Applying Offending Trajectory Analyses to Men Adjudicated for Child Sexual Exploitation Material Offenses

Abstract

We applied offending trajectory analyses to 387 men adjudicated for child sexual exploitation material (CSEM) offenses. After an average of 20 years, we found two trajectories of sexual offending and violent offending: one that peaked in late adolescence and was associated with higher rates of crimes, and one that peaked in the 30s and was associated with a lower rate of crime. We found four trajectories when modeling any crime. The findings highlight the heterogeneity of men with CSEM offenses. Although lifelong patterns of numerous sexual crimes were uncommon, men with more sexual crimes had greater indicators of sexual interest in children and a younger age of first contact with police. CSEM offenses were rarely the first offense in their criminal trajectories. As such, early intervention targeting youth before they are further advanced in their criminal careers may also reduce future CSEM offending.

Keywords

child pornography child sexual exploitation material (CSEM) offending indecent images offending trajectories

Child sexual exploitation material offenses (CSEM; legally referred to as child pornography in Canada and United States) are now a substantial proportion of sexual crimes (Adams & Flynn, 2017; Department of Justice Canada, 2019; McManus & Almond, 2014). Systematic reviews of individuals with CSEM offenses show that these individuals are generally less antisocial, have less criminal history, but have more issues in the sexual domain (e.g., pedophilia) than men with contact sexual offenses (Babchishin et al., 2011, 2015). Individuals with both contact sexual offenses and CSEM offenses (mixed CSEM offending) are riskier and reoffend more than individuals with exclusively CSEM offenses, and no other types of sexual offenses (CSEM-exclusive offending; Babchishin et al., 2015; Elliott et al., 2019; Henshaw et al., 2018).

There has been an influx of research on the characteristics of individuals with CSEM offenses (Babchishin et al., 2018; Seto, 2013), and some research examining future offending and risk factors for recidivism among this group (e.g., Elliott et al., 2019; Seto & Eke, 2015; Wakeling et al., 2011). There has yet to be a study that applies offending trajectory analyses to individuals with CSEM offenses, in line with developmental life course perspective on offending (see Lussier & McCuish, 2020). This is an important gap that could highlight different criminal career pathways, including age of onset, generalization versus specialization in offending, and escalation or desistance in offending.

Age and crime typically have a curvilinear relationship (Quetelet, 1831/1984) that emerges in early adolescence, peaks in early adulthood, and reduces with time (Loeber & Farrington, 2014), likely due to changes in the circumstances and resources of the individuals (e.g., psychological maturity, marriage, employment; Brandt, 2006; see Loeber & Farrington, 2012, for review). Although the age–crime curve is a robust finding in criminology, the peak has been found to vary based on offense type, cohort, and gender (Farrington, 1986; Kim & Bushway, 2018; Piquero et al., 2012). Some research has looked at the criminal trajectories of men who have sexually offended (e.g., Lussier et al., 2010), but there is a lack of this kind of research with men who have committed CSEM offenses.

Criminal trajectory research of individuals who commit sexual offenses categorizes most as generalists (both sexual and nonsexual offenses) and few as specialists (i.e., their criminal behaviors are restricted to sexual offenses; for example, Cale et al., 2016; Harris et al., 2009; Lussier & Blokland, 2014; Soothill et al., 2000). There is also little consistency across characteristics of sexual crime within an individual (Saramago et al., 2020). Rates of crossover in victim age for individuals with multiple sexual offenses range from 7% to 70% (median [Mdn] = 36%, samples [k] = 10), rates of crossover in relationship with victim (e.g., intrafamilial vs. extrafamilial) range from 7% to 80% (Mdn = 25%, k = 12), and rate of victim gender crossover range from 9% to 36% (Mdn = 17%, k = 8; see Saramago et al., 2020; see Saramago et al., 2020, for review¹). The lack of consistency across type of sexual offense and characteristics of sexual offenses suggests that men with CSEM offenses may be similarly inconsistent.

Knowing more about the trajectory of offending within individuals who commit CSEM offenses could inform management and intervention strategies. Trajectory studies suggest at least two sexual offending trajectories, with crime rates peaking in adolescence and then middle-adulthood, likely due to victim opportunity changes (from peers in adolescence to children and their friends as middle-aged individuals; for example, Smallbone et al., 2008). A proportion of individuals with CSEM offenses will exclusively commit CSEM offenses (these offenses are mainly committed online with the material being accessible from a variety of devices) and, thus, a differential opportunity for access to in-person victims may be less relevant to their offending trajectories. In addition, individuals with CSEM offenses are different from those with typical contact sexual offenses (Babchishin et al., 2015) and, thus, it would not be surprising if different offending trajectories were identified.

Current Study

The aim of this study was to examine the extent of offending across time in a sample of men adjudicated for CSEM offenses by applying criminal trajectory analyses. The data for reliable trajectory analyses require a long follow-up period and a representative sample. We used a representative cohort of all men charged or convicted of CSEM offenses in a large Canadian province from 1996 to 2011, who were between the ages of 12 (i.e., the age when criminal records can be recorded in Canada) and 74 at their first contact with police (see the age at first police involvement histogram in Figure S3 in Supplemental Part C). The study had criminal history information for the sample, on average, across 20 (range = 0.03–61) years, with the CSEM offense having occurred by 2011 at the latest, allowing for post-CSEM follow-up through to 2018 (see criminal history coverage histogram in Figure S4 in Supplemental Part C).

We hypothesized there were different trajectories among men with CSEM offenses given that some individuals with CSEM offenses also commit contact sexual offenses (Seto et al., 2011). Specifically, we expected at least two groups: (a) a low rate group that comprised men with CSEM offenses as their only sexual offense and (b) a higher frequency group comprised of men who commit CSEM and additional offenses. We hypothesized that both these groups would follow the age–crime curve, with earlier onset related to more crime and later onset related to less crime. We also examined violent offending, including contact sexual offending and any offending, to assess the extent to which criminal trajectories differed across type of crime for this population. A small proportion of adolescents with sexual offenses (about 12%) increase the frequency and seriousness of their crimes with age and these individuals have a younger age at first involvement with the criminal justice system and a history of both sexual and nonsexual crimes (Cale et al., 2016). As such, we hypothesized that early age of onset and other offenses would predict individuals with more frequent and serious sexual offending patterns. We also hypothesized that indicators of pedophilia and substance use would predict individuals with more frequent and serious sexual offending patterns, given these are robust risk factors for the onset and maintenance of sexual offending (Hanson & Morton-Bourgon, 2005; Whitaker et al., 2008) and there is variability in antisocial tendencies and atypical sexual interests among men with CSEM offending that are associated with variability in nonsexual and contact sexual offending (Babchishin et al., 2015).

Method

Sample

The sample consisted of 387 men who had been charged or convicted of at least one CSEM offense (i.e., possessing, distributing, accessing, or making CSEM) prior to 2011. In Canada, the legal definition for “making” CSEM is broader than the legal definition of “production” in the United States. “Making” can include actions beyond documenting a sexual offense against a child. While all “production” charges in the United States would result in a “making” charge in Canada, the reverse is not the case. Of those individuals charged with making CSEM (n = 90/387), 52.2% of cases documented their contact sexual abuse or sexual exploitation of a child (47/90). The remaining cases (35.6%; 32/90) had making charges relating to morphing images, reproducing images (e.g., saving images to CDs and printing text stories), writing text stories about sex with children, engaging in online chat with other adults about sex with children, or creating websites for sharing CSEM; 12.2% were unknown (n = 11).

