Do Developmental and Life-Course Theory Risk

Do Developmental and Life-Course Theory Risk Factors Equally Predict Age of Onset among Juvenile Sexual and Nonsexual Offenders?

Melanie Rosa University of South Florida Department of Criminology 4202 East Fowler Avenue Tampa, FL 33620-7200 Email: [email protected]

Bryanna Fox, PhD University of South Florida Department of Criminology

Wesley G. Jennings, PhD Texas State University School of Criminal Justice

Abstract: Previous empirical inquiries into the etiology of juvenile sex offending have been largely atheoretical. Consequently, a call for studies conducted utilizing Developmental and Life-Course (DLC) criminological theory has been made to better understand the onset, development, risk and protective factors of juvenile sex offending. Therefore, this study contributes to the discussion by testing key predictions proposed by the DLC framework regarding the theoretical correlates of early onset offending, as applied to juvenile sex offenders (JSOs) and juvenile nonsex offenders (JNSOs). Drawing upon a dataset of more than 64,000 youth referred to the Florida Department of Juvenile Justice, results indicate that although the number and severity of risk factors for early age of onset differ between the JSOs and JNSOs, the specific type of risk factors that emerged align with DLC theory predictions. The implications of these findings and contributions for DLC theory are also discussed. Keywords: developmental and life-course criminology, sexual offending, juveniles, risk factors Published in: Sexual Abuse Citation: Rosa, M., Fox, B.H., & Jennings, W.G. (2018). Do developmental and life-course theory risk factors equally predict age of onset among juvenile sexual and nonsexual offenders? Sexual Abuse, doi: 10.1177/1079063218797714. Go to: http://journals.sagepub.com/doi/abs/10.1177/1079063218797714

2 Do Developmental and Life-Course Theory Risk Factors Equally Predict Age of Onset among Juvenile Sexual and Nonsexual Offenders?

Research on juvenile sex offenders (JSOs) has been growing in recent years, perhaps in part due to the notable portion of sex crimes that these young offenders commit. Specifically, JSOs are responsible for 20% of all those arrested for sex crimes in the United States in 2013 (Federal Bureau of Investigation, 2014), and commit up to 20% of all rapes, half of all child sexual abuse, and 33% of all sex crimes against other juveniles in the United States. (Barbaree, Hudson, & Seto, 1993). In short, JSOs appear to be a sizable and problematic group of offenders who warrant further research on the development of this behavior, and how prevention strategies can be used to intervene prior to the onset of sexual offending. Several recent studies have made major contributions to our understanding of juvenile sex offending by examining trajectories of JSO behavior (Cale, Smallbone, Rayment-McHugh, & Dowling, 2016; Piquero, Farrington, Jennings, Diamond, & Craig 2012) and the continuity of sexual offending from adolescence into adulthood (Lussier, Corrado, & McCuish, 2016; Reingle, 2012; van Den Berg, Bijleveld, & Hendriks, 2017; Zimring, Piquero, & Jennings, 2007). This research on the rate, continuity, and desistance of JSOs has made significant headway toward understanding the development and life course of these young offenders, and led to important implications for both criminological theory and public policy. Specifically, this research has helped to illuminate the empirical realities that juvenile sex offending and recidivism are relatively uncommon, and sex offending continuity (i.e., sexual offending in adolescence and in adulthood) is even more rare. Thus, efforts to try and predict adult sex offending behavior (and develop policies such as registering JSOs for life) based on knowledge of sex offending behavior in adolescence almost always results in a false positive.

