Better Beginnings, Better Futures Study: Delinquency Trajectories of At-Risk Youth

Better Beginnings, Better Futures Study: Delinquency Trajectories of At-Risk Youth PDF Version (1.7 MB)

ISSN: 978-1-100-18384-8

Research Report 2011-03

Wendy Craig^{Footnote 1}

Kelly Petrunka^{Footnote 2}

Shahriar KhanFootnote 2a

Executive summary

Many studies of juvenile delinquency over the past two decades have focused on older, serious, and violent juvenile offenders. Younger delinquents have been ignored partly because their number is relatively small and their threat is often not as immediate. Understanding the trajectories of delinquency at a young age and the risk and protective factors associated with those developmental trajectories can inform the development of early risk assessments and the development of targeted prevention and intervention programs. The objectives of the research were to identify early trajectories of delinquency for both boys and girls from age 8 (Grade 3), age 11 (Grade 6), and age 14 (Grade 9) in a longitudinal sample of at-risk youth from a multi-informant perspective, assess risk and protective factors that may influence the likelihood that youth will engage in criminal behaviour in adolescence, and examine whether youth in the identified delinquency trajectories differ substantially in terms of delinquency, involvement with the criminal justice system, emotional and behavioural problems, experience of abuse, academic/school functioning, and health/health risk behaviours. Additionally, this study aimed at estimating the costs associated with each delinquency trajectory on utilization of government resources in the criminal justice system, remedial education, health care and social services, and social assistance.

In order to examine these research questions, analyses were conducted using the Better Beginnings, Better Futures data. These data followed 842 children living in five disadvantaged communities in Ontario. The same children were assessed when they were in Grades 3, 6, and 9 with measures largely based on the National Longitudinal Survey of Children and Youth (NLSCY). Three key informant sources were used to assess children's delinquency (parents, teachers and self-report youth ratings). In Grade 3, children's levels of delinquency were assessed by teachers. In Grade 6, the children were assessed by parents, teachers and the youth, while in Grade 9, they were assessed by parents and the youth. In addition to the above, 31 risk factors and 17 protective factors for delinquency were examined when the children were in Grade 3. When the children were in Grade 9, 41 outcome measures were examined in the following domains: emotional and behavioural problems, delinquency problems, abuse, involvement with the criminal justice system, functioning in school, and health and health risk activities. Finally, monetary costs associated with the criminal justice system, remedial education, health care and social services, and social assistance were estimated for each participant.

The literature on delinquent trajectories identifies three main delinquency groups among children and youth: a low delinquency group, a high delinquency group, and a desisting delinquency group. The trajectory analyses of the current research indicated that there were six delinquency trajectory groups. Children in two of the trajectories had very low ratings of delinquency across time (lowest delinquency group and the second lowest delinquency group). Two other trajectories showed a similar pattern of delinquency ratings that was decreasing over time. In the moderate desisters group, children had moderate levels of delinquency at Grade 3 followed by low levels of delinquency at Grades 6 and 9. In the highest desisters group children had the highest level of reported delinquency behaviours at Grade 3, followed by a marked decrease in reported delinquency at Grades 6 and 9. The fifth trajectory group, named escalators, had very low levels of reported delinquency at Grade 3 and increased markedly in their reported delinquency over time. By Grade 9, children in this trajectory group had the second highest delinquency scores. The final group, high delinquency, started with moderate levels of reported delinquency at Grade 3, marked by the highest reported levels of delinquency at Grades 6 and 9 of any of the trajectory groups.

Children at risk for delinquency (i.e., those in the high delinquency, escalators, and the two desisters trajectory groups) scored significantly higher on 17 of the 31 individual, family, peer, and neighbourhood risk factors. For example, children from these four trajectory groups experienced more hyperactive, oppositional-defiant, and physically aggressive behaviours; family risk factors included single parenthood, less parental education, public housing, and hostile-ineffective parenting. These results highlight the need to further develop and refine assessment tools to include these risk factors associated with delinquency. By Grade 9, the high delinquency and escalators groups also had significantly more problems than the other groups; they exhibited more emotional/behavioural, health, criminal, and school functioning problems. Early identification at school and involvement in special education programs early may have significantly reduced these negative outcomes in Grade 9.

Finally, the economic analyses identified that youth in the high delinquency, escalators, and the two desisters trajectory groups cost a significant amount of money; for example, approximately 80% of the estimated costs to society (e.g., on utilization of government resources in the criminal justice system, remedial education, health care and social services, and social assistance) were from these four trajectory groups which represent 18% of the sample. Furthermore, 80% of the estimated criminal justice costs were due to the youth in the high delinquency and escalators groups.

