that 7% had been reconvicted of a sex offence within 5 years, and 13% had been reconvicted within 35 years.

There are also studies of adults who have sexually offended against children which show that a majority of the adults did not sexually offend as adolescents (Marshall, Barbaree & Eccles, 1991; McKillop, Smallbone, Wortley & Andjic, 2012; Smallbone & Wortley, 2004). For example, a study of 495 adult sex offenders found that 28 (6%) of the men had offended as juveniles (Beaudry-Cyr, Jennings, Zgoba & Tewksbury, 2017).

In a review of longitudinal studies, Reingle (2012) concluded that most studies do not support the idea that adolescent sex offenders constitute a special group that continues to sexually offend. Instead, sex offending adolescents tend to be generalists regarding reoffending rather than specialists. Moreover, Ryan and Otonichar (2016) concluded that most juvenile sex offenders do not later develop paraphilic disorders or commit new sex offences. Instead, they are more like non-sexually delinquent adolescents. Caldwell (2010) suggests that when juvenile sex offenders become older they change from conducting sexually aggressive behaviors to presenting more acceptable forms of sexual behavior, and points out that little is known about the mechanisms behind this development.

Even though there seems to be a relatively small risk of committing new sexual offences among adolescents with previous sex offences, it is important to understand potential risk factors for reoffending, and according to Worling and Långström (2003) there is a need to find more empirical support for supposed risk factors. Several factors on their own or in combinations have been associated with recidivism. Having a history of family dysfunction such as paternal abandonment (Carpentier & Proulx, 2011) and experience of domestic violence especially in combination with low self-esteem (Smith, Wampler, Jones & Reifman, 2005) are factors associated with a higher risk of recidivism in sex offences. Childhood sexual victimization and association with significantly younger children are also risk factors (Carpentier & Proulx, 2011), as well as having a sexual interest in children (Worling & Curwen, 2000). Factors related to the original sex offence have also been studied, and having had younger victims, a higher number of victims, and having offended against a stranger are factors related to a higher risk for recidivism in new sex offences (Carpentier & Proulx, 2011).

Since most adolescents who have sexually offended and who commit new crimes do not reoffend in sex crimes but in other types of crimes, factors associated with general recidivism have also been studied. The use of violence during the first sexual offence and having offended against a stranger were linked with a higher risk of recidivism in violent crimes whereas school delay, a diagnosis of ADD, and association with delinquent peers were linked with a higher risk for overall recidivism (Carpentier & Proulx, 2011). Worling and Curwen (2000) reported that low socioeconomic status, negative parent–child relationship, a history of previous offences, low self-esteem, heightened anger, and antisocial personality were factors associated with recidivism in non-sexual crimes.

Even though research on adolescents who have sexually offended has increased, it is still unclear which specific factors are important in distinguishing between the group that desists and the group that continues to commit offences. Moreover, most recidivist studies have relatively short follow-up periods. Therefore, we aimed to investigate the recidivism rate and types of crimes in young adult men 10 years after they as adolescents committed their first sex crime. We also aimed to explore possible factors associated with recidivism using a broad assessment strategy measuring both background factors such as intellectual difficulties, ethnicity, and experiences of adversities in childhood, as well as using self-report questionnaires previously identified as possibly related to recidivism.

Method

Participants

The participants consisted of 45 adolescent males who had committed sexual offences during the years 2002–2005. In 2003–2007 the participants were in treatment for having sexually offended. Some of the participants were sentenced by court to inpatient compulsory care, others were placed by social services in inpatient or outpatient care. They were approached by Tidefors and agreed to take part in a research study (Tidefors, Goulding & Arvidsson, 2011). The age of the boys ranged from 13 to 22 years with a mean age of 16.2 years (SD = 1.9). Three participants were 19, 20, and 22 years, respectively, but these somewhat older participants were under 18 when they committed sexual offences. During the initial data collection, the participants were asked whether they could be contacted for follow-up studies and all of them consented.

Instruments

At the initial data collection, retrospective data derived from intake assessment files and questionnaire data was collected. Ten years later, data from the Swedish National Council for Crime Prevention was collected regarding criminal offences committed during the 10-year period after the sexual offence.