3 Other research has focused on the demographic, social, school, psychological, and family risk factors associated with juvenile sex offending (Baglivio et al., 2014; Calley, 2007; Fox, 2017; Hunter, Hazelwood, & Slesinger, 2000; Rice & Harris, 1997; Robertiello & Terry, 2007; Seto & Lalumière, 2010; Worling, 2001). In general, this research indicates that certain personality traits, psychopathologies, and childhood traumas are significant risk factors for juvenile sex offending. Specifically, Seto and Lalumière’s (2010) meta-analysis of 59 studies on the predictors and correlates of juvenile sex offending compared theoretically derived risk factors for adolescent male sex offenders with adolescent male nonsexual offenders. Results indicated that overall, deviant sexual interests, sexual and physical abuse in childhood, and certain psychopathologies were strong and significant predictors of juvenile sex offending versus nonsexual offending among the all-male samples (see also Hall & Hirschman, 1991, 1992; Marshall & Barbaree, 1990; Ward & Beech, 2006). Notably, JSOs scored lower than nonsex offenders in terms of criminal involvement, antisocial peer associations, and substance use, but JSOs still showed “extensive” histories of criminal behavior, conduct problems, and antisocial tendencies as compared to nonoffenders (Seto & Lalumière, 2010; see also Chewning, 1991; Etherington, 1993; Katz, 1990; Lindsey, Carlozzi, & Eells, 2001; Valliant & Bergeron, 1997). In a recent study examining the unique and significant risk factors for JSOs among male and female offenders in Florida, Fox (2017) found that having low empathy, high impulsivity, depression and/or psychosis; experiencing sexual abuse, an earlier age of criminal onset; and being male all significantly increased the risk of sexual versus nonsexual juvenile offending, even while controlling for all other theoretically derived measures in the analysis. Age of Onset and Juvenile Sex Offending Although this body of research has been an important step toward understanding the

4 correlates and criminal careers of these young offenders, far less research has examined the unique predictors of the age of onset for juvenile sexual offending. For instance, just two studies in Seto and Lalumière’s (2010) meta-analysis examined age of onset, and the results indicated that lower age at first intercourse predicted an earlier age of onset for JSOs. Specifically, Seto and Lalumière (2010) stated that childhood intercourse, likely caused by childhood sexual abuse, is related to the commission of a sex offense (Hanson & Bussiere, 1998; Hanson & MortonBourgon, 2005). Treating sexual abuse after it occurs is unlikely to reduce reoffending among identified JSOs, but preventing the abuse may also prevent sexual offending. Thus, identifying the risk factors which increase the likelihood of committing a sexual offense (and at an earlier age) may increase the opportunities to prevent sexual offending from ever occurring. Indeed, Seto and Lalumière’s (2010) meta-analysis prompted further inquiry into the age of onset for sex offenders (Lussier, Blokland, Mathesius, Pardini, & Loeber, 2015). In fact, Blokland and Lussier (2015) stated that “(t)he most important theoretical question to be answered is whether sex offenders should be regarded as similar or different to nonsexual offenders in terms of the etiology of their offending behavior” (p. 16). To address this question, Lussier and colleagues (2015) examined the childhood risk factors predictive of criminal onset in adolescence (below age 18) versus adulthood (age 18 and older) for sex offenders. This study, which drew upon a sample of 92 sex offenders, showed that youth onset offenders were more likely to live in a poor neighborhood, be exposed to structural deficits, have behavioral issue(s), have a psychological disorder, commit more delinquent acts, and be more sexually active, as compared with the adult onset offenders (Lussier et al., 2015). This study was among the first to demonstrate the wide variety of developmental correlates predictive of an adolescent versus adult onset of sex offending. Similarly, Carpentier, Leclerc,