The findings of the current study highlight some key conclusions. First, there are early indicators to the developmental pathways to delinquency. The risk factors associated with delinquency involvement (e.g., inattention/hyperactivity problems, oppositional defiant problems, low family functioning, having a teenage mother) can be identified as early as Grade 3 and can inform the implementation of an assessment and/or screening tool for children and youth at-risk of delinquency. Second, delinquency involvement does not just emerge, it develops over time, and without intervention, the problems accumulate and may become serious and significant by as early as Grade 9. Third, investment in prevention, such as educational support, can reduce criminal justice costs and delinquency involvement. The most at risk groups (high delinquency and escalators groups) for delinquency involvement accounted for the majority of the estimated reactive costs (e.g., criminal justice, health care and social services, social assistance) and not the preventative costs (e.g., remedial education). Specifically, the high delinquency and escalators groups accounted for 46% of the reactive costs compared to 32% for the two desisters groups and 22% for the two low delinquency groups; for the preventative costs, high delinquency and escalators groups accounted for 38% of the costs compared to 44% for the two desister groups and 18% for the two low delinquency groups.

Although more research is needed to understand the delinquency trajectories of girls, those at-risk of delinquency appear to require more support. Although our high risk group of girls was limited, there are some preliminary indications from this research that they are at a heightened risk for problems (e.g., emotional problems, having delinquent friends, police involvement) and the estimated costs associated with their problems may be higher than for boys because they appear not only through the criminal justice system, but also through the health care system.

Introduction

Delinquency is one of the most prevalent problem behaviours engaged in by Canadian youth. Statistics Canada (Savoie, 2006) indicates that over one-third of youth have been involved in some form of delinquency by the age of fourteen and that childhood delinquency tends to predict violent behaviours throughout the course of a lifetime. Although delinquency covers a wide range of behaviours, many of which do not go reported to the police, about 5% of Canadian youth have been charged with federal offences (Savoie, 2006). Engaging in early delinquent behaviour (i.e., before age ten) has been linked to negative psychological, emotional, health, social, academic, employment, and later criminal outcomes (Boyd et al., 2005; Lacourse, Nagin, Tremblay, Vitaro, & Claes, 2003). Nonetheless, not all early starters go on to become serious delinquents. The growing body of knowledge that forms developmental prevention science allows for the identification of risk factors associated with delinquency; the development of screening procedures to identify children at risk of delinquency; and the implementation of preventive intervention for changing the risk factors associated with delinquency and reducing children's probability of engaging in antisocial behaviour. In this paper, we examine the developmental trajectories of delinquency, and the associated individual, family, peer, and school correlates and outcomes in order to inform the prevention of delinquency. We also provide an economic analysis of the costs associated with early pathways associated with delinquent behaviours.

Development of Delinquent Behaviours

Several studies have used trajectory analysis to distinguish individual patterns of delinquent behaviour from childhood to adolescence (e.g., Hoeve, Blokland, Dubas, Loeber, Gerris, & Van Der Lann, 2008; Schonberg & Shaw, 2007; Wiesner & Windle, 2006). A review of these studies highlights several important themes. First, on average, between three and six groups of delinquent behaviours tend to be identified by the trajectory methodology. There are three consistent trajectories (although differentially labelled) across these studies: a low delinquency group (representing the majority of youth who rarely engage in delinquent behaviour), a high delinquency group (representing a small minority of youth with an early stage of high level of delinquent behaviour and increase over time), and a desisting delinquency group (representing a minority of youth who start with a high level of delinquent behaviour and decrease with time). In studies where more than three trajectories have been found, the three consistent groups are usually subdivided into other groups. For example, Lacourse, Côté, Nagin, Vitaro, Brendgen, and Tremblay (2002) found six trajectories that included the three above as well as a low rising, a low decline, and a medium decline of involvement in crime. The second important consistency across studies is that by the end of adolescence, most trajectory groups are on the decline with respect to delinquent behaviour.

Methodological differences may account for some of the discrepancies in these studies' results. First, while all the studies included self-report measures, some also included court records (Hoeve et al., 2008) and teachers' and parents' ratings (Lacourse et al., 2002). Second, the studies varied with the geographical region, for example some studies have participants from urban United States (Hoeve et al., 2008) or urban French-speaking Canadians (Lacourse et al., 2002). Third, the studies varied with respect to the age of the participants and have primarily focused on older students. Fourth, with a few exceptions, the studies included only boys. Although fewer girls than boys engage in high levels of problem behaviours, those girls who do start early and persist in antisocial behaviours experience mental health problems at levels equal to their antisocial male counterparts (Odgers et al., 2008). Thus, there are limited data available on the trajectories of delinquency in girls. Fifth, some of the studies conceptualized delinquency broadly and examined externalizing behaviours (i.e., conduct problems, physical aggression, oppositional behaviour, hyperactivity) as opposed to delinquency (defined by violations of the Criminal Code). Sixth, studies varied with respect to the number of assessments and the timing of assessments used to derive the trajectories. Thus, the differences in the shape and the number of the trajectories may in part be influenced by the operationalizing of delinquency and the study design. Despite these methodological differences across studies, the consistent finding of at least three similar trajectories on different populations and cultures provides strong test re-test reliability for the existence of the three main delinquent trajectories.

