An Exploration of the Subtypes of Male Youth Adjudicated for Sex Offenses: A Latent Class Analysis Based on Personal,Offense,and Victim Characteristics

Abstract

This study intended to explore possible variations among youth adjudicated for sexual offenses based on personal criminogenic factors, offense, and victim characteristics. Utilizing a data set collected from the juvenile court files in Turkey (n = 460), the Latent Class Analysis revealed that the study sample included three different subgroups with distinct features: “non-delinquent, peer victim-targeting,” “non-delinquent, younger victim-targeting,” and “delinquent, versatile” youth adjudicated for sex offenses. The first two of these groups were similar in terms of having low levels of delinquency, while the third group included the lowest number of youth with significantly broad delinquent activity patterns. These findings were in line with the results of previous studies, and the implications were discussed for future research and policy development.

Keywords

youth juvenile sex offense latent class analysis criminogenic factors Turkey

Sex offenses committed by minors have been observed to occur less frequently than sex offenses committed by adults. In the United States (US), only about 19% of all sex offenses were committed by juveniles in 2020 (FBI, 2021). This rate was 18% in Australia (Warner & Bartels, 2015) and 17% in England and Wales (Stevens et al., 2013; see also Ryan, 2018, for a more extensive review on the prevalence of sexual offending among youth). Turkish Judicial Statistics (2021) indicated that 13.1% of all sexual offenses, and 17.3% of all rape offenses, were committed by youth in 2020 in Turkey; this represents a 5% increase in the number of youth adjudicated for sex offenses (YAFSO)¹ over the past decade (from 5,108 in 2010 to 5,364 in 2020). In that same time frame, according to the same set of official statistics, the percentage of adjudicated sex offenses involving a youth defendant increased from 8.8% to 13.1% in Turkey. Sex offenses involving youth defendants cause particular concerns in light of the extent of the harm they inflict on their victims and those issues related to their specific characteristics and needs that may differ from adults adjudicated for sex offenses (Przybylski & Lobanov-Rostovsky, 2017). These concerns include how to treat YAFSO and prevent their future sexual and non-sexual offending behaviors. As such, in recent years, there have been increasing efforts to explore YAFSO empirically (Budd & Bierie, 2018).

Earlier studies on YAFSO examined criminogenic factors leading to their delinquent behavior (i.e., DeGue et al., 2013; Dwyer & Letourneau, 2011; Finkelhor et al., 2009; Fox, 2017; Hackett et al., 2013; Ryan, 2018; Ryan & Otonichar, 2016; Veneziano & Veneziano, 2002; Yoder et al., 2018), past victimization experiences (Fox, 2017; Karsten & Dempsey, 2018; Rasmussen & Miccio-Fonseca, 2007), the characteristics of their victims and offenses (Brown, 2019; Calleja, 2015; Christiansen & Vincent, 2013; Fox, 2017; Riser et al., 2013; Ryan & Otonichar, 2016), and their differences from other youth adjudicated for other offense types (Buker & Erbay, 2020; Fox, 2017; Margari et al., 2015; Pullman et al., 2014; Seto & Lalumière, 2010; Yoder et al., 2018). Through these studies, it has been determined that YAFSO do not constitute a homogeneous group but a heterogeneous one, resulting from differences in terms of criminogenic risk factors, victim or offense characteristics, and overall delinquent development (youth adjudicated for only sex offenses vs. multiple-type offenses; Fox, 2017; Fox & Delisi, 2018; Hunter, 2018; Margari et al., 2015). These prior studies exploring possible classes/types of YAFSO have utilized various approaches, but a recent trend toward that goal has been to employ a more empirical and self-directing approach known as “latent class analysis (LCA).” LCA reveals unobserved heterogeneity in a given group and helps to identify qualitatively different subgroups of individuals sharing a common profile (Hagenaars & McCutcheon, 2002). Understanding these subgroups could help in efforts to develop more effective intervention and prevention strategies for YAFSO (Fanelli, 2018; Riser et al., 2013). The current study aimed to contribute to the emerging research efforts designed to explore possible subgroups of YAFSO by analyzing data retrieved from the official court records of a group of and adjudicated youth in Turkey. In that sense, it is the first research endeavor, to our knowledge, testing the YAFSO heterogeneity hypothesis with a Turkish juvenile sample that has aimed to contribute to the body of prevailing comparative analyses, which is identified as a limitation of the current state of knowledge on the subject matter of YAFSO (Beaudry-Cyr et al., 2017; Hart, 2016).

Known Classes of Youth Adjudicated for Sex Offenses

Earlier efforts to understand possible subtypes of YAFSO originated with applying various theoretical perspectives based on personality differences of the YAFSO observed in clinical settings. For instance, O’Brien and Bera (1986) identified seven different types of YAFSO and provided case examples for each of those types. These types were categorized as (a) “the naive experimenter,” (b) “the under-socialized child exploiter,” (c) “the pseudo-socialized child exploiter,” (d) “the sexual aggressive offender,” (e)” the sexual compulsive type, “(f) “the disturbed impulsive,” and (g) “the group-influenced” type. Similarly, around the same time, Smith et al. (1987) identified four personality types in their sample of YAFSO: (a) a socially immature and isolated group (the largest group), (b) an over-controlled group, (c) an impulsive group, and (d) an emotionally disturbed group. Later on, Worling (2001) identified distinct types of personalities of YAFSO similar to Smith et al.’s (1987). In Worling’s (2001) study, “over-controlled/reserved” and “confident/aggressive” types presented fewer problematic features, but “antisocial/impulsive” and “unusual/isolated” types had “more pathological” features with an increased likelihood of committing other kinds of offenses. Along the same lines, Richardson et al. (2004) and Oxnam and Vess (2006, 2008) identified different subtypes based on personality differences among their sample of youth adjudicated for sex offenses. In these studies, the largest group of youth displayed personality features such as appearing withdrawn and having a socially inadequate personality and a poor social degree of awareness.

Along with the studies focusing on the personality features of the YAFSO, several earlier studies focused on the distinct characteristics of the offenses they committed and the individuals they targeted. In one of the early efforts with that orientation, for instance, Becker et al. (1986) examined a group of youth referred to a clinic due to a sex offense. They classified the participants as “pedophiles” if they mostly acted upon younger victims and as “rapists” if the victims were about the same age as or older than the perpetrators. Becker et al. (1986) also examined the nature of the offenses in their study. Accordingly, they identified subgroups of YAFSO committing “consensual incests” (see also O’Brien [1991] for a particular classification of youth targeting siblings), “mooning,” “voyeurism,” and “frottage.” Reviewing the findings of 13 studies, Graves (1993) reported that the YAFSO examined were classified as “sexually assaultive (targeting peers),” “pedophilic (targeting younger children),” or “undifferentiated (targeting indifferently)” based on the YAFSO-victim age differences reported in the reviewed studies.