Intake assessment files

These consisted of data documented when a boy arrived at an institution or became a patient at a psychiatric clinic. The following information was extracted from these files by a psychologist at the institutions or at the psychiatric clinics: ethnic origin, intellectual difficulties (impaired learning abilities), behavioral or acting out diagnosis, alcohol/drug abuse, victim of bullying, having bullied other children, truancy, other types of criminal activities than sexual offences, victim of sexual, psychological or physical abuse, was living or had lived in a foster-home, parental substance abuse, if the boy had a sexually abusive family member, separated parents, known to social services before being identified as a young sex offender, the relation between the boy and the victim(s), and finally if the boy was younger than 14 years at his first sexual offence occasion.

The adolescent sex offender assessment pack (ASOAP)

The Adolescent Sex Offender Assessment Pack (ASOAP) (Beckett, Brown & Gerhold, 1997) consists of 12 questionnaires, some with subscales. ASOAP has been translated to Swedish by Larsson, Linköping, and back-translated to English. ASOAP can be retrieved from the first author. Ten of the subscales were used here: (1) Social Desirability or the Personal Reaction Inventory, which is a 20-item questionnaire measuring the degree to which the subject reports being honest, truthful, responsible, of good intensions, and self-controlled. (2) Self-Esteem/Self Derogation which is an eight-item questionnaire in which the individual answers yes or no to statements regarding how he feels about himself. (3) The Children’s Assertiveness Behavior Scale which consists of 27 situations accompanied by five alternative responses with ratings of assertiveness. The Assertiveness scale measures both over- and under-assertiveness. (4) Emotional Loneliness which is a 20-item scale designed to detect variations in loneliness that occurs in everyday life. (5) The Interpersonal Reactivity Inventory which is a 28-item questionnaire measuring four dimensions of empathy (Perspective Taking, Empathic Concern, Fantasy, i.e., the ability to identify with fictional characters, and Personal Distress, i.e., negative emotions resulting from other’s feeling of distress). (6) Locus of Control which is a 40-item questionnaire measuring the extent to which the individual feels that events are contingent on his behavior and the extent to which the individual feels that events are controlled externally. (7) The Novaco Anger Questionnaire which consists of 73 items measuring partly cognitive mediation of anger, physiological arousal, and behavioral responses involved in anger, partly the sources of provocation. (8) Impulsivity which is a 20-item questionnaire assessing the degree to which the subject experiences himself as being in control of his own actions. (9) The Sexual Matters Questionnaire which is a modified version of the Sexual Knowledge and Beliefs and Sexual Desirability scales, the adolescent form of the Multiphasic Sex Inventory. The questionnaire has two parts. Part A Sexual Knowledge and Beliefs consists of 24 items addressing sexual knowledge and beliefs and part B Social Sexual Desirability consists of 35 items designed to measure social sexual desirability and denial versus openness about sex drives and interests. (10) The Children and Sex Questionnaire which also consists of two parts. Part A, the Cognitive Distortions Scale, is a 15-item scale which assesses an individual’s beliefs about children and their sexuality and part B, the Emotional Congruence Scale, is a 15-item scale which assesses the extent to which individuals can understand, relate to, and identify with what they believe to be the thoughts, feelings, and concerns of children. The internal consistency (Cronbach’s α) and test–retest reliability ranged between .70–.93 and .70–.98, respectively (van Outsem et al., 2006).

Procedure

Ethical committees in six different geographical areas in Sweden reviewed the study. The boys who had sexually offended were recruited for the initial data collection with the help of staff at the institutions or the psychiatric clinics in question. These persons, usually clinical psychologists, also fully informed first the parents and then the boys about the purpose of the study, that participation was voluntarily, and that the boys could drop out at any time they wanted. An informed consent form was administrated and signed, first by the parents, and thereafter by the boys. When both parents and the boy had signed the consent form, Tidefors was contacted to decide on a suitable occasion for data acquisition. The Chain of Word test (Jacobson, 1993) was used to make sure that the participants had the capacity to read and write. This test was administrated before the initial data collection started.

We submitted an application to the Swedish National Council for Crime Prevention in order to retrieve data about criminal offences committed by the participants during the 10-year period following their initial sexual offences. The application was granted after scrutiny of possible ethical problems and we were sent a password secured anonymous file that specified type of offences, date for the offences, and court decisions.