5 and Proulx (2011) found that sexual offenders with an early onset, prior to age 12, were more likely to have early aggressive behavior, deviant sexual behavior, and parents who were victims of sexual crimes, as compared with those with later criminal onset in adolescence and/or adulthood. Taken together, these studies support the notion that risk factors predictive of early onset offending will likely differ from predictors of adolescent onset offending among JSOs. However, there has been little consensus on the reason for the earlier age of onset among JSOs. To address this gap, scholars examining the etiology of juvenile sex offending have argued the importance of examining JSOs using a theoretical lens, specifically developmental and life-course (DLC) criminological theories (Lussier, 2017; McCuish & Lussier, 2017). Specifically, Lussier (2017) noted that although developmental changes have been shown to correlate with juvenile sexual offending (such as childhood trauma and abuse), very little is known on the risk factors for the deviant sexual behavior, and particularly, at such an early age. Consequently, Lussier (2017) called for more research on JSOs from a DLC perspective, specifically requesting studies that test the applicability of DLC theories to this unique population, and identify which risk factors are most predictive both of sexual offending and early (vs. adolescent) age of criminal onset. DLC Theoretical Perspective DLC theories emerged as a prominent perspective in the 1990s by criminological researchers who believed that the risk and protective factors for offending, and the resultant criminal career patterns, are not the same for all individuals (Farrington, 2003; Loeber & LeBlanc, 1990). Therefore, DLC theories aim to identify the within- individual causes of criminal offending, and the various correlates of criminal behavior over the life course. Unlike other theoretical perspectives in criminology, DLC is comprised of multiple subtheories, each aiming

6 to explain three main issues: 1) the onset and development of criminal behavior over the lifecourse; 2) the risk and protective factors for crime at different ages; and 3) the effect of life events on the course of criminal development (Farrington, 2003). Although DLC theories have been very successful and increasingly popular within criminology (Farrington, 2003), the tenets of DLC theory have not been applied to examine the unique risk factors contributing to variations in age of criminal onset among JSOs. As Farrington (1989, 1991, 1998) argued that specific explanations were unnecessary for individuals involved in specific types of crime, DLC theory provides an appropriate framework to test whether theoretically derived risk factors are predictive of onset of offending for JSOs and juvenile nonsex offenders (JNSOs). Therefore, this study seeks to determine whether predictions made regarding the risk factors for early versus adolescent onset of criminal behavior, as proposed by the DLC theoretical framework, equally apply for juvenile sex and nonsex offenders. To do this, we draw on the theoretical predictions made about early versus adolescent onset by DLC theories to evaluate the applicability of these theoretical predictions for juvenile sexual and nonsexual offenders. DLC Theory and Early Age of Onset Several major DLC perspectives make predictions regarding the risk factors for early versus adolescent onset of criminal behavior. In general, DLC research suggests that the etiology of early onset offenders arises from neuropsychological deficits that interact with negative social environments such as the family, school, and community to produce the early start and continuity of criminal behavior (Farrington, 2003; Moffitt, 1993; Patterson, DeBaryshe, & Ramsey, 1989; Patterson & Yoerger, 2002; Piquero et al., 2012; Thornberry & Krohn, 2005). Specifically,

7 neuropsychological issues including impulsivity, inattention (i.e. attention-deficit disorder [ADD]), difficult temperaments (i.e., conduct disorder, aggression), and cognitive deficits, as well as weak parental attachment or lack of affection, structural adversity, low interest and performance in school, and abuse and/or neglect in childhood are recurring risk factors in the DLC literature (see Farrington, 2003; Lahey, Moffitt, & Caspi, 2003; Patterson et al., 1989; Patterson & Yoerger, 2002; Piquero, 2001; Piquero & Moffitt, 2005; Thornberry & Krohn, 2005). Early onset offenders are predicted to have few protective factors in terms of family, school, and community support (Farrington, 2003; Moffitt, 1993, Patterson & Yoerger, 2002; Thornberry & Krohn, 2005). DLC Theory and Adolescent Age of Onset With respect to “later onset” offenders, who typically show delinquent and criminal behavior in their early- to midteenage years, DLC theories predict that these youth are more susceptible to peer influences, versus family, school, and neuropsychological deficits (Farrington, 2010; Patterson et al., 1989; Patterson & Yoerger, 2002). Instead, adolescent onset offenders are more likely to offend due to social pressures such as peer influence and emulation of deviant peers, and then desist in early adulthood (Moffitt, 1993). As adolescence is a transition period when children become less dependent on parents and more concerned with individualism and peer acceptance, higher rates of delinquency (particularly for “adult-like” behavior such as underage drinking, sex, and drug use) are more common (Moffitt, 1993). However, as prosocial behaviors are more promoted in adulthood and the need to fill the “maturity gap” decreases, adolescent onset offenders desist (Moffitt, 1993). Current Study The purposive selection of the DLC theoretical framework emphasizes the developmental