The criminal activity for this sample across ages was tracked for an average of 20 (age at end of follow-up – age of first offense; SD = 9.4; range = 0.03–61.1) years. During this period, the sample of 387 men accumulated a total of 1,747 offenses, with an average of 4.5 offenses per person (SD = 5.2). The mean age at first police involvement was 32 (SD = 13; range = 12–74) years, and the sample was, on average, 52 years old (SD = 13; range = 30–88) at the end of the follow-up period in 2018. Half the sample (n = 202 [52%]) had no prior charges or convictions for any offenses before their first CSEM offense. On average, the individuals were 39 years old at release from their first CSEM offense (i.e., index offense in the current study) between 1979 and 2012 (SD = 13; range = 18–76; Mdn = 37). The average length of follow-up (street time) since their release from their first CSEM offense was 12 years (SD = 4; range = 0.1–39).

Procedure

This study presents a reanalysis of two samples of men adjudicated for CSEM offenses from Seto and Eke (2015) and Eke et al. (2019) to increase the sample size and the range of years of the investigations. The first sample consisted of 301 men convicted of at least one CSEM offense following police CSEM investigations from 1993 through 2006 across a large Canadian province (Seto & Eke, 2015). To be consistent with how we defined other offending types (i.e., by charges or convictions), we included 15 additional cases not included in that original sample. These were men who were charged, but not convicted of CSEM at the time of the earlier study. With the longer follow-up, seven of the 15 cases ended with conviction for CSEM, and the remaining eight had alternative outcomes: death prior to additional court appearances (n = 3), charges were withdrawn in favor of a peace bond (n = 2), CSEM charges withdrawn in favor of a plea to contact offenses (n = 1), or withdrawn because the content was later deemed to be nudity rather than CSEM (n = 1). Only one individual was found not guilty, as the court could not prove the CSEM was his (shared computer) despite the suspect admitting CSEM use to police. The second sample was another 86 cases involving individuals charged or convicted of a CSEM offense from the same province (Eke et al., 2019).

Cases initially came to the attention of police in a variety of ways, including third-party reporting, victim complaints (for those who had also committed contact sexual offenses or sexual solicitation offenses), and online activity. Most (85%) of the cases involved the use of online technologies to access CSEM. There was no preselection of cases; police provided all of their available closed CSEM case files, and these cases were included if there was sufficient information. In the original work, the case that was investigated by the participating police service was considered to be the index CSEM offense; some men had a prior CSEM offense that was counted as part of their criminal history. In the current criminal career study, the “index” CSEM offense was the first known to police, with the earliest dating back to 1979.

Offense history and offense details, demographic information, and psychological variables were coded from police file information in the original studies (see Eke et al., 2019; Seto & Eke, 2015). We extended the follow-up period through 2018 using official criminal history from similar sources and we conducted new coding for the purpose of the current study. Specifically, dates relevant to all offenses were coded: (a) offense date, (b) suspect date (first indication person is a suspect in a case) or investigation date (when police began their formal investigation), (c) arrest date, (d) charge date, and (e) conviction date.

We coded for offense types by reviewing three sources: police occurrence reports, the Canadian Police Information Centre (CPIC), and the investigative file information from the collaborating police services. We create three offending categories: (a) any sexual offenses, which include contact and noncontact (including CSEM offending); (b) any violent (including sexual) offending, which included all crimes that involved a direct confrontation with the victim (including contact sexual offenses, but excluding noncontact sexual offenses, CSEM, and sexually motivated breaches); and (c) any offense included all crimes and all technical offenses (e.g., breach of conditional release), regardless of whether they were sexually motivated. In cases where CPIC query results and police or investigative files disagreed (e.g., due to a plea deal), police occurrence and investigative files were prioritized.

Interrater Reliability

This new set of coding was conducted by two independent raters (graduate-level students) using a standardized coding manual and form that is available upon request. Interrater reliability analyses for the new set of coding were conducted on six separate occasions using 60 cases (10 cases per time point). Interrater reliability was high. Continuous variables were assessed using absolute intraclass correlations (ICCs) based on a two-way mixed design, and ranged from .96 to 1.00 (Mdn = 1.00, n variable = 22). Categorical variables, including the outcome variables, were assessed using Cohen’s kappa statistic and percent agreement and ranged from 75% to 100% agreement (Mdn = 98%, n variable = 32; κ = .50–1.00, Mdn = .93, n variable = 30). None of the variables were excluded due to unacceptable interrater reliability (Cicchetti, 1994; Landis & Koch, 1977). The original scoring of demographic and psychological variables were used, which also had high interrater reliability (r = .94–1.00, κ = .70–1.00; described in Seto & Eke, 2015).

Measures

A variety of measures, available in police case files, were included as predictors. Demographic characteristics were (a) age at different offense dates (e.g., first contact with police, first CSEM offense), (b) employed (not working vs. student, unskilled, semi-skilled, professional), and (c) never married (single vs. married/common-law or separated/divorced/widowed). Indication of sexual interest in children was coded based on (a) the number of child (aged below 12) victims in a contact sexual offense (scored for those with contact sexual offenses); (b) the number of male victims in a CSEM sexual offense involving contact (i.e., “making”) against a child; (c) pedophilic interest that was scored based on admission to police regarding a pedophilic or hebephilic interest/admission of sexual interest in CSEM and/or a clinical diagnosis of sexual interest in children; (d) more boy than girl CSEM material; (e) more boy than girl other child material (nudity/clothed images); and (f) volunteered in a role with high access to children. Items c through f are items of the Child Pornography Offender Risk Offender Tool (CPORT) and the Correlates of Admission of Sexual Interest in Children (CASIC) scale.² These items are associated with pedophilic sexual interests (e.g., Seto & Eke, 2017).

We included all available psychological variables: (a) substance use problems scored on a 4-point scale (1 = does not use, 2 = uses occasionally, 3 = some problems with use such as problems at work or marital issues related to use, and 4 = severe problems with use such as impaired driving charges, marital breakup relating to use, job loss, or diagnosis of dependence), and (b) major mental illness or personality disorder (scored as not evident, some indication based on evidence, or known diagnosis). These variables were included to examine the criterion validity of the trajectories, given they were not used in the trajectory assignment.

For the current study, information of predictors was coded based on an individual’s first CSEM offense, as this information was obtained from the research that was the source of the samples (Eke et al., 2019; Seto & Eke, 2015). As a result, there is missing information for some cases due to (a) the current CSEM index offense being different from that in the originating study due to how the index was determined in the original research (17 cases), and (b) the current sample including cases that were excluded in prior research (15 cases). In the previous research, a coin flip was used to determine the index offense in cases where an individual was investigated multiple times by the collaborating police service, and an investigative case file was available to reduce potential selection bias relating to offense history and recidivism (Seto & Eke, 2015). In contrast, in the current trajectory work, the first CSEM offense was the index offense for our analysis.

Analytic Strategy

Group-Based Trajectory Modeling

We performed semi-parametric group-based trajectory modeling (SPGM; Nagin & Land, 1993) using the SAS macro pro traj (Jones et al., 2001; Jones & Nagin, 2007). Semi-parametric group-based modeling identifies different cut points of individuals to explain heterogeneity in criminal trajectories and the shape of the trajectory (e.g., linear, quadratic, cubic, and quartic curves). Pro traj is a SAS procedure to identify clusters of individuals by fitting a group-based model to longitudinal data. Given that criminal offenses, particularly sexual offenses, are relatively rare and that the reporting of offenses to the police is low (Moreau, 2019), there were considerable periods with zero offenses across age and follow-up time. As such, the zero-inflated Poisson (ZIP) model was used to estimate trajectories of count data. The ZIP model allows the parameters to have “temporary” spells of the zero-offending period without considering the change in their overall rate of offending (Bushway et al., 2003). As a result, the ZIP model displays relatively smoothed trajectories despite the significant fluctuation of counts. The extraction of groups becomes more stable with samples greater than 200; we had 387 cases (D’Unger et al., 1998; Sampson et al., 2004). The shape of the curve indicates whether the function changes across time. If linear, the function is constant, and if nonlinear (i.e., quadratic, cubic, quartic), the function varies across time.