An age-based classification of YAFSO seems to represent a further continuation of the existing efforts to classify adults adjudicated for sex offenses and explore possible variations in the level of criminogenic risk factors for and the needs of those youth targeting children compared to the others targeting same-age or older victims (Aebi et al., 2012; Robertiello & Terry, 2007). Seto and Lalumière (2010), for instance, a meta-analysis of prior studies have indicated that those studies testing the hypothesis mentioned above have tended to find that those YAFSO targeting younger victims were more likely to have a history of sexual abuse compared to the YAFSO who tends to target same-age or older victims, while other studies are reporting no significant difference in that measure between the two groups (Fanniff & Kolko, 2012). The earlier studies examining the differences between the YAFSO targeting younger versus peer/older victims reported mixed findings on the mental health conditions, socialization, and general delinquency/antisocial behavior of the members of these two types of YAFSO. In that sense, Hunter et al. (2003) reported that YAFSO targeting younger victims were more likely to display psycho-social functioning deficits and depression. Likewise, t’Hart-Kerkhoffs et al. (2009) reported more internalizing and (psychosexual) developmental problems among YAFSO targeting younger victims, but no significant differences between the two groups in terms of externalizing behaviors. Aebi et al. (2012) reported significant differences between YAFSO targeting younger victims and the YAFSO targeting peers or older victims in an all-male sample. The YAFSO in the former group (ones victimizing younger children) were more likely to prefer male victims, use less aggression, and be younger than the latter group members. However, they indicated that classifying YAFSO with a victim-age focus was ineffective. This classification did not yield differences with possible other classification approaches (i.e., based on co-offending status or non-sex offense status). Fanniff and Kolko (2012) also reported differences between the two groups regarding using force against victims, targeting strangers rather than acquaintances, and victimizing primarily males. Fanniff and Kolko’s (2012) study indicated that YAFSO who victimize children were more likely to target males and related victims. In contrast, YAFSO with more peer/adult victims were more likely to lack parental monitoring and re-offend. However, they also concluded that classifying YAFSO based on the individual instead of offense characteristics could be more promising.

As psychopathological characteristics have commonly been cited for explaining the practice of committing sex offenses among youth (Fox, 2017; Seto & Lalumière, 2010), those factors were also utilized to assess possible subgroups of YAFSO in previous studies. Using a model linking psychological functioning to one’s sexual deviances, Gamache et al. (2012) identified six different clusters of YAFSO in their study group with psychological features including impaired reality, psychopathy, and narcissism. Using a psychological and behavioral index, Dunton (2020) identified four distinct groups of YAFSO (about 25% of the study group was not adjudicated): (a) a sexually aggressive group, (b) a disturbed revenge group, (c) a disturbed group, and (d) a revenge group. Among those, youth classified under the “disturbed revenge” group were more likely to have behavioral and psychological problems, along with those YAFSO classified as belonging to the “disturbed aggressive” group. On the other hand, the “sexually aggressive” group included the highest numbers of youth, but they demonstrated the lowest behavioral and psychological problems.

A group of earlier studies further determined groupings of various types of YAFSO from a criminal career development perspective. Butler and Seto (2002), for instance, grouped YAFSO as either “sex-only offenders” or “sex-and-other offenders.” As such, the YAFSO in the first group are those who commonly commit sex offenses and refrain from other types of offenses. However, in the second group, sex offenses are among the different offenses a juvenile commits. Nisbet et al. (2010) also made a similar distinction of YAFSO and pointed out two distinct groups (a) “specialist sex offenders” if the YAFSO in their group had no prior nonsexual offending and (b) “versatile offenders” if they had previous or concurrently occurring non-sexual offenses. McCuish et al. (2016) argue that such a classification is limited as the assignment of adjudicated youth to either of these groups is made at an early stage of possible criminal career development. Thus, it fails to address possible subsequent changes in life throughout adolescence and adulthood. Supporting that argument was the finding that 88.2% of the youth initially identified as youth adjudicated for “ex-only” offenses in their study committed other crimes when they reached the age of 23. Following up on a group of youth adjudicated for sex and non-sex offenses from age 12 to 32, Lussier et al. (2012) pointed out that a group of YAFSO in their study group was “adolescent limited,” meaning that their sexual offenses continued only throughout adolescence, but another distinct group of YAFSO (high-rate slow disasters) continued their non-sexual and sexual offending after adolescence. Using a similar approach, Cale et al. (2016) designed a retrospective longitudinal study to examine the offending trajectories of the YAFSO and identified differences between a group of YAFSO that were defined as criminally “versatile” and “non-versatile” YAFSO. Over half of their sample was classified as “rare” and “non-versatile” because most of the youth in that group were mainly adjudicated for a sex offenses and did so only throughout their adolescence. On the other hand, the smallest group in their sample was identified as being “high-rate chronic” YAFSO. They started offending at an earlier age, generally with non-violent and violent non-sex offenses, followed by sex offenses.

The studies mentioned above have all contributed to the current knowledge on the heterogeneity of the YAFSO, especially regarding between-class differences. However, the methodologies employed in those studies had limitations in terms of fostering an understanding of how each factor applied in the various YAFSO classifications contributed to the distinction of the determined classes. In response to those methodological limitations, recent studies aiming to understand how YAFSO could be classified empirically utilized an approach known as Latent Class Analysis (LCA), and applied this approach to determine data-driven classes/types of YAFSO. This approach enables researchers to describe qualitative variations across the latent classes based on regularities and patterns revealed by the interaction of various indicators in data (Brown, 2019). Utilizing LCA, Fox and Delisi (2018) pointed out four distinct types of male (non-disordered, impulsive/unempathetic, early-onset/chronic, and male-victim) and two types of female (non-disordered and (b) female-victim) YAFSO. The emerging groups in their study displayed differences in past sexual victimization, prior criminal (arrest) record, psychological or behavioral features, and age of the first delinquent activity. For instance, most of the male YAFSO (54%) fell under the “non-disordered” group and displayed the lowest level of overall criminogenic risk factors, while the “male-victims” group was the smallest in size (10%), but the members of that group carried the highest level of criminogenic risk factors such as past victimization and psychological problems. Utilizing LCA, again, Barra et al. (2018) identified five distinct groups of varying male YAFSO based on the risk factors extracted from the Adverse Childhood Experiences (ACE) framework. Similar to what Fox and Delisi (2018) found, they also pointed out that the largest YAFSO group in their study had minor risk factors (0.55 ACE on average), while the smallest group had the highest level of risk (7.55 ACE on average). In a more recent study, Baglivio and Wolf (2021) used the LCA approach to classify sex offenses committed by juveniles based on the ACE and victim characteristics. They identified six groups: (a) schoolmate victims, (b) stranger victims, (c) acquaintance victims, (d) sibling victims, (e) diverse victims, and (f) other relative victims. Their study provided further information on how ACE significantly impacted distinctions between these classes. For instance, the positive relationship between sexual abuse history and victimizing siblings and household substance abuse and more diverse and multiple victimizations was noteworthy in their findings.

Using sex-offense-specific variables (age/sex of the victim, age of the adjudicated youth at the first-sex offense, use of penetration/force during the sex offense), Brown (2019) also utilized LCA to explore possible subtypes of YAFSO. As such, Brown also aimed to address the limitations of utilizing the criminogenic risk factors that might overlap with general delinquent behaviors (e.g., Barra et al., 2018; Fox & Delisi, 2018) as a basis of YAFSO classification. Brown (2019) identified four distinct groups of YAFSO; (a) child victims/nonviolent (48%), (b) female peer victims-only (20.5%), (c) male child-focus (16.5%), and (d) early starter/multiple victim (14.8%). While Brown’s basis of YAFSO classification was different from Fox and Delisi (2018) and Barra et al. (2018), the findings were similar in the sense that the members of the largest group, “child victims/nonviolent” carried the lowest criminogenic risk factors. On the contrary, the “early starter/multiple victims” group had the fewest number of YAFSO, but they were characterized by various criminogenic factors and displayed more problematic sexual offending patterns (multiple victims, targeting both children, and teen/adult victims).

As presented above, existing literature on the classification of youth who were adjudicated for sex offenses has three broad bases to assess the differences between these youth: personal characteristics, offense characteristics, and victim characteristics, as visualized in Figure 1.

Figure 1.

Attributes used to classify youth adjudicated for sex offenses.