8 aspect of offending, and unique relationships between psychological, biological, social, parental, and peer influences that may influence onset of offending at different ages. Given the largely atheoretical examination of JSOs and specific call for studies to utilize DLC theory to better understand the onset, development, and risk and protective factors of juvenile sex offending, this study tests whether DLC perspectives on early onset offending can be generalized to JSOs with criminal histories exclusive to sex offending. This study contributes to the discussion prompted by Lussier (2017) by focusing on two specific developmental stages (childhood and adolescence), and the corresponding DLCinformed risk factors for offending during each developmental stage. By examining the onset of JSO behavior, and the associated risk factors for it, we may be able to better understand the etiology of juvenile sex offending (and how it differs from early onset nonsexual offending) in three novel ways. First, including and testing DLC-informed risk factors for early onset juvenile sex offending can help address a notable limitation in existing research (Lussier, 2017). Second, this is one of the few studies conducted to specifically examine the predictors of age of onset among JSOs. Although it is the norm to employ age of onset as a control variable, in this study, onset of offending will be examined as an outcome measure to determine whether DLCinformed risk factors for early onset offending apply to JSOs. Finally, this study aims to determine whether specific DLC risk factors are unique to early onset juvenile sex offending versus nonsexual offending. These aims and novel contributions are also important as research on the etiology of juvenile sex offending has often not included a comparison group of nonsex offenders. To this end, Zakireh, Ronis, and Knight (2008) noted, “few studies have included appropriate comparison groups (e.g., juvenile offenders who have not committed sexual crimes). Without

9 such control groups, it is difficult to determine whether observed results are linked with sexual offending in particular or with delinquency in general” (p. 324). Specificity-design studies such as this allow for the determination as to whether a variable distinguishes early onset JSOs from nonsexual and adolescent onset offenders, and is, therefore, a potential causal candidate in the etiology of juvenile sex offending (Garber & Hollon, 1991). Ultimately, the current study represents an innovative approach to further understanding the etiology of juvenile sex offending. Building on previous studies and calls for this type of research (e.g., Fox, 2017; Lussier, 2017; McCuish & Lussier, 2017; Seto & Lalumière, 2010), this study draws on a large sample of male and female juvenile offenders to determine whether the DLC-informed risk factors for early onset (i.e., neuropsychological, family, structural deficits, school) versus adolescent onset can explain early onset offending specifically for JSOs with criminal histories exclusive to sex offending. Method Data The data set utilized in this study was retrieved from the Florida Department of Juvenile Justice (FDJJ), following approval from the lead author’s university institutional review board. Because Florida does not specify a minimum age for criminal responsibility, referrals for all juvenile offenders, as young as age 7, are included in the FDJJ dataset. In total, 64,329 juvenile sex and nonsex offenders were referred to FDJJ and followed until they aged out of the system between 2007 and 2012. In short, if a juvenile committed an offense (status, misdemeanor, felony) prior to age 18, the offense and the offender are included in this database. Per Florida statute, JSOs must have committed a misdemeanor or felony that met the requirements of a sex offense. Therefore, juveniles with a criminal history comprised entirely of