Analyses

The participants were grouped depending on whether they had committed any criminal offences during the 10-year follow-up period or not. Recidivism was thus defined according to convictions in a court of law. The participants were also grouped depending on types of offences they had committed during the follow-up period.

We used t-tests to compare recidivists with non-recidivists regarding continuous variables and the Mann–Whitney U-test in cases of skewed distributions or small samples. We used the Fisher’s exact test for group comparisons on categorical variables. Due to the small sample size, we did not conduct analyses between the sub-groups of offenders. Differences across groups were considered statistically significant if p (two-tailed) <.05. The software used was SPSS 24.

Results

According to the Swedish National Council for Crime Prevention, 29 of the 45 participants (64.4%) had committed new criminal offences during the follow-up period. The offences were divided into three groups; sexual and other offences (seven participants), physical (i.e., violence against others), and other offences (nine participants), and other offences (13 participants). No participant had committed sexual or physical offences exclusively. The number of offences committed ranged between 1 and 79 offences (M = 7.8, SD = 15.2, Md = 2.0). Half of the recidivists had committed over 10 criminal offences.

Ten years ago, when the first data collection took place, half of the 45 boys (n = 22, mean age 16.2 years) had committed sex offences exclusively. Of these boys, 10 had reoffended during the follow-up period and 2 of these recidivists had committed sex crimes, but none of them had committed sex offences exclusively.

We analyzed group differences regarding the same background and offending behavior variables that were described in the Tidefors et al. study (2011), see Table 1. Fisher’s exact tests showed statistically significant differences between the recidivist and nonrecidivist groups regarding truancy, having experienced parental separation, involvement in other criminal activities than sexual offences, abusing alcohol or drugs, ethnic origin, and first sexual offence against a child or against a peer/adult, see Table 1. The effect sizes (ϕ) associated with these variables were of medium to large strength (Cohen, 1988). The variable having been bullied was associated with a medium strength effect size, although the statistical analysis showed a non-significant result, see Table 1.

Table 1 Frequencies of participants in the nonrecidivist and recidivist groups concerning background and offending behavior variables gathered 2003–2007, and group comparison results.

Background variables	Nonrecidivist group (n)	Recidivist group (n)	χ^2 (p)	φ
No intellectual difficulties	6	16	1.29	.17
Intellectual difficulties	10	13	(.353)
No acting out diagnosis	9	21	1.21	.–16
Acting out diagnosis	7	8	(.331)
No truancy	12	8	9.39	.46
Truancy	4	21	(.004)
Not victim of sex abuse	10	17	.06	.04
Victim of sex abuse	6	12	(1.00)
No sex abusive family member	13	20	.80	.13
Sex abusive family member	3	9	(.491)
Not victim of emotional abuse	6	5	2.29	.23
Victim of emotional abuse	10	24	(.161)
Not victim of physical abuse	7	13	.01	.–01
Victim of physical abuse	9	16	(1.00)
Nonseparated parents	8	6	4.13	.30
Separated parents	8	23	(.053)
Not lived in foster-home	12	15	2.33	.28
Lived in foster-home	4	14	(.204)
Not known to social services	7	8	1.21	.16
Known to social services	9	21	(.331)
Not victim of bullying	3	14	3.82	.–29
Victim of bullying	13	15	(.062)
Have not bullied others	12	17	1.21	.16
Have bullied others	4	12	(.341)
No other criminal activities	12	10	6.78	.39
Other criminal activities	4	19	(.013)
No alcohol/drug abuse	16	16	10.09	.47
Alcohol/drug abuse	0	13	(.001)
Parents not abusing alcohol/drugs	14	19	2.55	.24
Parents abusing alcohol/drugs	2	10	(.164)
Both parents born in Sweden	14	15	7.62	.42
At least one parent born abroad	1	14	(.007)
First offence against a child	15	17	6.19	.37
First offence against peer/adult	1	12	(.016)
First offence against sibling	7	14	.08	.04
First offence against nonsibling	9	15	(1.00)
Under 14 at first offence	10	15	.48	.10
Over 14 at first offence	6	14	(.544)