Current Study

The current study strove to improve the knowledge base regarding the possible subtypes of YAFSO in several ways. First of all, it is one of the rare studies on YAFSO conducted in a non-western context (i.e., Wood et al., 2000—South Africa), especially on the classification of YAFSO, and, thus, provides a new perspective on the types of YAFSO in comparison with the findings of the earlier studies conducted in the US and other western contexts. Furthermore, it is one of the rare studies (Buker & Ebay, 2020; Şengül et al., 2012) on the subject matter exploring sex offenses in a Turkish sample. Among those studies, Yildiz (2012) reported that 26% of the individuals adjudicated for sex offenses this researcher examined in a prison setting were under 18. Şengül et al. (2012) reviewed the court files of a group of Turkish youth adjudicated at courts and found that 39.8% of the cases examined involved sex offenses. However, these two studies were not informative in explaining the characteristics of the YAFSO in the Turkish context. On the other hand, Buker and Erbay (2020) specifically focused on the individual and offense characteristics of male YAFSO in a Turkish sample. They found the victims of the YAFSO in their study sample were primarily female, younger than the adjudicated youth, and commonly known to the adjudicated youth (family, partners, and friends). Buker and Erbay (2020) also reported how YAFSO were different from other youth adjudicated for various other offense types regarding individual and family characteristics. Considering the evident lack of scientific study on the subject matter in that context, an essential objective of the current research was to further inform researchers and policymakers specifically interested in Turkish YAFSO while providing a broad comparative perspective to the international audience generally interested in the subject matter.

Second, above and beyond the particular contributions made due to the unique features of the data source, this study aimed to advance the knowledge on the classification of the YAFSO by utilizing a novel analytical approach known as Latent Class Analysis (LCA). While various studies have attempted to provide a sound knowledge basis on how YAFSO could be classified through multiple approaches as reviewed above, utilization of LCA has been relatively recent, and the number of studies utilizing LCA to classify YAFSO has been limited (Baglivio & Wolf, 2021; Barra et al., 2018; Brown, 2019; Fox & Delisi, 2018).

Lastly, this study strove to advance current knowledge on the possible subtypes of YAFSO by utilizing data compiled from an adjudicated/convicted youth sample (as opposed to a clinical sample), which includes both personal and the offense/victim characteristics. In a recent study, Brown (2019) pointed out that utilizing only criminogenic factors while exploring possible subtypes of the YAFSO (i.e., Barra et al., 2018; Fox & Delisi, 2018) could be limited as those factors also predict non-sex offenses committed by youth. In this study, criminogenic factors and offense/victim characteristics specific to the sex offenses (i.e., victim-YAFSO age difference, type of coercion used) are utilized together to explore if these two groups of commonly utilized predictors of the group membership in the previous YAFSO typology studies could drive a significant separation among the YAFSO subtypes in the current sample. In other words, this study explores if a group of YAFSO-specific criminogenic factors (i.e., psychopathology, exposure to violence, and academic success) and a group of offense/victim characteristics could act synchronously to identify qualitatively different latent classes of the YAFSO.

Methods

Data

The inclusion criterion for the subjects in the current study who were known to have a “youth adjudicated for sex offenses” status was based on how the Turkish legal system defined “youth (child)” and “sex crimes” with the #5395 Child Protection Code (CPC)² and the #5237 Turkish Criminal Code (TCC)³ where every individual under the age of 18 was considered a “child.” The CPC requires every child defendant to be adjudicated by a special juvenile court (though some exceptions may apply in certain circumstances). According to these laws, the minimum age of criminal liability is 12, and until the age of 14, youth have the lowest level of limited criminal liability. Between the ages of 15 and 18, they have a higher but still limited level of criminal liability. Articles 102 to 105 in the TCC include “Offenses against Sexual Immunity,” including sexual assault (rape), child molestation, statutory rape, and sexual harassment. Only the sexual assault (Article 102) and child molestation (Article 103) convictions were included when choosing the study sample in an effort to provide a clear and global description of a “sex offense.”

The study data were collected retrospectively by examining the court files of the juveniles adjudicated in Istanbul Juvenile Felony Courts, one of the largest judicial jurisdictions in Turkey with about 3.5 million residents.⁴ The court files provided the researchers with rich information on the offense and personal characteristics in a given case and further information on the YAFSO’s past criminal history and psycho-social aspects reported through the “Comprehensive Social Examination (CSE)” which is a legally mandated process conducted through a personal interview with the adjudicated/prosecuted youth. To quantify data from the case files and CSE reports therein, a data collection instrument (spreadsheet) was first developed, including available information in the court files. Then a coding strategy was developed and implemented to enter the available information from the court documents into the spreadsheet. Examining the 3,301 total cases adjudicated in that juvenile court between 2006 (the first year when the court was established) and 2018, the researchers identified 460 sex offense trials that had resulted in a conviction and included valid responses aligned with the study variables. In all these cases, the youth was a male who had committed a single sex-related offense over his lifetime. Those case files were then examined, and the selected target variables were coded using the data collection instrument. To assure intercoder reliability, 70 case files were randomly selected and re-coded by another coder, and the results indicated 99.1% consistency between the coders.⁵ The rate of consistency was much higher for the items like age, educational level, and gender since they were retrieved from the official records included in the court files.

Variables

The LCA model utilized two sets of variables as potential predictors of class membership among the YAFSO: personal criminogenic factors and offense/victim characteristics. As required for employing the LCA approach, all variables were binary coded where 1 (Yes) indicated a positive response to a given victim/youth characteristic, and 0 (No) indicated the lack of this feature. The first set of variables included; being a school dropout, smoking, using alcohol, having a non-traditional family structure (all but biological parents live together with the child), being 15 years of age or older, being exposed to domestic violence, a lack of familial communication, at or below-average education level (compared to other YAFSO in the study population), a risky lifestyle (reporting at least one of the following conditions: drug use, having a delinquent friend, or runaway history), having a criminal record (non-sexual), and having a psychological disorder as documented during the court process. The second set of variables included characteristics related to the victims and the cases. In that group were the three variables identifying the victim-YAFSO age difference: younger victim status (three or more years younger than the YAFSO), older victim status (three or more years older than the YAFSO), and peer victim status (all but older or younger victim). In addition, the following variables were in that group: non-acquaintance victim status (all but family, partner, relative, and friend), male victim, and using physical violence or threat to kill to force the victim for sex (see Table 1).

Table 1.

Descriptive Statistics for the Study Variables (n = 460).

Variables (categorical—Yes/No)	Explanation	Yes%
School dropout	Whether or not dropped out of a formal education system	65.20
Criminal record	Whether or not having a non-sexual criminal record other than the adjudicated sex offense	23.70
Smoking	Whether or not smoke cigarettes regularly	68.90
Alcohol	Whether or not consume alcohol	28.50
Non-traditional family structure	Live with biological parents—traditional; all others—non-traditional	32.60
YAFSO age 15 years or older	Age in years	63.30
Exposure to domestic violence	Any kind of domestic violence mentioned by the YAFSO in the social examination process	25.90
Lack of familial communication	Whether or not the YAFSO shared problems with family members	42.80
At or below-average education level	Whether or not being at the formal education level of their age group	29.30
Risky lifestyle	Having friends who use drugs or have criminal records or having a runaway history or abusing the drugs	41.10
Psychological disorder	Whether or not diagnosed any kind of psychological disorder by the psychiatrist before the offense	48.00
Younger victim	Whether or not the victim is three or more years younger than the YAFSO	36.30
Peer victim	Whether or not the age gap between victim and YAFSO is or lower than 3 years	59.10
Older victim	Whether or not the victim is three or more years older than the YAFSO	4.60
Non-acquaintance victim	Whether or not the victim is a stranger (all but family, partner, relative, and friend)	18.70
Use of physical force or threat to kill	Whether or not the YAFSO used physical force or threat to kill the victim during the offense	51.70
Male victim	Whether or not the victim is male	27.20

Analytical Strategy

The structure and characteristics of possible YAFSO subtypes in the study sample were predicted by including 17 variables in the LCA, as described in the previous section. The LCA is an evolving analytical strategy used in many applications such as disease subtypes, marketing research, sociology, psychology, education, and criminology. The broad objective of the LCA is to extrapolate the structure of the latent variables or classes from the observed variables. As such, LCA is used to identify possible unobserved (latent) classes moving from the individual characteristics (as responses to a set of categorical [binary] variables) of the subjects in a sample (Weller et al., 2020). The classification of the individuals into unobserved (latent) classes can display the entire univariate or multivariate distribution of potential predictor variable(s) as a mixture of a finite number of component distributions as well as a finite mixture model (Goodman, 1974; Lazarsfeld & Henry, 1968).