10 either misdemeanor and/or felony sex offense(s) made up the population of JSOs under study1 . Conversely, juveniles with a criminal history exclusive to nonsex offenses (violent, property, drug) comprised the population of JNSOs. In other words, juveniles classified as sex offenders could not have criminal histories with nonsex offenses and nonsex offenders could not have criminal histories with sex offenses. Therefore, out of the 64,329 juvenile offenders, 4,153 were JSOs whereas 60,176 were classified as JNSOs. Measures In addition to records of criminal history, a referral to FDJJ prompts the issuance of the Positive Achievement for Change Tool (PACT). This tool is used to assess the unique needs and future risk of each adjudicated juvenile offender for programming and placement purposes. The PACT provides extensive information on not readily available information, such as adverse childhood experiences (ACEs) and the presence of psychopathologies (for a complete review, see Fox, Perez, Cass, Baglivio, & Epps, 2015). This method of data collection has high validity because self-reported responses are verified by official records, to substantiate the veracity of claims. The FDJJ offender database and the PACT assessment are widely used in criminology due to the high validity of the measures (e.g. Baglivio et al., 2014; Baglivio et al., 2015; Fox, Perez, Cass, Baglivio, & Epps, 2015; Fox 2017; Fox & Delisi, 2018; Perez, Jennings, & Baglivio, 2018; Wolff, Baglivio, & Piquero, 2017) and large population of offenders. Interviews and file reviews are conducted to ensure the most accurate responses are obtained (Fox, Perez, Cass, Baglivio, & Epps, 2015) and Baglivio and colleagues performed quality checks on the

1

Due to data limitations, the temporal order and frequency of offenses among youth who commit sexual and nonsexual offenses (i.e. sex-plus offenders) are unknown. Therefore, to most clearly determine the risk factors predictive of early onset sex offending (and not general offending within the sex offender group), we use the sexonly (JSO) group.

11 PACT data which indicate the data are highly reliable (see Baglivio, 2009; Baglivio & Jackowski, 2013). Furthermore, several studies have utilized the same data and exact dichotomous coding as the current study (Fox 2017; Fox & Delisi, 2018) to indicate the presence or absence of specific risk factors for juvenile sex offending. Age of Onset This study draws on insights from Erikson’s (1963) psychosocial developmental stages to guide the operationalization of age of onset. Erikson proposed two major phases of development relevant to this study (i.e., ages where youth in this study began offending). The first stage, industry versus inferiority, emphasizes the school years in childhood (ages 5-12). This is a period when children begin to juggle the growing demands of school, friendships, and extracurricular activities. Conversely, the later ego identity versus role confusion stage (ages 13-18) demarcates childhood from adolescence, where autonomy takes precedence (McMaken, 2000). To determine whether DLC perspectives can be generalized to onset of juvenile sex offending, and not general offending with sex offense histories, the population was partitioned based on the youths’ offense histories (exclusive sex offenses vs. exclusive nonsex offenses) and age of onset according to Erikson’s (1963) definition. Specifically, all those with a criminal onset at 12 years of age or less were classified as having an “early” onset, while those who began their offender careers between 13 and 18 were classified as “adolescent” onset. The resultant outcome variable represents a dichotomy, where 1 = early onset and 0 = adolescent onset. Of the JSOs, 1,405 were identified as early onset, whereas 2,748 were identified as adolescent onset offenders. Of the JNSOs, 14,613 were early onset offenders, whereas 45,563 identified as adolescent onset offenders.

12 DLC Risk Factors for Early Age of Onset Demographics. The demographic control measures of gender and race are included to determine whether, and to what extent, the likelihood of being an early onset versus adolescent onset JSO versus JNSO varies by gender or race (see Ellis, Beaver, & Wright 2009; Fix, Cyperski, & Burkhart, 2017; Murphy, DiLillo, Haynes, & Steere, 2001). Gender is a dichotomous variable, where 1 = male and 0 = female. Race is coded 1 = White and 0 = nonWhite. Structural deficits. DLC theories suggest that less than ideal structural factors may place a strain on caretakers, serving as a hindrance to a successful parent-child relationship (Farrington, 2005; Murray & Farrington, 2010). To tap into potential financial strains, annual family income is included as an ordinal variable, where 0 = less than US$15,000, 1 = US$15,000 - US$34,999, 2 = US$35,000 - US$49,999, and 3 = US$50,000+. Child Maltreatment. Research indicates that childhood trauma, abuse, and maltreatment relate to higher risk of offending, known as the victim-offender overlap (see Baglivio, Wolff, DeLisi, Vaughn, & Piquero, 2017; Jennings, Higgins, Tewksbury, Gover, & Piquero, 2010; Widom, 1989). These ACEs, particularly childhood sexual abuse, have also been shown to be predictors of sexual offending among juveniles (Fox, 2017; Morais, Alexander, Fix, & Burkhart, 2018; Seto & Lalumière, 2010). For example, McCuish, Cale, and Corrado (2017) found that adolescent sex offenders are more likely than adolescent nonsex offenders to have sexual abuse histories and that family histories of sexual abuse have a strong and significant effect on youth sexual abuse and, consequently, sex offenses committed by youth (McCuish et al., 2017). This study, therefore, examines four types of child maltreatment as predictors of sexual offending and age of onset: emotional neglect, physical abuse, sexual abuse, and