We performed t-tests to investigate group differences between the recidivist and nonrecidivist groups regarding the ASAOP questionnaires, see Table 2. Similarly to the Tidefors et al.’s study (2011), the Victim Empathy scale of the ASAOP was not used due to attrition problems, and the Self-Esteem/Self-Derogation scale had a negatively skewed distribution, so the Mann–Whitney U-test was used to investigate differences on this scale. We found statistically significant group differences regarding the Impulsivity and Sexual Knowledge and Beliefs scales from the ASAOP and these were associated with large effect sizes (Cohen, 1988). The recidivist group had a higher degree of impulsivity and sexual knowledge compared to the nonrecidivist group, see Table 2. The group differences regarding the Locus of Control and Cognitive Distortions scales were statistically nonsignificant but associated with medium effects, see Table 2. The recidivist group had a higher degree of locus of control and a lower degree of cognitive distortions compared to the nonrecidivist group.

Table 2 Descriptives regarding the ASOAP questionnaire for the non-recidivist and recidivist groups, and group comparison results.

ASOAP questionnaire	Nonrecidivist group M (SD)	Recidivist group M (SD)	t (p)	Cohen’s d
Social desirability	51.0 (14.9)	52.0 (13.2)	–.23 (.822)	–.07
Self-esteem	7.0	7.0	197.0	–.13
self-derogation^a	(25.2)	(21.8)	(.393)
Children’s assertiveness	14.4	16.0	–.69	–.21
behavior scale	(7.7)	(7.5)	(.496)
Emotional loneliness	34.6	37.1	–.91	–.29
	(8.5)	(9.0)	(.367)
Interpersonal reactivity	48.1	51.4	–.72	–.23
inventory	(11.5)	(16.7)	(.476)
Locus of control	11.5	14.5	–1.91	–.63
	(3.6)	(5.7)	(.062)
Novaco anger	138.6	143.2	–.60	–.19
questionnaire	(22.9)	(25.4)	(.550)
Impulsivity	6.3	9.7	–2.39	–.76
	(4.2)	(4.7)	(.021)
Sexual knowledge and beliefs	12.9	16.8	–2.30	–.75
	(4.8)	(5.5)	(.027)
Social sexual	19.0	19.1	–.04	–.02
desirability	(6.1)	(6.6)	(.971)
Cognitive distortions	19.7	14.6	1.68	.54
scale	(10.4)	(8.3)	(.101)
Emotional congruence	16.6	18.7	–.58	–.19
	scale(11.4)	(11.2)	(.563)

^a Median and (mean rank) are the descriptive values for this variable which was analysed with the Mann–Whitney U-test and r for the effect size value.

Discussion

The recidivism rate in our material was 64.4%, and 15.5% of the men had reoffended in both sexual crimes and other criminality. The recidivism rate of 64.4% is higher compared to for example the recidivism-rate of 43.4% found in the meta-study by Caldwell (2010). Caldwell also found a mean sex recidivism rate of 7.1% and in our sample a higher proportion, 15.5%, had committed a new sex offence plus other criminal acts. Perhaps the higher recidivism rate in the present study can be explained by our relatively long follow-up period.

In studies, recidivists are sometimes divided into three groups; those who commit new sex crimes exclusively, those who commit other crimes than sex crimes exclusively, and those who commit sex crimes and other crimes. However, we found that although half of the 45 boys initially had committed sex offences exclusively, none of them had relapsed into new sex crimes exclusively at follow-up. The results here are thus in line with other studies showing that the majority of those who sexually offend in adolescence do not continue to sexually offend as adults (Lussier & Blokland, 2014; Nisbet et al., 2004) and those who continue with criminal behavior mostly commit other crimes than sex crimes. Accordingly, most studies do not support the idea that sex offenders are a specialized group. Instead, sex offenders tend to offend sexually occasionally and sex-offending adolescents tend to be generalists regarding reoffending rather than specialists. Thus, our results are in line with what has been labeled the antisocial trajectory hypothesis (Christiansen & Vincent, 2013). Moreover, our results can in some way support Caldwell’s (2010) suggestion that when juvenile sex offenders become adults, they change from coercive sexuality to more acceptable forms of sexual behavior.