The latent class model without explanatory variables assumes a latent variable Z where each possible outcome of a response y₁,y₂,. . .,y_T (from variables Y₁, Y₂,. . ., Y_P, ) and each category z of Z. If the number of classes is K then the latent class assumes the joint multinomial probabilities are (Agresti, 2018):

\begin{array}{l} P (Y_1 = y_1, \dots, Y_p = y_P) = \sum_(z = 1)^K (\prod_(t = 1)^ \\ P P (Y_t = y_t | Z = z)) P (Z = z) \end{array}

(1)

An extension of this model allows covariates to predict latent class membership (Bandeen-Roche et al., 1997; Dayton & Macready, 1988). Covariates are included in the latent class model via their effects on the prior probabilities of latent class membership. This model is called the latent class regression model.

The analysis in this study was performed using the R package polCApolka (Linzer & Lewis, 2011). To assess the models’ goodness of fit, we began by fitting several models by adding a class each time. Statistics such as the Akaike information criterion (AIC) and Bayesian information criterion (BIC) are also commonly utilized to compare the models and to determine the ideal number of latent classes that can also be explained theoretically. The best models should have the lowest values of BIC and/or AIC while BIC is often more useful for latent class models since it balances the model’s goodness of fit and conveys parsimony.

Results

Ten latent class models were fitted. Figure 1 presents the models with an increasing number of classes, from 1 to 10. It shows the AIC and BIC values of each model. The AIC decreases when the number of latent classes increases, but BIC values decrease rapidly from a model with one class to three classes and become stable after the model of four classes around the value of 8,600 and become slightly higher with seven classes and more. The LCA model with three subgroups (BIC = 8,606.07, AIC = 8,387.11) was determined to be a reasonable model for the current study since it balances the goodness of fit (smaller BIC) and parsimony criteria (not a complex model; Table 2).

Table 2.

Conditional Probabilities Across YAFSO Latent Classes (n = 460).

	Class 1	Class 2	Class 3
Estimated class population shares → predictor variables↓	30%	28%	42%
School dropout (Yes)	0.53	0.94	0.55
Criminal record (Yes)	0.08	0.65	0.08
Smoking (Yes)	0.53	0.99	0.61
Alcohol (Yes)	0.10	0.73	0.13
Non-traditional family structure (Yes)	0.23	0.51	0.27
YAFSO age 15 or older (Yes)	0.56	0.80	0.58
Exposure to domestic violence (Yes)	0.21	0.47	0.16
Lack of familial communication (Yes)	0.47	0.19	0.55
At or below-average education level (Yes)	0.23	0.46	0.23
Risky lifestyle (Yes)	0.20	0.98	0.18
Psychological disorder (Yes)	0.34	0.90	0.31
Younger victim (Yes)	0.94	0.29	0.00
Peer victim (Yes)	0.00	0.61	1.00
Older victim (Yes)	0.06	0.10	0.00
Non-acquaintance victim (Yes)	0.26	0.30	0.06
Use of physical force or threat to kill (Yes)	0.54	0.67	0.40
Male victim (Yes)	0.44	0.23	0.18

The elements of the three-subgroup model were examined using the conditional probabilities of the response given a latent class. Table 3 provides the likelihood of observing a response for every YAFSO subgroup. These probabilities also indicate what percentage of each item response can be observed in the unobserved subgroup/cluster (see Figure 2). The findings are presented below in an order starting from the largest class.

Table 3.

Comparing Frequencies Among the Three-Class LCA Model.^a

	School dropout			Criminal record
	No, n = 160	Yes, n = 300	p-Value^b	No, n = 351	Yes, n = 109	p-Value^b
Classes			<.001			<.001
1	87 (54%)	104 (35%)		177 (50%)	14 (13%)
2	8 (5.0%)	121 (40%)		45 (13%)	84 (77%)
3	65 (41%)	75 (25%)		129 (37%)	11 (10%)
	Smoking			Alcohol
	No, n = 143	Yes, n = 317	p-Value^b	No, n = 329	Yes, n = 131	p-Value^b
Classes			<.001			<.001
1	77 (54%)	114 (36%)		169 (51%)	22 (17%)
2	1 (0.7%)	128 (40%)		35 (11%)	94 (72%)
3	65 (45%)	75 (24%)		125 (38%)	15 (11%)
	Family type			YAFSO age 15 or older
	No, n = 310	Yes, n = 150	p-Value^b	No, n = 169	Yes, n = 291	p-Value^b
Classes			<.001			<.001
1	139 (45%)	52 (35%)		82 (49%)	109 (37%)
2	63 (20%)	66 (44%)		26 (15%)	103 (35%)
3	108 (35%)	32 (21%)		61 (36%)	79 (27%)
	Domestic violence			Familial sharing
	No, n = 341	Yes, n = 119	p-Value^b	No, n = 263	Yes, n = 197
Classes			<.001			<.001
1	162 (48%)	29 (24%)		85 (32%)	106 (54%)
2	68 (20%)	61 (51%)		104 (40%)	25 (13%)
3	111 (33%)	29 (24%)		74 (28%)	66 (34%)
	Education categorical			Risky lifestyle
	No, n = 325	Yes, n = 135	p-Value^b	No, n = 271	Yes, n = 189	p-Value^b
Classes			<.001			<.001
1	147 (45%)	44 (33%)		159 (59%)	32 (17%)
2	70 (22%)	59 (44%)		1 (0.4%)	128 (68%)
3	108 (33%)	32 (24%)		111 (41%)	29 (15%)
	Psychological disorder composite			Young victim
	No, n = 239	Yes, n = 221	p-Value^b	No, n = 293	Yes, n = 167	p-Value^b
Classes			<.001			<.001
1	134 (56%)	57 (26%)		191 (65%)	0 (0%)
2	11 (4.6%)	118 (53%)		93 (32%)	36 (22%)
3	94 (39%)	46 (21%)		9 (3.1%)	131 (78%)
	Peer victim			Old victim
	No, n = 188	Yes, n = 272	p-Value^b	No, n = 439	Yes, n = 21	p-Value^b
Classes			<.001			<.001
1	0 (0%)	191 (70%)		191 (44%)	0 (0%)
2	48 (26%)	81 (30%)		117 (27%)	12 (57%)
3	140 (74%)	0 (0%)		131 (30%)	9 (43%)
	Non-acquaintance victim			Physical violence or threat to kill
	No, n = 374	Yes, n = 86	p-Value^b	No, n = 222	Yes, n = 238	p-Value^b
Classes			<.001			<.001
1	180 (48%)	11 (13%)		115 (52%)	76 (32%)
2	90 (24%)	39 (45%)		43 (19%)	86 (36%)
3	104 (28%)	36 (42%)		64 (29%)	76 (32%)
	Male victim
	No, n = 335	Yes, n = 128	p-Value^b
Classes			<.001
1	157 (47%)	34 (27%)
2	99 (30%)	30 (24%)
3	79 (24%)	61 (49%)

n(%).