13 witnessing household violence. All are dichotomous measures, where 1 = history of neglect, abuse, or exposure to violence, and 0 = no history of childhood trauma. School. A dichotomous variable of youth’s involvement in school activities was used, where 1 = not interested in school activities and 0 = interested or involved in school activities. Familial adversity. Put forth as one of the dominant risk factors during childhood, DLC theories emphasize the salient role parents play, and if that role becomes tenuous, the effects may lead to an early onset of offending (Baglivio, Wolff, Piquero, & Epps, 2015; Fox, Jennings, & Farrington, 2015). To examine whether, and to what extent, familial adversity uniquely identifies onset among juvenile offenders, four measures were used: parental supervision (1 = sporadic or inadequate supervision, 0 = consistent supervision), punishment (1 = consistently overly severe, erratic, or insufficient punishment, 0 = consistently appropriate punishment), mental health issues (1= yes, 0= no), and incarceration history (1= yes, 0= no), which were all dichotomous. Neuropsychological deficits. Although neuropsychological deficits cannot be measured directly, five psychopathologies are used in the analysis: ADD/attention-deficit hyperactivity disorder (ADHD), anger/irritability, empathy, impulsivity, and psychosis. These psychopathologies are theoretically predicted to relate to an earlier onset of offending according to DLC theory, and have been shown to predict sexual offending, as well as the type of sexual offending behavior and the rate of sex crime recidivism (Becker, Harris, & Sales, 1993; DeLisi et al., 2008; Fagan & Wexler, 1988; Fox, 2017; Fox & DeLisi, 2018; Hanson & Bussiere, 1998; Robertiello & Terry, 2007; Worling, 2001). Both impulsivity and ADD/ADHD measures are dichotomized, with 1 = presence of the specific psychopathology. Conversely, empathy is a binary measure, where 1 = unempathetic towards victims and 0 = empathetic to victims. Anger/irritability is a dichotomous variable, where 1 = consistently gets angry or irritable and 0

14 = never or occasionally gets angry or irritable. To distinguish between juveniles with psychosis versus those without, a dichotomous variable was included, where 1 = psychotic symptoms and 0 = no psychotic symptoms. Neuropsychological deficit (ND) * Familial adversity (FA). To determine whether, and to what extent, the presence of both neuropsychological deficit (ND) and familial adversity (FA) measures are predictive of an early onset of offending for both JSOs and JNSOs, an interaction term was included (Moffitt, 1993). To appropriately test Moffitt’s ND*FA prediction, variables were created for both ND (ADD + anger + empathy + impulsivity + psychosis) and FA (parental supervision + parental punishment + parental incarceration + parental mental illness), which were then multiplied together to create the interaction term. Therefore, juvenile offenders with nonzero values must be afflicted by both ND and FA measures. DLC Risk Factor for Adolescent Age of Onset Peers. Indicative of an adolescent onset of offending, DLC predicts that peer influence is highly pronounced during adolescence, when parental influence begins to wane and is replaced by peer influence and social factors (Agnew, 1991; Akers, Krohn, Lanza-Kaduce, & Radosevich, 1979; Chung & Steinberg, 2006; Simons, Whitbeck, Conger, & Conger, 1991). To measure peer influence, a single dichotomized variable was included in the model, where 1 = strongly or somewhat admires/emulates antisocial peers and 0 = does not admire/emulate antisocial peers. Analytic Technique To determine whether, and to what extent, DLC theoretical predictions regarding risk factors for early versus adolescent offending can be generalized to JSOs, the analysis proceeds in three stages. First, descriptive statistics and chi-square tests of association and odds ratio (OR) effect sizes are presented to determine the bivariate relationships between DLC theoretically