When analyzing background data for the 45 participants our results are similar to for example those of Worling and Curwen (2000). We found that the group that had committed new criminal offences to a higher degree had experienced parental separation, had an immigrant background, had had serious truancy, had a higher degree of alcohol or drug abuse, and had at an early age been involved in criminal activities other than sexual perpetration. These relationships point to the fact that the participants who had the most troubled backgrounds were those who continued to live a life afflicted with social problems that included criminality.

In earlier studies, factors related to the characteristics of the first sex offence have been studied and having had younger victims, a higher number of victims, and having offended against a stranger are factors related to a higher risk for recidivism in new sex offences (Carpentier & Proulx, 2011). In the present study, the group who committed new criminal offences had to a higher degree offended against peers/adults 10 years ago. This result might indicate that what we found here was an ongoing criminal lifestyle rather than a propensity to externalize coercive sexuality.

We also found differences in the self-rating scales between those who had relapsed and those who had not. Concerning self-reported impulsivity and degree of sexual knowledge 10 years ago, those who had relapsed scored higher on impulsivity and had a higher degree of sexual knowledge. The higher degree of impulsivity found here could indicate that those who offend and reoffend have a reduced capacity for affect regulation. The result that those who continued a criminal lifestyle had a higher degree of sexual knowledge as teenagers is puzzling and this result needs to be replicated and studied further before it can be discussed. Moreover, the results from the self-rating scales need to be interpreted with caution since this group of participants scored significantly higher concerning social desirability compared to both norms and a comparison group when they answered the questionnaires 10 years ago (Tidefors et al., 2011). It is possible that the results from the self-report measures are compromised due to social desirability.

We found the strongest relationship between relapse-rate and being an immigrant (i.e., being born in another country or having parents born in another country). Having an immigrant background can be related to experiences of segregation and discrimination (Gyberg, Frisén, Syed, Wängqvist, & Svensson, 2018). In Sweden, segregation and marginalization are growing problems that tend to make immigrants synonymous with outsiders. Many immigrant boys might expect little from Swedish society and instead subcultures are formed where macho ideals are praised (Sernhede, 2003). However, in a study of men with refugee backgrounds in Australia, different approaches to adjust to new ways of pursuing intimate relationships were found. For some men, this process was overwhelmingly difficult, for others it was an opportunity (Muchoki, 2015).

The use of court verdicts to measure recidivism in the present study might underestimate the actual reoffence rate, which is a limitation. However, Caldwell (2010) found that the use of arrests versus convictions did not significantly influence the reported rate of sexual recidivism in juvenile sex offenders. A strength of the present study is the relatively long follow-up period of 10 years. However, although we had a relatively long follow-up period, we do not know how long period of time the 45 participants had lived outside institutions after their first sexual offence. It is possible that some participants had spent much less time than 10 years in society and therefore it is possible that the recidivism rate was underestimated. Moreover, the small sample size prevented us from being able to conduct more powerful statistical analyses such as regression analyses, and from conducting analyses between reoffence subgroups.

The results of the present study indicate areas of importance to study further, such as research of living conditions for immigrant children in Sweden. Some of the young men were raised by caregivers whose views of sexuality might have differed from that of the majority in Sweden, raising questions about what it might be like to be an immigrant teenage boy trying to develop his own sexuality. Being an immigrant teenage boy and also experiencing early trauma, in combination with being a “tough guy,” can possibly lead to difficulties understanding the boundary between consensual and nonconsensual sexuality. Furthermore, problems such as truancy, alcohol and drug abuse, and being a victim of bullying are signals visible outside of the family system. For many, an early intervention, before criminality and drugs entered their lives, would have been desirable. Further efforts from authorities and schools to identify such boys in time and provide them with help and support seems of utmost importance. The self-report questionnaire data indicated some areas of interest for further exploration, such as impulsivity, locus of control, cognitive distortions, and sexual knowledge. The relations between these variables and recidivism need to be replicated and further explored. Lastly, the findings here point to the necessity to focus not only on recidivism in sexual offences nor on general criminality. Instead the focus for interventions should be on discovering both those boys who have difficult lives, and those who need support with their view of sexuality in order to enable the possibility to live a decent life.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

This work was supported by Swedish Research Council for Health, Working life and Welfare [grant number 2012-0791].