Pearson’s chi-squared test.

Figure 2.

Latent class selection.

Class 3 of YAFSO was the largest subgroup (42% of the population) that emerged during the LCA. The victims of the YAFSO in this group were exclusively peers (100%), with a 40% chance of being physically coerced and a slim chance of being a non-acquaintance (6%), the lowest probabilities for both victim characteristics in the three subgroups. Only 18% of the victims of the YAFSO in this group were male, which was, again, the lowest probability among all three subgroups. Another noteworthy finding in this subgroup was that the chance of using physical force or threat to kill to coerce the victims was the lowest among this group (40%) compared to the other subgroups. The YAFSO in this subgroup tended to lack communication with their families (55%), were smokers (61%), and were 15 years or older (58%). On the other hand, the chances for other criminogenic factors were relatively low. Only 8% of the YAFSO in this class had a prior criminal record, only 16% were exposed to domestic violence, and 18% had a risky lifestyle. Accordingly, the most distinctive feature in the largest subgroup was that they exclusively targeted their peers, mostly their acquaintances. The group members had the lowest likelihood of being physically aggressive against their victims and a relatively lower chance of displaying criminogenic factors, such as having a criminal record, a risky lifestyle, a psychological disorder, and exposure to domestic violence. Hence, the largest subgroup can be described as “Non-delinquent, peer victim-targeting” YAFSO.

Classes 1 and 2 were closer in size, while Class 1 was slightly larger and made up 30% of the study population. A 94% percent of the victims in Class 1 were younger than the YAFSO. Closer to half of these victims were male (44%—the largest of all subgroups) and subjected to physical coercion (54%). While there was no peer victim in this group, the chances for the victims to be older than the adjudicated youth were slim (6%). The YAFSO in this group were more likely than not to be school dropouts (53%), smokers (53%), and older than 15 years of age (56%). Lack of familial communication was recorded for 47% of these group members. While this subgroup displayed almost similar characteristics to the previous subset (Class 3) in terms of criminogenic factors (i.e., being a school dropout, having a criminal record, risky lifestyle, and psychological disorders), they were distinctively separable from the members of Class 1 as they targeted younger victims, as opposed to peers, and were more likely to victimize non-acquaintances through physical coercion (compared to Class 3 members). Hence the members of Class 1 can be qualitatively described as “non-delinquent, younger victim-targeting” YAFSO.

Lastly, Class 2 was the smallest of all subgroups consisting of 28% of the study population. Unlike the previous two subgroups, the YAFSO-victim age comparison variables (younger, older, or peer victims) did not yield a distinct probability for the group membership in this class. While the chances of having peer-victims were relatively higher (61%), there was a 29% chance of having younger victims and a 10% chance of having older victims in this subgroup (the highest of all subgroups). The probability of using physical force or threat to kill (67%) was the highest among the YAFSO in this subgroup compared to the other YAFSO classified under the previous two subgroups. Likewise, the probability of having non-acquaintance victims (30%) in this subgroup was also the highest. The YAFSO in this subgroup tended to have the highest likelihood of being school dropouts (94%), having criminal records (65%), being smokers (99%), consuming alcohol (73%), having a risky lifestyle (98%), suffering from psychological disorders (90%), and living within a non-traditional family structure (51%), compared to the other YAFSO classified under the other two subgroups. The members of this group also had the highest chances of being 15 years or older of age (80%). Though Class 2 had the fewest number of YAFSO, they displayed highest level of criminogenic factors and distinct victim/case characteristics. Accordingly, this subgroup was described as “Delinquent, versatile” YAFSO as they had the highest probability of having criminogenic factors and being versatile in terms of offense types that they were adjudicated for as well as the ages of victims they targeted (see Figure 3).

Figure 3.

YAFSO latent class patterns.

Pearson’s chi-squared tests were run to compare frequencies between each factor and the three classes from the LCA model. Table 3 presents the results of the statistical tests. All associations are statistically significant (p < .001). To better understand the latent class patterns, a multiple correspondence analysis (MCA) was also run to explore and visualize the data. Multiple correspondence analysis (MCA) can be seen as a generalization of principal component analysis (PCA) when the variables are categorical instead of continuous (Abdi & Valentin, 2007; Benzécri , 1977). MCA is useful to describe, summarize, and visualize information contained within a data table of n (rows) individuals described by M (columns) categorical variables. Figure 4 presents the MCA biplot where the labels from the LCA groups were superposed to the biplot. The MCA biplot showed that the three groups’ characteristics are consistent with those highlighted in the LCA model. This further provides evidence of the validation of the LCA model.

Figure 4.

MCA biplot for YAFSO data with the three LCA classes (n = 460).

Discussion

Using a data-driven empirical approach called Latent Class Analysis (LCA), the study results indicated that the male YAFSO included in this study did not constitute a homogeneous group but rather consisted of three distinctive latent classes: (a) a non-delinquent, peer-targeting, (b) a non-delinquent younger victim-targeting, and (c) a delinquent versatile YAFSO group. In that sense, the results of this study were in line with the results of earlier studies that indicated that YAFSO show differences among themselves in terms of criminogenic factors as well as the characteristics of their victims (Fox, 2017; Fox & Delisi, 2018; Hunter, 2018; Margari et al., 2015).

The findings related to the latent classes of the YAFSO in the study population resonated with the results of earlier studies regarding the YAFSO typologies in several other ways. First of all, the non-delinquent, peer-targeting subgroup was the largest class in this study, and similar to the largest types determined through earlier LCA, the YAFSO in that group tended to display a relatively low level of criminogenic factors (Barra et al., 2018; Brown, 2019; Fox & Delisi, 2018). This group’s characteristics were also in line with the previous studies’ findings which indicated that YAFSO were more likely to victimize their peers (Finkelhor et al., 2009; Riser et al., 2013). On the contrary, the members of the subgroup with the fewest members, delinquent, versatile YAFSO, had the highest probability of displaying various criminogenic factors as well as signs of aggression in their sexual offenses (i.e., physical coercion, targeting strangers, and victimizing younger and older individuals). This particular finding also resonated with the results of earlier studies, which determined that the classes of YAFSO with the fewest members tended to display higher levels of criminogenic factors and more problematic/versatile offending patterns (i.e., Brown’s [2019] early starter-multiple victims and Cale et al.’s [2016] high-rate chronic groups).

In addition, the latent classes that emerged from the current study confirmed some of the earlier approaches to YAFSO typologies. In that respect, the outcomes of the present study indicated that classifying the YAFSO based on the age differences between them their alleged victims as several earlier studies did (Aebi et al., 2012; Barra et al., 2018; Brown, 2019; Fanniff & Kolko, 2012; Graves, 1993) could be empirically possible. This study’s data-driven approach distinguished YAFSO based exclusively on victimizing peers or younger/older individuals. These two groups also differed on specific offense/victim characteristics–the non-delinquent, younger victim targeting YAFSO in this study were more likely to victimize males and strangers and use physical coercion against their victims than the non-delinquent peer-targeting YAFSO. The results were comparable with the findings of those earlier studies where the differences in various aspects of peer versus younger victim-targeting YAFSO were noted (Aebi et al., 2012; Fanniff & Kolko, 2012; Seto & Lalumière, 2010). On the other hand, the probabilities of displaying certain criminogenic factors were pretty close in these subgroups–the peer-targeting versus younger victim targeting YAFSO yielded almost the same types and levels of criminogenic factors. Hence, the findings of the current study support earlier conclusions that classifying YAFSO based on the offense/victim characteristics per se, without including personal characteristics of those youth, will be limited, at best, in both developing a meaningful YAFSO typology and effective prevention and intervention programs based on those typologies (Fanniff & Kolko, 2012; ‘t Hart-Kerkhoffs et al., 2009).