Decision Rules

There were three key decisions when identifying trajectory groups: (a) the number of groups that best fit the data; (b) polynomial order that best fit the trajectory (e.g., linear, quadratic, cubic); and (c) range of data availability. Different models with a varying number of groups and shapes were compared to find the model that best fits the data. To determine the optimal number of offending trajectories, the Bayesian information criterion (BIC) index was used. The BIC index measures the improvement in model fit obtained by adding more parameters (e.g., more trajectory groups), but also rewards parsimony by imposing penalties for more parameters (Nagin, 2005). For deciding on the number of groups, the model with the larger BIC value was chosen. In addition, the Bayes factor approximation was examined to determine whether the difference in BIC values between the two models is substantive (exp^{[BICi – BICj]}; Nagin, 2005). Weak, moderate, and strong evidence the reference model is, a better fit is represented by Bayes factor between 1 and 3, 3 and 10, and values greater than 10, respectively (Nagin, 2005). Once the best-fitted model was chosen (i.e., Bayes factor < 3 compared with the subsequent model), further analyses were stopped. To avoid problems associated with having only a small number of individuals defining offending trajectories at the youngest or oldest ages and latest follow-up period, we limited the offending trajectories to ages (18–60) and to a follow-up period (1–15 years) for which data were available on a substantial number of individuals.

Missing Data

If people were deceased, deported, or in custody for a whole year during the observation period, the unobserved years were coded as missing and did not contribute to estimating the trajectories of criminal behavior. Of the 387 individuals, 6.7% (n = 26; M_age = 55, SD = 11, range = 28–74) were deceased and 3.4% (n = 13; M_age = 40, SD = 10, range = 25–63) were deported. If someone was incarcerated multiple times (e.g., year 1, 4, and 6), these years were coded as missing.

Procedure for Computing Age–Crime Curve

We applied one-, two-, three-, four-, and five-group models of offense trajectories until models with the best fit were identified for each offense type. We ended our test at five groups because fit decreased at this point. Once the number of groups was selected, we examined which shape of the curve fit the data best. The BIC and Bayes factor for model selections and averaged assignment probabilities for the selected groups are presented in Supplemental Part A. The age–crime curve for additional offense types (CSEM and contact sexual offenses) is also presented in Supplemental Part B. The number of groups was replicated when we defined the x-axis as a follow-up period rather than age (see Supplemental Part C for these analyses).

Group Comparisons

We examined the extent to which demographic and psychological variables could predict trajectory group membership across sexual, violent, and all types of offending. Area under the curve (AUC) analysis (Swets et al., 2000) was used to compare trajectories. AUC, which can vary between 0 and 1, can be interpreted as the probability that a randomly selected member of Group 1 will have a higher score than a randomly selected member of Group 2: AUCs above .50 indicate Group 1 is higher, whereas AUCs below .50 indicate Group 2 is higher. An AUC value is statistically significant if the 95% confidence interval (CI) does not contain .50. AUCs have the benefits of being robust against low base rates (Babchishin & Helmus, 2016; Hsu, 2002). As a heuristic, an AUC of .56 corresponds to a small effect size, .64 reflects a moderate effect, and .71 reflects a large effect size (Rice & Harris, 2005). For three or more trajectory groups, chi-square (categorical variables) or analysis of variance (ANOVA) (continuous variables) were used to compare groups. The omega-squared (ω²) values of .01, .06 and .14 for ANOVA represent small, medium, and large effects, respectively (Kirk, 1996). The Cramer’s V is a measure of the strength of association between two categorical variables (0 ≤ V ≤ 1); values of .06, .17 and .29 for chi-square represent small, medium, and large effects, respectively (Cohen, 1988).

Results

Any Sexual Offending (Contact and Noncontact Including CSEM Offending)

For frequency of any sexual offending, a two-group model with a quadratic (∩ shaped) curve provided the best model fit: There was a late onset group with earlier periods of desistance (Group 1; 98%, n = 379) and an early onset group with longer periods before desistance (Group 2; 2%; n = 8). The early onset group had an earlier age of police involvement (CSEM offending: 23 years old, any offending: 18 years old) than the late onset group (CSEM offending: 36 years old, any offending: 32 years old). The early onset group (Group 2) displayed a higher frequency of sexual offenses than Group 1 until approximately their mid-30s, at which point both groups had similar frequencies of sexual offenses. The late onset group (Group 1) had much lower frequencies of sexual offenses. Most men with CSEM offenses were classified as sexual offending specialists (98%), meaning they did not commit any other kinds of sexual offenses (see Figure 1).

Figure 1:

Age–Crime Curves of Men With CSEM Offenses: (A) Any Sexual Offenses: n = 640, (B) Violent (Including Contact Sexual) Offenses: n = 322, and (C) Any Offenses: n = 1,498

The early onset group was younger at first police involvement (AUC = .88, 95% CI = [.82, .93]) and CSEM offense (AUC = .81, 95% CI = [.68, .93]), and committed more sexual and violent crimes over time (AUC = .75–.96, Mdn = .83). CSEM offending was less likely to be the first offense of the early onset group (AUC = .70, 95% CI = [.55, .86]); men in the late onset group offended less frequently overall and were more likely to have their first identified sexual offense be a CSEM offense (see Table 1). A greater number of sexual contact child victims (AUC = .82, 95% CI = [.63, .996]), indicators of sexual interest in children (pedophilia/hebephilia; AUC = .75, 95% CI = [.59, .91]), and a preference for boys over girls in CSEM material (AUC = .78, 95% CI = [.59, .98]) also distinguished groups. The presence of a personality disorder also strongly distinguished the early onset group (AUC = .96, 95% CI = [.91, .999]).

Table 1:

Predicting Higher Frequency Group for Sexual Offending

Variable	Group 1: Late onset (n = 379)	Group 2: Early onset (n = 8)	AUC	95% CI
Variable	M (SD) or n/N (%)		AUC	95% CI
Demographic
Age at first police involvement^a	32.3 (13.5)	17.9 (2.5)	.875	[.815, .934]
Age at first CSEM offense^a	36.0 (12.7)	23.0 (7.8)	.806	[.681, .932]
Never married at the first CSEM offense	188/355 (53.0)	7/8 (87.5)	.673	[.514, .831]
Unemployed	27/379 (7.1)	1/8 (12.5)	.527	[.317, .737]
Criminal history
Total number of offenses	4.3 (4.9)	13.5 (10.8)	.856	[.757, .954]
Total number of CSEM offenses	1.3 (0.6)	2.6 (1.7)	.748	[.533, .963]
Making CSEM offenses	0.9 (0.6)	1.0 (1.3)	.659	[.456, .862]
Total number of contact sexual offenses	0.5 (0.9)	3.1 (3.2)	.805	[.602, .999]
Total number of any sexual offenses	1.8 (1.2)	5.9 (2.5)	.959	[.923, .995]
Total number of any violent offenses	0.9 (1.4)	4.1 (4.0)	.829	[.663, .994]
Was CSEM offense the first offense?^a	201/379 (53.0)	1/8 (12.5)	.703	[.550, .855]
Indicators of sexual interest in children
Volunteered with children (CASIC Item 5)	31/358 (8.7)	0/8 (0.0)	.457	[.271, .643]
Indication of pedophilic/hebephilic interest (CPORT Item 5)	108/308 (35.1)	6/7 (85.7)	.753	[.594, .913]
More boy than girl CSEM (CPORT Item 6)	50/340 (14.7)	5/7 (71.4)	.784	[.587, .981]
More boy than girl child other material (CPORT Item 7)	53/338 (15.7)	4/7 (57.1)	.707	[.486, .928]
Number of boy victims in a making CSEM offense	0.3 (0.5)	1.0 (0.0)	.837	[.690, .984]
Number of minor victims in contact sexual offense	1.3 (0.8)	3.3 (2.9)	.815	[.633, .996]
Psychological variables
Major mental illness known at the first CSEM offense	47/118 (39.8)	2/3 (66.7)	.634	[.320, .948]
Personality disorder known at the first CSEM offense	9/110 (8.2)	2/2 (100.0)	.959	[.912, .999]
Substance (alcohol and/or drug) use problem, generally	48/260 (18.5)	2/6 (33.3)	.574	[.327, .821]
Drug use problem, scored as severe	7/160 (4.4)	1/5 (20.0)	.578	[.296, .860]
Alcohol problems, scored as severe	13/179 (7.3)	0/5 (0.0)	.464	[.224, .703]
Charge or conviction for DUI/drive while impaired	31/363 (8.5)	0/7 (0.0)	.457	[.259, .656]

Note. Larger AUC indicates that higher scores (or higher risk) on the predictor (e.g., number of offenses) were related to Group 2 (early onset group). Boldfaced values were statistically significant at p < .05. AUC = area under the curve; CI = confidence interval; CSEM = child sexual exploitation material; CASIC = Correlates of Admission of Sexual Interest in Children; CPORT = Child Pornography Offender Risk Tool; DUI = driving under the influence.