15 informed risk factors for early and adolescent onset of offending among JSOs and JNSOs (Table 1)2 . Next, multivariate binary logistic regressions are conducted to determine whether DLC risk factors significantly predict early versus adolescent criminal onset among JSOs and JNSOs, after controlling for all other theoretical and demographic covariates. As DLC theory emphasizes the cumulative impact of neuropsychological deficits (NDs) and familial adversity (FA) in increasing the risk of chronic, violent, and other types of severe offending (i.e., that the cumulative effect of experiencing familial adversity and neuropsychological deficits is the strongest risk factor for early onset for all offending types; see, for example, Moffitt, 1993), moderation analyses examining the interaction between ND and FA on the risk of early versus adolescent onset among the JSO and JNSOs are run for each analysis. Finally, a multinomial logistic regression is conducted to determine whether DLC risk factors significantly distinguish early onset JSOs and early onset JNSOs, when compared with juvenile offenders of any kind with adolescent onset. The interaction term of ND and FA is also included to evaluate the moderating effect of these cumulative risks on early onset JSO and JNSOs as compared with all forms of adolescent onset juvenile offenders. Together, this analytic plan will allow for an assessment of whether, and to what extent, DLC theoretical predictions align with the observed risk factors for early and adolescent onset of offending among the two distinct groups of juvenile offenders. Results Multivariate Analysis: JSOs As noted in the DLC literature, criminal behavior, family adversity, neuropsychological deficits, and childhood trauma tend not to occur in isolation. Therefore, prior to running

2

For a description of all variables across early and adolescent onset for juvenile sex and nonsex offenders, including chi-square measures of associations and odds ratio (OR) effect sizes, please see Table 1.

16 multivariate regression models, an analysis of multicollinearity was first conducted. Diagnostic results show that although several variables tend to occur in tandem, the variance inflation factors (VIFs) for all covariates in the analyses were satisfactory (range = 1.05 - 1.71). As the next step of the analytic process, a series of multivariate logistic regression models were estimated to determine the ability of the above noted DLC risk factors to predict early versus adolescent onset of offending, when controlling for all other covariates, for both juvenile sex and nonsex offenders (Table 2). The initial models (labeled Models 1 and 3 in Table 2) evaluate the theoretical and demographic risk factors’ ability to predict age of onset for the JSO and JNSO groups. A second set of models containing the summed ND and FA interaction term was added into the models for both JSO and JNSO groups (labeled Models 2 and 4 in Table 2), to perform a more precise test of the predictions made by DLC theory. Results of the multivariate binary logistic regression models distinguishing early versus adolescent onset among the JSO and JNSO youth are presented in Table 2 below. In each case, early onset is the predicted outcome, and adolescent onset is the reference category. Four DLC risk factors emerged as statistically significant predictors of an early onset of juvenile sex offending in Model 1, after holding all other theoretical and demographic variables constant: emotional neglect, parental incarceration, ADD/ADHD, and anger/irritability. JSOs who experienced emotional neglect were 24% more likely to have early versus adolescent onset (OR = 1.24, p < .05), having a parent incarcerated increased the odds of early onset among JSOs by 41% (OR = 1.41, p < .001), a history of ADD/ADHD increased the odds of early onset by 52% (OR = 1.52, p < .001), and anger/irritability increased the odds of early onset by 28% (OR = 1.28, p