Likewise, the latent classes that emerged from the current study data were supportive of an approach to classifying YAFSO as “sex-only” or “sex-plus” (i.e., specialized vs. versatile; Butler & Seto, 2002; Cale et al., 2016; Nisbet et al., 2010). The YAFSO that tended to constitute the two classes (Classes 1 and 3) in the current study were primarily the youth adjudicated for only sex offenses with relatively fewer and low-level criminogenic characteristics and included the majority (almost 60%) of all study members. On the other hand, the smaller group (Class 1) included youth adjudicated for various types of offenses (versatile or sex-plus) who demonstrated a significantly higher probability of having various criminogenic factors. In other words, the YAFSO who also committed different types of crimes were more likely to score high on all measures of criminogenic factors and problematic sex offense patterns (i.e., targeting non-acquaintance victims from different age groups, using physical coercion).

Conclusion

Exploring the YAFSO in an understudied non-western social context, this study aimed to determine if the study population constituted separate latent classes with different offense and personal characteristics rather than being homogenous, as pointed out by earlier studies based in the US or other western contexts. In addition to being conducted in an understudied population and providing a comparative perspective on the issue at hand, this study employed an emerging analytical strategy called Latent Class Analysis, which allows one to determine if a sample can be empirically classified based on regularities that emerge from the data, rather than a theoretical position. This study also intended to contribute to the current state of knowledge on various types of YAFSO by employing a combination of offense and personal characteristics as predictor variables rather than relying solely on general criminogenic factors of juveniles.

The current study also certain limitations, and the outcomes should be interpreted accordingly. First of all, the data utilized in this study was based on court files, and those files may not always include ideal types of information or measurements. For instance, the measures of violence, risky behavior, or psychological disorders included in this study were only based on the self-reports or the manifestation of certain behaviors as captured in the official reports; however, a more rigorous assessment of these features could be made with specifically designed measures if the data were based on a clinical sample. As such, this official data set should be interpreted more closely to assess those quantifiable characteristics of the youth that could be recognized and adjudicated by the formal juvenile justice system in Turkey. For this study, another limitation is the lack of recidivism information. Without information on how the members of each group we identified continue or discontinue their sexual or non-sexual delinquent behaviors, the implications drawn from the current study or similar studies without recidivism information are limited. Therefore, future studies in this respect can collect and utilize recidivism information to further understand how latent groups within YAFSO vary in terms of recidivism during and beyond adolescence. In addition, the data was collected in a social context where sex offenses could be more conservatively reported and adjudicated. The largest number of YAFSO in this study being in the group adjudicated for only sex offense with significantly low criminogenic factors could be the result of that cultural perspective on how peer-to-peer sexual relations among youth were perceived even though the sexual behavior itself was not necessarily coercive or deceptive. Likewise, the legal definition of sexual abuse of a child where the age of consent is determined as 15 in the Turkish Criminal Code may also affect how consensual sexual activities between youth with a three or more years of age difference could be criminally adjudicated. In other words, some of the cases reported and adjudicated in the Turkish courts could actually represent consensual sex between youth but were still reported to and adjudicated by the courts as offenses due to the cultural or conservatively designed legal norms that mitigate against adolescent girls’ pre-marital sexual activities (i.e., being considered promiscuous for having sex before marriage). Unfortunately, during this study’s data collection phase, such cases were not specified. Future studies should consider identifying the prevalence of court cases where youth were adjudicated for sex offenses, which could possibly be a consensual relationship, but were adjudicated due to cultural norms or legal standards determined based on these norms. Lastly, since the study was based on official data resources and was not designed to compare findings with previous studies, the classes determined in this study should be compared cautiously with those defined by the earlier studies. Future studies should be designed purposefully to explore different types of YAFSO with comparable sets of variables to offer more concrete policy and research suggestions.

Future research efforts should continue exploring variations among YAFSO by employing both personal and offense characteristics as predictor variables and utilizing data-driven analytical strategies (i.e., LCA, cluster analysis). However, forthcoming studies on this topic must develop a research strategy that includes a set of purposefully chosen criminogenic risk factors and offense/victim characteristics instead of relying on those provided by the official (secondary) data sets to build a comparable classification across different study samples. The most visible limitation of the current state of knowledge regarding the YAFSO typologies is the differences between the existing studies in terms of the types of study participants (court data vs. clinical data) and the included predictor variables determining the group memberships (i.e., including only one type of personal characteristics [i.e., ACEs] and limited variables on offense/victim characteristics). Another gap in the literature related to the YAFSO typologies is the lack of self-reported data. Purposefully designed criminological studies could capture sexually aggressive behaviors that are not officially recorded and classify those youth who demonstrate sexually deviant behavior and would not otherwise be included in research studies relying on official data. Lastly, the literature on YAFSO typologies needs more comparative information. This particular study makes a contribution in this regard by indicating that a non-western sample of YAFSO could also yield qualitatively and quantitatively different latent classes in a similar way to the samples of the YAFSO in those previous studies conducted in the US and other western contexts.

In conclusion, the findings of this study have important implications for policy and research related to YAFSO. Primarily, the majority of YAFSOs have a very low level of delinquent involvement other than the sexual offenses they committed. The present study’s findings point out that the YAFSO with only sexual offenses bears a relatively low probability of bearing criminogenic factors and does not usually display other types of deviant behavior. Therefore, the intervention strategies for this youth should prioritize reintegration and pro-social development rather than disintegration with a punitive approach. On the other hand, more focus should be placed on the sex-plus category of YAFSO, who also manifest higher levels of various criminogenic factors and display multiple types of deviant and violent behavior.

Footnotes

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

ORCID iDs

Hasan Buker

Ayhan Erbay

Notes

Author Biographies

Hasan Buker is an Associate Professor of Criminology and Criminal Justice at the University of West Florida. His research interest covers the issues at the intersection of children, crime, and law, along with policing, administration, and theory. His recent studies were published in Violence & Victims, Crime & Delinquency, Journal of Interpersonal Violence, and Mankind Quarterly.

Achraf Cohen is an Assistant Professor of Statistics at the University of West Florida. His research focuses on statistical models for monitoring (SPM) and anomaly detection. Dr. Cohen has a broad statistics and data science background, with specific expertise in control charts, applied statistical modeling, and simulation studies. Dr. Cohen maintains and seeks collaborations with researchers across diverse disciplines, such as criminology, sports medicine, engineering, and biology, to enable the appropriate statistical modeling. He produced several peer-reviewed publications.

Ayhan Erbay is an Associate Professor of Applied Psychology at the Istanbul Kultur University. His research focuses on juvenile delinquency, the cognitive development of children, criminal responsibility, and developmental psychology. His recent studies were published in Crime & Delinquency, Journal of Interpersonal Violence, and International Journal of Offender Therapy and Comparative Criminology.

References

Abdi

Valentin

(2007). Multiple correspondence analysis. Encyclopedia of Measurement and Statistics, 2(4), 651–657.

Aebi

Vogt

Plattner

Steinhausen

H.-C.

Bessler

(2012). Offender types and criminality dimensions in male juveniles convicted of sexual offenses. Sexual Abuse: A Journal of Research and Treatment, 24, 265–288. https://doi.org/10.1177/1079063211420449

Agresti

(2018). An introduction to categorical data analysis. John Wiley & Sons.