These variables were reversed scored so that larger AUC indicated higher risk (i.e., younger age) for Group 2 (early onset trajectory).

Any Violent (Including Contact Sexual) Offending

For violent (including contact sexual) offenses, 59% of men (n = 228) were classified into a lower frequency group (Group 1) and 41% (n = 159) into a higher frequency group (Group 2) that had an earlier average age at first police involvement (28 vs. 34 years old). The higher frequency group started showing a lower frequency of violent offenses around their late 30s, but continued to violently offend at a higher frequency than the lower frequency group. As expected, there were many more men with CSEM offending classified as generalists when defining offending as violent (including contact sexual) offending than when we defined offending as any sexual offending (59% vs. 2%).

The higher frequency group had younger age at first police involvement (AUC = .62, 95% CI = [.57, .68]) and a greater number of all types of offending, including violent offending (AUCs = .58–.98, Mdn = .78). In addition, their first offense was typically not CSEM (AUC = .76, 95% CI = [.71, .81]; see Table 2). These two groups did not differ on indicators of sexual interest in children. Of the psychological variables examined, only substance use distinguished the two trajectories, with a small effect size (AUC = .59, 95% CI = [.52, .66]).

Table 2:

Predicting Higher Frequency Group for Violent Offending

Variable	Group 1: Lower frequency (n = 228)	Group 2: Higher frequency (n = 159)	AUC	95% C.I.
Variable	M (SD) or n/N (%)		AUC	95% C.I.
Demographic
Age at first police involvement^a	34.5 (14.7)	28.3 (10.7)	.621	[.565, .676]
Age at first CSEM offense^a	34.9 (13.6)	36.9 (11.5)	.437	[.380, .495]
Never married at the first CSEM offense	123/211 (58.3)	72/152 (47.4)	.445	[.385, .505]
Unemployed	15/228 (6.6)	13/159 (8.2)	.508	[.449, .567]
Criminal history
Total number of offenses	2.4 (2.6)	7.6 (6.7)	.856	[.819, .894]
Total number of CSEM offenses	1.2 (0.6)	1.5 (0.8)	.577	[.518, .636]
Making CSEM offenses	0.3 (0.5)	0.6 (0.7)	.630	[.572, .687]
Total number of contact sexual offenses	0.1 (0.3)	1.3 (1.3)	.862	[.820, .905]
Total number of any sexual offenses	1.3 (0.7)	2.6 (1.7)	.777	[.728, .826]
Total number of any violent offenses	0.1 (0.4)	2.2 (1.7)	.982	[.969, .994]
Was CSEM offense the first offense?^a	167/228 (73.2)	35/159 (22.0)	.756	[.706, .806]
Indicators of sexual interest in children
Volunteered with children (CASIC Item 5)	17/212 (8.0)	14/154 (9.1)	.505	[.445, .565]
Indication of pedophilic/hebephilic interest (CPORT Item 5)^a	63/183 (34.4)	51/132 (38.6)	.521	[.456, .586]
More boy than girl CSEM (CPORT Item 6)	25/199 (12.6)	30/148 (20.3)	.539	[.477, .600]
More boy than girl child other material (CPORT Item 7)	26/199 (13.1)	31/146 (21.2)	.541	[.479, .603]
Number of boy victims in a making CSEM offense	0.4 (0.8)	0.3 (0.5)	.503	[.255, .751]
Number of minor victims in contact sexual offense	1.2 (0.8)	1.4 (1.1)	.570	[.392, .748]
Psychological variables
Major mental illness known at the first CSEM offense	28/70 (40.0)	21/51 (41.2)	.506	[.401, .611]
Personality disorder known at the first CSEM offense	4/68 (5.9)	7/44 (15.9)	.550	[.439, .661]
Substance use (alcohol and/or drug) use problem^a	17/155 (11.0)	33/111 (29.7)	.594	[.523, .664]
Drug use problem, scored as severe	3/88 (3.4)	5/7 (6.5)	.515	[.427, .604]
Alcohol use problem, scored as severe	2/96 (2.1)	11/88 (12.5)	.552	[.468, .636]
Charge or conviction for DUI/drive while impaired	11/216 (5.1)	20/154 (13.0)	.539	[.479, .600]

Note. Boldfaced values were statistically significant at p < .05. Larger AUC indicates that higher scores (or higher risk) on the predictor (e.g., number of offense) were related to Group 2 (high frequency group). AUC = area under the curve; CI = confidence interval; CSEM = child sexual exploitation material; CASIC = Correlates of Admission of Sexual Interest in Children; CPORT = Child Pornography Offender Risk Tool; DUI = driving under the influence.

These variables were reversed scored so that larger AUC indicated higher risk (i.e., younger age) for Group 2 (higher frequency group).

Table 3:

Predicting Higher Frequency Group Across All Offending

Variable	Group 1: Lowest frequency (n = 266)	Group 2: Late onset (n = 31)	Group 3: Low frequency (n = 66)	Group 4: Highest frequency (n = 24)	F (df) or χ²(df)	ω²/Cramer’s V
Variable	M (SD) or n/N (%)				F (df) or χ²(df)	ω²/Cramer’s V
Demographic
Age at first police involvement¹	35.9^a (13.8)	32.7^a (7.5)	21.1^b (4.6)	17.8^c (2.9)	F (3, 91.0) = 120.7	.23
Age at first CSEM offense¹	36.7^a (13.6)	43.8^b (7.2)	30.4^c,d (8.5)	29.6^d (10.6)	F (3, 73.1) = 24.3	.07
Never married at the first CSEM offense	133/246 (54.1)	10/29 (34.5)	39/64 (60.9)	13/24 (54.2)	χ²(3, N = 363) = 5.7	.13
Unemployed	18/266 (6.8)	0/31 (0.0)	9/66 (13.6)	1/24 (4.2)	χ²(3, N = 387) = 6.9	.13
Criminal history
Total number of offenses¹	2.1^a (1.8)	8.2^b,c (2.9)	7.1^c (2.9)	19.5^d (7.5)	F (3, 57.6) = 132.7	.72
Total number of CSEM offenses^1a	1.2^a (0.5)	1.7^a (1.0)	1.5^a (1.0)	1.5^a (1.0)	F (3, 57.4) = 5.0	.05
Making CSEM offenses¹	0.4 (0.6)	0.5 (0.9)	0.5 (0.7)	0.5 (0.5)	F(3, 63.0) = 1.0	.001
Total number of contact sexual offenses	0.3^a (0.6)	1.0^b (1.2)	1.2^b (1.4)	1.3^b (2.1)	F (3, 56.6) = 13.5	.14
Total number of any sexual offenses	1.4^a (0.7)	2.8^b (1.8)	2.8^b (1.6)	3.0^b (2.6)	F (3, 56.3) = 24.4	.23
Total number of any violent offenses	0.4^a (0.7)	2.3^b (1.9)	1.8^b (1.7)	3.3^b (2.7)	F (3, 55.9) = 33.4	.35
Was CSEM offense the first offense?	187/266^H (70.3)	7/31^L (22.6)	8/66^L (12.1)	0/24^L (0.0)	χ² (3, N = 387) = 114.5	.54
Indicators of sexual interest in children
Volunteered with children (CASIC Item 5)	23/249 (9.2)	3/30 (10.0)	2/63 (3.2)	3/24 (12.5)	χ²(3, N = 363) = 3.1	.09
Indication of pedophilic/hebephilic interest (CPORT Item 5)	73/214 (34.1)	7/23 (30.4)	25/56 (44.6)	9/22 (40.9)	χ²(3, N = 315) = 2.7	.09
More boy than girl CSEM (CPORT Item 6)	38/234 (16.2)	2/29 (6.9)	9/61 (14.8)	6/23 (26.1)	χ²(3, N = 347) = 3.0	.10
More boy than girl child other material (CPORT Item 7)	39/233 (16.7)	2/29 (6.9)	11/60 (18.3)	5/23 (21.7)	χ²(3, N = 345) = 2.6	.09
Number of boy victims in a making CSEM offense	0.4 (0.6)	0.0 (0.0)	0.3 (0.5)	0.7 (0.6)	F(3, 47) = 1.5	.03
Number of minor victims in contact sexual offense	1.2 (0.5)	1.5 (0.8)	1.6 (1.1)	1.3 (2.1)	F(3, 29.1) = 2.0	.02
Psychological variables
Major mental illness known at the first CSEM offense	35/90 (38.9)	3/5 (60.0)	9/20 (45.0)	2/6 (33.3)	χ²(3, N = 121) = 1.2	.08
Personality disorder known at the first CSEM offense	7/87 (8.0)	0/4 (0.0)	4/18 (23.5)	0/4 (0.0)	χ²(3, N = 112) = 4.8	.21
Substance (alcohol and/or drug) use problem, generally	24/189 (12.7)	5/17 (29.4)	13/44 (29.5)	8/16^H (50.0)	χ² (3, N = 266) = 19.4	.27
Drug use problem, scored as severe	4/109 (3.7)	0/14 (0.0)	3/29 (10.3)	1/13 (7.7)	χ²(3, N = 165) = 3.2	.14
Alcohol problems, scored as severe	4/122 (3.3)	2/15 (13.3)	3/33 (9.1)	4/14^H (28.6)	χ² (3, N = 184) = 13.6	.27
Charge or conviction for DUI/drive while impaired	8/253^L (3.2)	8/30^H (26.7)	9/64 (14.1)	6/23^H (26.1)	χ² (3, N = 370) = 34.1	.30

Note. Continuous variables were tested using ANOVA. The ω² value of .01, .06 and .14 for ANOVA test represent small, medium, and large effects, respectively (Kirk, 1996). Dichotomous variables were compared across groups using chi-square tests. The Cramer’s V (for χ² test, df = 3) value of .06, .17, and .29 for χ² test represent small, medium, and large effects, respectively (Cohen, 1988). Boldfaced values were statistically significant (p < .05) for the omnibus test. CSEM = child sexual exploitation material; CASIC = correlates of admission of sexual interest in children; CPORT = Child Pornography Offender Risk Tool; ANOVA = analysis of variance.

Robust test of equality of means (Welch) were used when equality of variance was not assumed. The values in the same row not sharing the same superscript are significantly different based on the Games-Howell post hoc test (p < .05). ^HIndicates the observed number is significantly higher than the expected number. ^LIndicates the observed number is significantly lower than the expected number based on z-score (1.96).

Any Offending

For frequency across all types of offending, a four-group model provided the best model fit. The majority of men with CSEM offenses were in the lowest frequency group (Group 1; 69%, n = 266) with low rates of crime, including non-CSEM crime (i.e., they were specialists). The other groups were varying levels of generalists. One tenth of the cases were in a late onset group (Group 2, 8%, n = 31), with crimes appearing later, in their 30s, peaking in their late 40s, and reducing again in the late 50s. The second largest group had a slightly higher frequency of crime than Group 1, where the frequency of crime started reducing in the late 30s and became similar to Group 1 in the early 50s (this is Group 3 or low frequency group; 17%, n = 66). Finally, a fourth group—the highest frequency group—started offending in their teens with crime frequency reducing in their late 20s but still meaningfully higher than the other three groups until their late 50s (Group 4; 6%, n = 24).

Any offending trajectory groups differed on age at first police involvement (Cramer’s V = .23; a large effect), age at first CSEM offense (Cramer’s V = .07; a moderate effect), and number of offenses (Cramer’s V ranged from .05 to .72, Mdn = .29; moderate to large effect); see Table 3. The highest frequency offending group was younger in age at first police involvement (18 years old vs. lowest frequency: 36 years old; late onset: 31 years old; low frequency = 21 years old). In the lowest frequency group, CSEM was significantly more likely to be their first offense (70%), compared with the other groups (late onset: 22.6%; low frequency: 12.1%; highest frequency: 0%). Indicators of sexual interest did not differ across the four groups. The highest frequency offending group had the greatest number of substance use indicators (e.g., 50% with alcohol use problems, 26% driving while impaired/driving under the influence charges or convictions).

Discussion

Our findings indicate that men with CSEM offenses are heterogeneous in their criminal trajectories with an average of 4.5 offenses during the average 20 years follow-up (range = 0.03–61 years). The number and shape of the trajectories depended on the type of crime analyzed. Two groups emerged when we focused on sexual offending and violent offending (including contact sexual) offending. There were four trajectory groups when we examined the frequency of all types of offending. These trajectories were replicated whether we used age or year since first CSEM offense. Our sexual crime trajectory findings replicated the findings of Lussier and Davies (2011), with 96% of men with sexual offenses in their sample and 98% of men with CSEM offenses in our sample committing few (detected) sexual offenses. Although individuals with CSEM offenses and no known other sexual offending were mainly in the low sexual offending frequency group, results demonstrate things are not that simple. Some men with CSEM offenses are still prolific in their offending, either for their continual CSEM offending or for their other sexual and nonsexual crimes.

We found that offending trajectories were differentially associated with characteristics of these individuals. Similar to other offense types (e.g., Moffitt, 2006), age at first police involvement was a robust predictor of a more prolific criminal trajectory in men with CSEM offenses. Individuals with CSEM offenses who have more varied criminal careers tended to have a non-CSEM offense as their first charge and more substance use than those with less varied criminal careers. Sexual interests and behaviors outside of CSEM offending (e.g., never married, evidence of pedohebephilia) distinguished the early onset sexual trajectory group, but did not distinguish violent and any crime trajectories.

Implications for Practice

We found evidence that men who have committed CSEM offenses can also be understood within a developmental life course perspective. The heterogeneity in offending observed in the current study suggests that CSEM offending may be a transitory phase or one crime in a varied criminal career. Younger age of onset, substance use, and first police involvement that does not include CSEM may be reliable predictors of more varied and lengthy criminal trajectories among men with CSEM offenses. Sexual interests and behaviors outside of CSEM offending, including never having been married and indicators of sexual interest in children, predict a more prolific sexual offending trajectory but did not predict violent or any offending trajectories.

Prevention programs that target children and youth before they are further advanced in their criminal careers are important to reduce future criminal offenses (Letourneau et al., 2017; see also, Laws, 2000). The current study suggests that general intervention can also reduce CSEM offending. Indeed, CSEM was rarely the first offense that brought the individuals in contact with police. The highest frequency group in the any offending trajectory, for example, was in contact with police, on average, 12 years before their first CSEM offense (18 years old at first police contact vs. 30 years old at first CSEM offense). Only half (53%) of the late onset group of the sexual offending trajectory (i.e., men with CSEM offenses who tend to restrict their behavior to sexual offending and have fewer crimes) had a CSEM offense as their first offense; the remainder had non-CSEM offenses as their first contact with police.

Men with CSEM offenses with indicators of sexual interest in children were more likely to be classified as part of an early onset group in the sexual offending trajectory. As such, assessing for sexual interest in children is essential for those interested in understanding the risk posed by men with CSEM offenses. There are some opportunities to assess sexual interest in children during an investigation—such as using CASIC items (Seto & Eke, 2017), police asking about sexual interests during interviews (Seto & Eke, 2015), and other information gathered via online conversations (e.g., individuals indicating to others their lack of interest in adult relationships or specifically stating a sexual preference for children; Eke et al., 2019). There are also a selection of physiological and indirect assessments (e.g., Nunes & Pedneault, 2020, for review) and offense-based assessments (Seto & Lalumière, 2001) available to assess sexual interest in children.

Exclusive sexual offending (contact or noncontact, including CSEM) was rare among men with CSEM offenses. Other nonsexual crimes were much more frequent. The lack of evidence for specialization within CSEM offending suggests that risk assessment tools that predict general reoffending (e.g., Level of Service Inventory–Revised [LSI-R], Andrews & Bonta, 1995)—not just sexual reoffending—will be useful to identify individuals at risk of more varied and persistent criminal careers. Those with multiple offenses may also be a higher risk for past undetected contact sexual offenses (e.g., Long et al., 2016).

Limitations and Future Research Directions

Strengths of this study included our access to descriptive data about offenses and a relatively long follow-up period, averaging 20 years across individuals. A limitation of our trajectory modeling analyses was the small number of individuals in some groups (e.g., 8 men in the early onset sexual crime group). We used analytical strategies that were robust to low base rate events (AUC, Babchishin & Helmus, 2016; Hsu, 2002). Nevertheless, some findings relating to significant predictors could be unstable. The greater likelihood of missing information for some variables (e.g., psychological variables) also limited what we could learn about group differences.

We also did not have varied measures of sexual behaviors (e.g., risky sexual experiences) that have been predictive of problematic criminal trajectories (e.g., Mason et al., 2010). With information on classical criminological variables, such as school maladjustment and parental criminality, we may have found further evidence of differences (or similarities) between trajectories. There is also theoretical and clinical importance in considering how criminal career trajectories may relate to opportunity to offend and the relevance of life events (e.g., children, marriage) and interpersonal factors (e.g., personality, sex drive) in offending. The current study did not have time-varying factors, such as multiple assessments on dynamic risk tools or employment status, and, as such, could not examine more detailed models of desistance. Future research could apply latent growth modeling through a structural equation modeling framework to determine whether change on factors predict different trajectory outcomes (e.g., Lee et al., 2020). Studies that include social service records (e.g., child protection and welfare), self-reported behavior, and criterion validity measures would provide novel information and help validate our findings and further increase confidence in the identified groupings. In short, there is much more research needed, with the lengthy follow-up required for this type of analyses just now becoming available for CSEM offending.

Like almost all trajectory studies (see Jolliffe et al., 2017, for review), the current study is not a full life course analysis as we do not have data from birth to death necessary for strong life course conclusions. Available studies of the age–crime curve tend to sample multiple cohorts (cross-sectional design) rather than one cohort across time and, therefore, cannot disentangle the effect of aging with cohort and period effects (Loeber & Farrington, 2014). A benefit of the current study is that it follows one cohort of individuals; however, it is important to note that individuals’ first criminal involvement varied from 1979 to 2007 (Mdn = 2000). The crime curve findings suggest that offending typically emerges in adolescence and peaks in early adulthood, as with other types of offending. For a segment of our participants, their adolescence, however, would have occurred prior to the current ubiquity of the internet. This is relevant as access to the internet has been a well-documented factor in CSEM offending (e.g., Babchishin et al., 2011; Seto, 2014). Other research comparing data across the years has documented a decrease in perpetrator age in online offending (e.g., Wolak et al., 2011). The changing nature of the internet is different from other sexual or nonsexual offending where potential victims and opportunities have always been available, such as access to children or having intimate partners. As such, it is possible that a more recent cohort of individuals with CSEM offenses would find a greater number of individuals in an earlier onset group. It is also possible that some of our later adult onset group might actually have shown up in our early onset group if they had the opportunity to access CSEM online when they were younger.

Although the current study had a lengthy follow-up period, it was based on official criminal records, which underestimate the onset, frequency, and severity of offending (Loeber & Farrington, 2014; Piquero et al., 2012; Weinrott & Saylor, 1991). Offenses that are detected by police or reported to police are not a random subset of all offenses that a person commits. This issue is not unique to this study (Piquero et al., 2012). Our offense information was extensive; however, individuals might have had police contacts of which we were unaware. Not all charges or convictions are included or submitted to the national system we used, including some lesser or “summary” offenses, offenses diverted from the criminal justice system, and offenses committed outside of Canada. Some charges or convictions can be purged, including offenses committed as a youth or offenses as an adult that were pardoned. Juvenile offending is expected to be underrepresented in our sample, in part because the original sampling strategy necessitated that each individual had an adult conviction for CSEM. Therefore, juveniles who had committed CSEM offenses but did not have a CSEM offense in adulthood are not in our sample. Youth offending data were known if noted in the file information, rather than it being based on an examination of juvenile records, which are sealed or purged if an individual remains offense-free as an adult for a number of years, as specified in the Canadian Youth Criminal Justice Act. For those individuals in our sample who were younger at the time of their index CSEM offense, there is a greater likelihood that juvenile data would be noted in an investigative file. Analyses removing participants with offenses prior to the age of 18, however, found the same number of trajectories than when we included these participants.

Conclusion

We found evidence for two offending trajectories for sexual and violent offending, with a significant portion of our sample categorized as generalists rather than specialists. Lifelong patterns of sexual crimes were not common among individuals with CSEM offenses; nonsexual offenses were far more common. As such, general risk assessment tools would be important additions in risk evaluations of individuals with CSEM offenses. This study also suggests that early intervention of youth and young adults involved in the criminal justice system could also reduce future CSEM offending.

Supplemental Material

sj-docx-1-cjb-10.1177_00938548211040849 – Supplemental material for Applying Offending Trajectory Analyses to Men Adjudicated for Child Sexual Exploitation Material Offenses

Supplemental material, sj-docx-1-cjb-10.1177_00938548211040849 for Applying Offending Trajectory Analyses to Men Adjudicated for Child Sexual Exploitation Material Offenses by Kelly M. Babchishin, Angela W. Eke, Seung C. Lee, Nicole Lewis and Michael C. Seto in Criminal Justice and Behavior

Footnotes

Authors’ Note:

We would like to thank Detective Sergeant Bill Gofton, Detective Sergeant Kristina Truax, and Staff Sergeant Tanya Sampson for their comments on an earlier version of the manuscript. We also thank Kimberly Mularczyk for coding some of the additional data for this study. The views expressed in this article are that of the authors and do not necessarily reflect the views of their organizations.

ORCID iD

Kelly M. Babchishin

Supplemental Material

Supplemental Parts A, B, and C are available in the online version of this article at

Notes

Kelly M. Babchishin, PhD, is an assistant professor in the Department of Psychology, Carleton University and also serves as an action editor of the journal Sexual Abuse. She conducts research to identify factors associated with the onset and maintenance of sexually harmful behaviors with the ultimate goal of informing effective prevention programs. She has published extensively on sexual offending, pedophilia, and risk assessment.

Angela W. Eke, PhD, is a research coordinator in the Criminal Behaviour Analysis Unit of the Behavioural Sciences and Analysis Services, Ontario Provincial Police and an adjunct faculty member at Laurentian University, Ontario, Canada. She has published extensively on crime scene behavior, intimate partner violence, child sexual exploitation material, and risk assessment.