Baglivio

M. T.

Wolff

K. T.

(2021) Adverse childhood experiences distinguish violent juvenile sexual offenders’ victim typologies. International Journal of Environmental Research Public Health, 18, 11345–11359. https://doi.org/10.3390/ijerph182111345

Bandeen-Roche

Miglioretti

D. L.

Zeger

S. L.

Rathouz

P. J.

(1997). Latent variable regression for multiple discrete outcomes. Journal of the American Statistical Association, 92(440), 1375–1386. https://doi.org/10.1080/01621459.1997.10473658

Barra

Bessler

Landolt

M. A.

Aebi

(2018). Patterns of adverse childhood experiences in juveniles who sexually offended. Sexual Abuse: A Journal of Research and Treatment, 30, 803–827. https://doi.org/10.1177/1079063217697135

Beaudry-Cyr

Jennings

W. G.

Zgoba

K. M.

Tewksbury

(2017). Examining the continuity of juvenile sex offending into adulthood and subsequent patterns of sex and general recidivism. International Journal of Offender Therapy and Comparative Criminology, 61(3), 251–268. https://doi.org/10.1177/0306624X15594442

Becker

J. V.

Cunningham-Rathner

Kaplan

M. S.

(1986). Adolescent sexual offenders: Demographics, criminal and sexual histories, and recommendations or reducing future offenses. Journal of Interpersonal Violence, 1, 431–445. https://doi.org/10.1177/088626086001004003

Budd

K. M.

Bierie

D. M.

(2018). Victims of sexual assault perpetrated by female youth: An exploratory study of gender differences. Journal of Sexual Aggression, 24(3), 274–293. https://doi.org/10.1080/13552600.2018.1528793

10.

Buker

Erbay

(2020). Juvenile sex offenders in Turkey: A study on offense characteristics, risk factors, and the differences between juvenile sex offenders and other juvenile offenders. Crime & Delinquency, 67(5), 689–716. https://doi.org/10.1177/0011128720930667

11.

Brown

(2019). Using latent class analysis to explore subtypes of youth who have committed sexual offenses. Youth Violence and Juvenile Justice, 17(4), 413–430. https://doi.org/10.1177/1541204018820578

12.

Butler

S. M.

Seto

M. C.

(2002). Distinguishing two types of adolescent offenders. Journal of the Academy of Child and Adolescent Psychiatry, 41(1), 83–90. https://doi.org/10.1097/00004583-200201000-00015

13.

Cale

Smallbone

Rayment-McHugh

Dowling

(2016). Offense trajectories, the unfolding of sexual and non-sexual criminal activity, and sex offense characteristics of adolescent sex offenders. Sexual Abuse: A Journal of Research and Treatment, 28, 791–812. https://doi.org/10.1177/1079063215580968

14.

Calleja

N. G.

(2015). Juvenile sex and non-sex offenders: A comparison of recidivism and risk. Journal of Addictions & Offender Counseling, 36(1), 2–12. https://doi.org/10.1002/j.2161-1874.2015.00031.x

15.

Christiansen

A. K.

Vincent

J. P.

(2013). Characterization and prediction of sexual and non-sexual recidivism among adjudicated juvenile sex offenders. Behavioral Sciences and the Law, 31(4), 506–529. https://doi.org/10.1002/bsl.2070

16.

Dayton

C. M.

Macready

G. B.

(1988). Concomitant-variable latent-class models. Journal of the American Statistical Association, 83(401), 173–178. https://doi.org/10.2307/2288938

17.

DeGue

Massetti

G. M.

Holt

M. K.

Tharp

A. T.

Valle

L. A.

Matjasko

J. L.

Lippy

(2013). Identifying links between sexual violence and youth violence perpetration: New opportunities for sexual violence prevention. Psychology of Violence, 3(2), 140–156. https://doi.org/10.1037/a0029084

18.

Dunton

C. A.

(2020). Toward an empirically derived classification of sexually aggressive juveniles in a treatment setting. Violence and Victims, 35(5), 751–779.

19.

Dwyer

R. G.

Letourneau

E. J.

(2011). Juveniles who sexually offend: Recommending a treatment program and level of care. Child and Adolescent Psychiatric Clinics of North America, 20(3), 413–429. https://doi.org/10.1016/j.chc.2011.03.007

20.

Fanelli

C. A.

(2018). Comparing Juvenile Sex Offenders and Juvenile Non-sex Offenders in an Adjudicated Group (Unpublished doctoral dissertation). Fordham University.

21.

Fanniff

A. M.

Kolko

D. J.

(2012). Victim age-based subtypes of juveniles adjudicated for sexual offenses. Sexual Abuse: A Journal of Research and Treatment, 24(3), 224–264. https://doi.org/10.1177/1079063211416516

22.

FBI (2021). National Incident-Based Reporting System (NIBRS), 2020. https://www.fbi.gov/news/pressrel/press-releases/fbi-releases-2020-incident-based-data

23.

Finkelhor

Ormrod

Chaffin

(2009). Juveniles who commit sex offenses against minors. PsycEXTRA Dataset. http://unh.edu/ccrc/pdf/CV171.pdf

24.

Fox

B. H.

(2017). What makes a difference? Evaluating the key distinctions and predictors of sexual and non-sexual offending among male and female juvenile offenders. Journal of Criminal Psychology, 7(2), 134–150. https://doi.org/10.1108/JCP-12-2016-0047

25.

Fox

B. H.

Delisi

(2018). From criminological heterogeneity to coherent classes. Youth Violence and Juvenile Justice, 16(3), 299–318. https://doi.org/10.1177/1541204017699257

26.

Gamache

Diguer

Laverdiere

Rousseau

(2012). Development of an object relation–based typology of adolescent sex offenders. Bulletin of the Menninger Clinic, 76(4), 329–364. https://doi.org/10.1521/bumc.2012.76.4.329

27.

Goodman

L. A.

(1974). Exploratory latent structure analysis using both identifiable and unidentifiable models. Biometrika, 61, 215–231. https://doi.org/10.2307/2334349

28.

Graves

R. B.

(1993). Conceptualizing the youthful male sex offender: A meta-analytic examination of offender characteristic by offense type. [PhD thesis, Unpublished, Utah State University].

29.

Hackett

Phillips

Masson

Balfe

(2013). Individual, family and abuse characteristics of 700 British child and adolescent sexual abusers. Child Abuse Review, 22(4), 232–245. https://doi.org/10.1002/car.2246

30.

Hagenaars

J. A.

McCutcheon

A. L.

(2002). Applied latent class analysis. Cambridge University Press.

31.

Hart

S. D.

(2016). Culture and violence risk assessment: The case of Ewert v. Canada. Journal of Threat Assessment and Management, 3(2), 76–96. https://doi.org/10.1037/tam0000068

32.

Hunter

J. A.

(2018). Patterns of sexual offending in juveniles and risk factors. In Ryan

E. P.

Hunter

J. A

Murrie

D. C.

(Eds.), Juvenile sex offenders: A guide to evaluation and treatment for mental health professionals, pp. 34–53. Oxford University Press.

33.

Hunter

J. A.

Figueredo

A. J.

Malamuth

N. M.

Becker

J. V.

(2003). Juvenile sex offenders: Toward the development of a typology. Sexual Abuse, 15(1), 27–48. https://doi.org/10.1177/107906320301500103

34.

Karsten

T. M.

Dempsey

(2018). Neuropsychological risk factors to consider when assessing for sexually abusive youth. Journal of Child Sexual Abuse, 27(8), 1–19. https://doi.org/10.1080/10538712.2018.1542419

35.

Lazarsfeld

P. F.

Henry

N. W.

(1968). Latent structure analysis. Houghton Mifflin.

36.

Linzer

D. A.

Lewis

J. B.

(2011). Polca: An r package for polytomous variable latent class analysis. Journal of Statistical Software, 42(10), 1– 29. https://doi.org/10.18637/jss.v042.i10

37.