Seung C. Lee, PhD, is a research analyst for Public Safety Canada (Corrections Research Unit). He conducts research to investigate the international generalizability of risk assessment tools relevant to violent and sexual offending. He has published on the validity of risk assessment instruments of violent and sexual offending, as well as their applicability across ethnic minority groups (e.g., Hispanic, Black, Indigenous, and Asian) in North America.

Nicole Lewis is a research assistant in the Criminal Behaviour Analysis Unit of the Behavioural Sciences and Analysis Services, Ontario Provincial Police.

Michael C. Seto, PhD, is a registered clinical and forensic psychologist, director of the Forensic Mental Health Research Unit at the Royal’s Institute of Mental Health Research (IMHR), and professor in psychiatry at the University of Ottawa. He also serves as the editor-in-chief of Sexual Abuse. He has published extensively on pedophilia, sexual offending, and offending by individuals with mental disorders.

References

Adams

Flynn

(2017). Federal prosecution of commercial sexual exploitation of children cases, 2004–2013. Office of Justice Programs, Bureau of Justice Statistics, U.S. Department of Justice. https://www.bjs.gov/content/pub/pdf/fpcsecc0413.pdf

Andrews

D. A.

Bonta

J. L.

(1995). The level of service inventory-revised. Multi-Health Systems.

Babchishin

K. M.

Hanson

R. K.

Hermann

C. A.

(2011). The characteristics of online sex offenders: A meta-analysis. Sexual Abuse, 23(1), 92–123. https://doi.org/10.1177/1079063210370708

Babchishin

K. M.

Hanson

R. K.

VanZuylen

(2015). Online child pornography offenders are different: A meta-analysis of the characteristics of online and offline sex offenders against children. Archives of Sexual Behavior, 44(1), 45–66. https://doi.org/10.1007/s10508-014-0270-x

Babchishin

K. M.

Helmus

L. M.

(2016). The influence of base rates on correlations: An evaluation of proposed alternative effect sizes with real-world data. Behavior Research Methods, 48(3), 1021–1031. https://doi.org/10.3758/s13428-015-0627-7

Babchishin

K. M.

Merdian

H. L.

Bartels

R. M.

Perkins

(2018). Child sexual exploitation material offenders: A review. European Psychologist, 23(2), 130–143. https://doi.org/10.1027/1016-9040/a000326

Brandt

(2006). Delinquency, development and social policy. Yale University Press.

Bushway

Thornberry

Krohn

(2003). Desistance as a developmental process: A comparison of static and dynamic approaches. Journal of Quantitative Criminology, 19(2), 129–153. http://www.jstor.org/stable/23366838

Cale

Smallbone

Rayment-McHugh

Dowling

(2016). Offense trajectories, the unfolding of sexual and nonsexual criminal activity, and sex offense characteristics of adolescent sex offenders. Sexual Abuse, 28(8), 791–812. https://doi.org/10.1177/1079063215580968

10.

Cicchetti

D. V.

(1994). Guidelines, criteria, and rules of thumb for evaluating normed and standardized assessment instruments in psychology. Psychological Assessment, 6(4), 284–290. https://doi.org/10.1037/1040-3590.6.4.284

11.

Cohen

(1988). Statistical power analysis for the behavioral sciences (2nd ed.). Lawrence Erlbaum.

12.

Department of Justice Canada. (2019). Just facts: Sexual violations against children and child pornography. https://www.justice.gc.ca/eng/rp-pr/jr/jf-pf/2019/docs/mar02.pdf

13.

D’Unger

Land

McCall

Nagin

(1998). How many latent classes of delinquent/criminal careers? Results from mixed Poisson regression analyses. American Journal of Sociology, 103(6), 1593–1630. https://doi.org/10.1086/231402

14.

Eke

A. W.

Helmus

L. M.

Seto

M. C.

(2019). A validation study of the Child Pornography Offender Risk Tool (CPORT). Sexual Abuse, 31(4), 456–476. https://doi.org/10.1177/1079063218762434

15.

Elliott

I. A.

Mandeville-Norden

Rakestrow-Dickens

Beech

A. R.

(2019). Reoffending rates in a U.K. community sample of individuals with convictions for indecent images of children. Law and Human Behavior, 43(4), 369–382. https://doi.org/10.1037/lhb0000328

16.

Farrington

D. P.

(1986). Age and crime. In Tonry

Morris

(Eds.), Crime and justice: An annual review of research (Vol. 7, pp. 189–250). University of Chicago Press.

17.

Hanson

R. K.

Morton-Bourgon

(2005). The characteristics of persistent sexual offenders: A meta-analysis of recidivism studies. Journal of Consulting and Clinical Psychology, 73(6), 1154–1163. https://doi.org/10.1037/0022-006X.73.6.1154

18.

Harris

D. A.

Mazerolle

Knight

R. A.

(2009). Understanding male sexual offending: A comparison of general and specialist theories. Criminal Justice and Behavior, 36(10), 1051–1069. https://doi.org/10.1177/0093854809342242

19.

Henshaw

Ogloff

J. R. P.

Clough

J. A.

(2018). Demographic, mental health, and offending characteristics of online child exploitation material offenders: A comparison with contact-only and dual sexual offenders. Behavioral Sciences and the Law, 36(2), 198–215. https://doi.org/10.1002/bsl.2337

20.

Hsu

L. M.

(2002). Diagnostic validity statistics and the MCMI-III. Psychological Assessment, 14(4), 410–422. https://doi.org/10.1037/1040-3590.14.4.410

21.

Jolliffe

Farrington

D. P.

Piquero

A. R.

Loeber

Hill

K. G.

(2017). Systematic review of early risk factors for life-course-persistent, adolescence-limited, and late-onset offenders in prospective longitudinal studies. Aggression and Violent Behavior, 33, 15–23. https://doi.org/10.1016/j.avb.2017.01.009

22.

Jones

B. L.

Nagin

D. S.

(2007). Advances in group-based trajectory modeling and an SAS procedure for estimating them. Sociological Methods & Research, 35(4), 542–571. https://doi.org/10.1177/0049124106292364

23.

Jones

B. L.

Nagin

D. S.

Roeder

(2001). A SAS procedure based on mixture models for estimating developmental trajectories. Sociological Methods & Research, 29(3), 374–393. https://doi.org/10.1177/0049124101029003005

24.

Kim

Bushway

S. D.

(2018). Using longitudinal self-report data to study the age–crime relationship. Journal Quantitative Criminology, 34(5), 367–396. https://doi.org/10.1007/s10940-017-9338-9

25.

Kirk

R. E.

(1996). Practical significance: A concept whose time has come. Educational and Psychological Measurement, 56(5), 746–759. https://doi.org/10.1177/0013164496056005002

26.

Landis

J. R.

Koch

G. G.

(1977). The measurement of observer agreement for categorical data. Biometrics, 33(1), 159–174. https://doi.org/10.2307/2529310

27.

Laws

D. R.

(2000). Sexual offending as a public health problem: A North American perspective. Journal of Sexual Aggression, 5(1), 30–44. https://doi.org/10.1080/13552600008413294

28.

Lee

Moon

Garcia

(2020). The pathways to desistance: A longitudinal study of juvenile delinquency. Deviant Behavior, 41(1), 87–102. https://doi.org/10.1080/01639625.2018.1519138

29.

Letourneau

E. J.

Schaeffer

C. M.

Bradshaw

C. P.

Feder

K. A.

(2017). Preventing the onset of child sexual abuse by targeting young adolescents with universal prevention programming. Child Maltreatment, 22(2), 100–111. https://doi.org/10.1177/1077559517692439

30.

Loeber

Farrington

D. P.

(2012). From juvenile delinquency to adult crime: Criminal careers, justice policy and prevention. Oxford University Press.

31.

Loeber

Farrington

D. P.

(2014). Age-crime curve (pp. 12–18). In Bruinsma

Weisburd

(Eds.), Encyclopedia of criminology and criminal justice. Springer. https://doi.org/10.1007/978-1-4614-5690-2_474

32.