Lussier

Berg

V. D. C.

Bijleveld

Hendriks

(2012). A developmental taxonomy of juvenile sex offenders for theory, research, and prevention: The adolescent-limited and the high rate slow desister. Criminal Justice and Behavior, 39(12), 1559–1581. https://doi.org/10.1177/0093854812455739

38.

Margari

Lecce

P. A.

Craig

Lafortezza

Lisi

Pinto

Stallone

Pierri

Pisani

Zagaria

Margari

Grattagliano

(2015). Juvenile sex offenders: Personality profile, coping styles and parental care. Psychiatry Research, 229(1–2), 82–88. https://doi.org/10.1016/j.psychres.2015.07.066

39.

McCuish

E. C.

Lussier

Corrado

R. R.

(2016). Criminal careers of juvenile sex and non-sex offenders: Evidence from a prospective longitudinal study. Youth Violence and Juvenile Justice, 14(3), 199–224. https://doi.org/10.1177/1541204014567541

40.

Nisbet

Smallbone

Wortley

(2010). Developmental, individual, and family characteristics of specialist, versatile, and short-duration adolescent sex offenders. Sexual Abuse in Australia and New Zealand, 2, 85–96.

41.

O’Brien

M. J.

(1991). Taking sibling incest seriously. In Patton

M. Q.

(Ed.), Family sexual abuse: Frontline research and evaluation (pp. 75–92). Sage Publications, Inc. https://doi.org/10.4135/9781483325613.n5

42.

O’Brien

Bera

(1986). Adolescent sexual offenders: A descriptive typology. Newsletter of the National Family Life Education Network, 1, 1–5.

43.

Oxnam

Vess

(2006). A personality-based typology of adolescent sexual offenders using the Millon adolescent clinical inventory. New Zealand Journal of Psychology, 35(1).

44.

Oxnam

Vess

(2008). A typology of adolescent sexual offenders: Millon adolescent clinical inventory profiles, developmental factors, and offense characteristics. The Journal of Forensic Psychiatry Psychology, 19(2), 228–242. https://doi.org/10.1080/14789940701694452

45.

Przybylski

Lobanov-Rostovsky

(2017). Unique considerations regarding juveniles who commit sexual offenses. In Lobanov-Rostovsky

Przybylski

(Eds.), U.S. Department of Justice Office of Justice Programs, Sex Offender Management Assessment and Planning Initiative (pp. 221–226). USDJ.

46.

Pullman

L. E.

Leroux

E. J.

Motayne

Seto

M. C.

(2014). Examining the developmental trajectories of adolescent sexual offenders. Child Abuse & Neglect, 38(7), 1249–1258. https://doi.org/10.1016/j.chiabu.2014.03.003

47.

Rasmussen

L. A.

Miccio-Fonseca

L. C.

(2007). Paradigm shift: Implementing mega, a new tool proposed to define and assess sexually abusive dynamics in youth ages 19 and under. Journal of Child Sexual Abuse, 16(1), 85–106. https://doi.org/10.1300/J070v16n01_05

48.

Richardson

Kelly

T. P.

Graham

Bhate

S. R.

(2004). A personality-based taxonomy of sexually abusive adolescents derived from the millon adolescent clinical inventory (MACI). British Journal of Clinical Psychology, 43, 283–298. https://doi.org/10.1348/0144665031752998

49.

Riser

D. K.

Pegram

S. E.

Farley

J. P.

(2013). Adolescent and young adult male sex offenders: Understanding the role of recidivism. Journal of Child Sexual Abuse, 22(1), 9–31. https://doi.org/10.1080/10538712.2013.735355

50.

Robertiello

Terry

K. J.

(2007). Can we profile sex offenders? A review of sex offender typologies. Aggression and Violent Behavior, 12, 508–518. https://doi.org/10.1016/j.avb.2007.02.010

51.

Ryan

(2018). Sexual offending in society and the legal system. In Ryan

E.P.

Hunter

J. A.

Murrie

D. C.

(Eds.), Juvenile Sex Offenders: A Guide to Evaluation and Treatment for MentalHealth Professionals, pp. 1–20. Oxford University Press.

52.

Ryan

E. P.

Otonichar

J. M.

(2016). Juvenile sex offenders. Current psychiatry reports, 18(7), 67. https://doi.org/10.1007/s11920-016-0706-1

53.

Şengül

H. A.

Cantürk

Dağalp

(2012). Evaluation of the children on trial in the Ankara 1st children’s High Criminal Court for peer abuse. Turkish Archives of Pediatrics, 47(4), 283–289. https://doi.org/10.4274/tpa.909

54.

Seto

M. C.

Lalumière e

M. L

. (2010). What is so special about male adolescent sexual offending? A review and test of explanations through meta-analysis. Psychological Bulletin, 136(4), 526–575. https://doi.org/10.1037/a0019700

55.

Smith

W. R.

Monastersky

Deisher

R. M.

(1987). MMPI-based personality types among juvenile sexual offenders. Journal of Clinical Psychology, 43, 422–430. https://doi.org/10.1002/1097-4679(198707)43:4<422::AID-JCLP2270430414>3.0.CO;2-3

56.

Stevens

Hutchin

French

Craissati

(2013). Developmental and offense-related characteristics of different types of adolescent sex offenders: A community sample. Journal of Sexual Aggression, 19(2), 138–157. https://doi.org/10.1080/13552600.2011.645889

57.

‘t Hart-Kerkhoffs

Doreleijers

T. A. H.

Jansen

L. M. C.

van Wijk

A. P. H.

Bullens

R. A. R

. (2009). Offense related characteristics and psychosexual development of juvenile sex offenders. Child and Adolescent Psychiatry and Mental Health, 3, 19. https://doi.org/10.1186/1753-2000-3-19

58.

Turkish Judicial Statistics (2021). Ministry of Justice. General Directorate of Judicial Record and Statistics. https://adlisicil.adalet.gov.tr/Resimler/SayfaDokuman/501202216013122420211449082020H%C4%B0ZMETE%C3%96ZELK%C4%B0TAP.pdf

59.

Veneziano

(2002). Adolescent sex offenders. Trauma, Violence, Abuse, 3(4), 247–260. https://doi.org/10.1177/1524838002237329

60.

Warner

Bartels

(2015). Juvenile sex offending: Its prevalence and the criminal justice response. University of New South Wales Law Journal, 38(1), 48–75.

61.

Weller

B. E.

Bowen

N. K.

Faubert

S. J.

(2020). Latent class analysis: A guide to best practice. Journal of Black Psychology, 46(4), 287–311. https://doi.org/10.1177/0095798420930932

62.

Willis

G. M.

Letourneau

E. J.

(2018). Promoting accurate and respectful language to describe individuals and groups. Sexual Abuse, 30(5), 480–483. https://doi.org/10.1177/1079063218783799

63.

Wood

Welman

Netto

(2000). A profile of young sex offenders in South Africa. Southern African Journal of Child and Adolescent Mental Health, 12, 45–58. https://doi.org/10.1080/16826108.2000.9632367

64.

Worling

J. R.

(2001). Personality-based typology of adolescent male sexual offenders: Differences in recidivism rates, victim-selection characteristics, and personal victimization histories. Sexual Abuse: A Journal of Research and Treatment, 13(3), 149–166. https://doi.org/10.1023/A:1009518532101

65.

Yıldız

(2012). Crime of sexual abuse of children and the problem of secondary victimization, [Unpublished doctoral dissertation] Istanbul University, Institute of Forensic Medicine, Department of Social Sciences.

66.

Yoder

J. R.

Leibowitz

G. S.

Peterson

(2018). Parental and peer attachment characteristics: Differentiating between youth sexual and non-sexual offenders and associations with sexual offense profiles. Journal of Interpersonal Violence, 33(17), 2643–2663. https://doi.org/10.1177/0886260516628805