Long

Alison

Tejeiro

Hendricks

Giles

(2016). KIRAT: Law enforcement’s prioritization tool for investigating indecent image offenders. Psychology, Public Policy, and Law, 22(1), 12–21. https://doi.org/10.1037/law0000069

33.

Lussier

Blokland

(2014). The adolescence-adulthood transition and Robins’s continuity paradox: Criminal career patterns of juvenile and adult sex offenders in a prospective longitudinal birth cohort study. Journal of Criminal Justice, 42(2), 153–163. https://doi.org/10.1016/j.jcrimjus.2013.07.004

34.

Lussier

Davies

(2011). A person-oriented perspective on sexual offenders, offending trajectories, and risk of recidivism: A new challenge for policymakers, risk assessors, and actuarial prediction? Psychology, Public Policy, and Law, 17(4), 530–561. https://doi.org/10.1037/a0024388

35.

Lussier

McCuish

(2020). A developmental life-course view of sexual offending: Taking stock of research on the life-course development of antisocial and criminal behavior. In Proulx

Cortoni

Craig

L. A.

Letourneau

E. J.

(Eds.), The Wiley handbook of what works with sexual offenders: Contemporary perspectives in theory, assessment, treatment, and prevention (pp. 39–70). Wiley. https://doi.org/10.1002/9781119439325

36.

Lussier

Tzoumakis

Cale

Amirault

(2010). Criminal trajectories of adult sex offenders and the age effect: Examining the dynamic aspect of offending in adulthood. International Criminal Justice Review, 20(2), 147–168. https://doi.org/10.1177/1057567710368360

37.

Mason

A. W.

Hitch

J. E.

Kosterman

McCarty

C. A.

Herrenkohl

T. I.

Hawkins

D. J.

(2010). Growth in adolescent delinquency and alcohol use in relation to young adult crime, alcohol use disorders, and risky sex: A comparison of youth from low- versus middle-income backgrounds. Journal of Child Psychology and Psychiatry, 51(12), 1377–1385. https://doi.org/10.1111/j.1469-7610.2010.02292.x

38.

McManus

Almond

(2014). Trends of indecent images of children and child sexual offences between 2005/2006 and 2012/2013 within the United Kingdom. Journal of Sexual Aggression, 20(2), 142–155. https://doi.org/10.1080/13552600.2014.893031

39.

Moffitt

T. E.

(2006). Life-course-persistent versus adolescence-limited antisocial behavior (pp. 570–598). In Cicchetti

Cohen

D. J.

(Eds.), Developmental psychopathology: Risk, disorder, and adaptation. Wiley.

40.

Moreau

(2019). Police-reported crime statistics in Canada, 2018. Statistics Canada. https://www150.statcan.gc.ca/n1/en/pub/85-002-x/2019001/article/00013-eng.pdf?st=huoDIRDR

41.

Nagin

D. S.

(2005). Group-based modeling of development. Harvard University Press.

42.

Nagin

D. S.

Land

(1993). Age, criminal careers, and population heterogeneity: Specification and estimation of nonparametric, mixed Poisson model. Criminology, 31(3), 327–362. https://doi.org/10.1111/j.1745-9125.1993.tb01133.x

43.

Nagin

D. S.

Tremblay

R. E.

(2001). Analyzing developmental trajectories of distinct but related behaviors: A group-based method. Psychological Methods, 6(1), 18–34. https://doi.org/10.1037/1082-989X.6.1.18

44.

Nunes

K. L.

Pedneault

C. I.

(2020). Indirect and physiological approaches to assessing deviant sexual interests. In Proulx

Cortoni

Craig

L. A.

Letouneau

E. J.

(Eds.), The Wiley handbook of what works with sexual offenders: Contemporary perspectives in theory, assessment, treatment, and prevention (pp. 123–138). Wiley. https://doi.org/10.1002/9781119439325.ch7

45.

Piquero

A. R.

Hawkins

J. D.

Kazemian

(2012). Criminal career patterns. In Loeber

Farrington

D. P.

(Eds.), From juvenile delinquency to adult crime: Criminal careers, justice policy and prevention (pp. 14–16). Oxford University Press.

46.

Quetelet

(1984). Adolphe Quetelet’s research on the propensity for crime at different ages ( Sylvester

S. F.

, Trans.). Anderson. (Original work published 1831)

47.

Rice

M. E.

Harris

G. T.

(2005). Comparing effect sizes in follow-up studies: ROC area, Cohen’s d, and r. Law and Human Behavior, 29(5), 615–620. https://doi.org/10.1007/s10979-005-6832-7

48.

Sampson

Laub

Eggleston

(2004). On the robustness and validity of groups. Journal of Quantitative Criminology, 20(1), 37–42. http://www.jstor.org/stable/23366819

49.

Saramago

M. A.

Cardoso

Leal

(2020). Victim crossover index offending patterns and predictors in a Portuguese sample. Sexual Abuse, 32(1), 101–124. https://doi.org/10.1177/1079063218800472

50.

Seto

M. C.

(2013). Internet sex offenders. American Psychological Association.

51.

Seto

M. C.

(2014). Sex offender management assessment and planning initiative: Chapter 4—Internet-facilitated sexual offending. U.S. Department of Justice.

52.

Seto

M. C.

Eke

A. W.

(2015). Predicting recidivism among adult male child pornography offenders: Development of the Child Pornography Offender Risk Tool (CPORT). Law and Human Behavior, 39(4), 416–429. https://doi.org/10.1037/lhb0000128

53.

Seto

M. C.

Eke

A. W.

(2017). Correlates of admitted sexual interest in children among individuals convicted of child pornography offenses. Law and Human Behavior, 41(3), 305. https://doi.org/10.1037/lhb0000240

54.

Seto

M. C.

Hanson

R. K.

Babchishin

K. M.

(2011). Contact sexual offending by men with online sexual offenses. Sexual Abuse, 23(1), 124–145. https://doi.org/10.1177/1079063210369013

55.

Seto

M. C.

Lalumière

M. L.

(2001). A brief screening scale to identify pedophilic interests among child molesters. Sexual Abuse, 13(1), 15–25. https://doi.org/10.1023/A:1009510328588

56.

Smallbone

Marshall

W. L.

Wortley

(2008). Preventing child sexual abuse: Evidence, policy and practice. Willan Publishing. https://doi.org/10.4324/9781843925606

57.

Soothill

Francis

Sanderson

Ackerley

(2000). Sex offenders: Specialists, generalists-or both? The British Journal of Criminology, 40(1), 56–67. https://doi.org/10.1093/bjc/40.1.56

58.

Swets

J. A.

Dawes

R. M.

Monahan

(2000). Psychological science can improve diagnostic decisions. Psychological Science in the Public Interest, 1(1), 1–26. https://doi.org/10.1111/1529-1006.001

59.

Wakeling

H. C.

Howard

Barnett

(2011). Comparing the validity of the RM2000 scales and OGRS3 for predicting recidivism by internet sexual offenders. Sexual Abuse, 23(1), 146–168. https://doi.org/10.1177/1079063210375974

60.

Weinrott

M. R.

Saylor

(1991). Self-report of crimes committed by sex offenders. Journal of Interpersonal Violence, 6(3), 286–300. https://doi.org/10.1177/088626091006003002

61.

Whitaker

D. J.

Hanson

R. K.

Baker

C. K.

McMahon

P. M.

Ryan

Klein

Rice

D. D.

(2008). Risk factors for the perpetration of child sexual abuse: A review and meta-analysis. Child Abuse and Neglect, 32(5), 529–548. https://doi.org/10.1016/j.chiabu.2007.08.005

62.

Wolak

Finkelhor

Mitchell

(2011). Child pornography possessors: Trends in offender and case characteristics. Sexual Abuse, 23(1), 22–42. https://doi.org/10.1177/1079063210372143

63.

Youth Criminal Justice Act (S.C. 2002, c. 1)

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.38 MB