Risk and Protective Factors

Identifying the trajectories of delinquency provides an understanding of how this behaviour changes with age, gender and other risk factors. There are two types of risk factors that are typically defined as static and dynamic. Static risk factors refer to historical variables that are resistant to change such as age at first offence, prior criminal history whereas dynamic risk factors are changeable (Andrews & Bonta, 1998). The most useful risk factors to identify from a prevention and intervention perspective are dynamic, because these factors are amenable to change. Identification of the individual, family, peer, and community risk and protective correlates of each of the trajectory groups can provide specific direction for the development of prevention and intervention programs.

The development of delinquent behaviour is influenced by risk and protective factors residing both within individuals and their environments. Risk factors are those that lead directly to problem behaviour whereas protective factors operate to buffer risk. Protective factors generally refer to influences that modify, ameliorate, or alter a person's response to some risky environmental conditions that may result in maladaptive behaviour. Rutter (1986) points out that: (1) protective factors do not necessarily mean positive experiences; (2) protective factors are detectable only for high-risk individuals; and, (3) protective factors can be non-environmental and part of the biological make-up of the individual. Protective factors should not be considered to be merely flip sides of risk factors. Protective factors operate under conditions of risk. That is to say, protective factors operate to prevent delinquency under high-risk conditions or among high-risk individuals.

There is a cumulative effect of risk and protective factors both within and across time. At a given point in time, children are at greater risk for juvenile delinquency if they experience multiple risk factors (Lerner, 1996). Over time, there is a progressive accumulation of the consequences of individual factors (cumulative continuity) and the responses they elicit during social interactions (interactional continuity). Within this developmental framework, life phases and transitions are particularly important in understanding behaviour because they present either crises or challenges, engendering stress that can undermine development or revealing resources and opportunities (Lerner, 1996). A developmental perspective considers both stability and transformations in behaviour in their developmental context. The challenge is to explain the emergence and the change in form and frequency of antisocial and delinquent behaviours over the course of development.

The correlates of juvenile delinquency are similar in males and females. It remains unclear, however, the extent to which the outcomes of early externalizing problems are the same for both genders. The developmental trajectories of aggressive girls may involve similar processes to those of boys but result in different outcomes. For example, girls' trajectories to delinquency indicate there is strong comorbidity with depression and suicidal ideation, as well a physical and sexual victimization (Moffitt, Caspi, Rutter, & Silva, 2001). The developmental trajectories of aggressive girls exemplify the joint processes of cumulative and interactional continuity. They are maintained by individual characteristics of the girls themselves and by their interactions within the family, school, peer, and marital systems. There is emerging evidence that the risks experienced by aggressive girls may be transferred to the next generation through their ineffective parenting practices as well as their genes (Serbin et al., 2004). In summary, many risk and protective factors have been identified by researchers. However, there exist limited data on girls' involvement in delinquency and whether there are specific or nonspecific risk and protective factors for girls.

Estimated Costs Associated with Delinquency

There are significant individual, justice, health and social services, and societal costs associated with delinquency. These high intra-personal, interpersonal, and societal costs highlight the need to increase our understanding of delinquency behaviour, before it emerges. Despite the well documented individual, physical, psychological and mental health, social, and criminal outcomes of engaging in delinquent behaviours, there are limited data available in Canada on the costs associated with it. There is limited research on the costs of delinquency beyond costs savings of early prevention programs on future delinquency, and the costs to the criminal justice system. Antisocial youth tend to be multiple offenders and Cohen (1998) found that the average delinquent commits 68-80 crimes over their delinquency time period and costs society between $1.3-$1.5 million (U.S. dollars). Early intervention programs have the potential to reduce the long term costs of delinquency. Cohen and Piquero (2009) estimated that a beneficial prevention program of diverting a 14-year-old high risk juvenile from a life of crime could save from $2.6 million to $5.3 million (U.S. dollars).

Few studies of early childhood prevention programs for children have included an economic analysis (e.g., Barnett & Masse, 2007; Karoly, Kilburn, & Cannon, 2005; Mrazek & Brown, 2002; Nores, Belfield, Barnett, & Schweinhart, 2005; Peters et al., 2010; Reynolds, Temple, Robertson, & Mann, 2002; Waddell, Hua, Garland, Peters, & McEwan, 2007). All these early childhood intervention studies have reported economic analyses based on follow-up data for children, and in some cases their parents, to the child's age of 15, 21, and/or 40. Economic analyses results from these studies provide the rationale to policy makers for investing in early childhood interventions. For most economic analyses of early childhood education programs, economic benefits are typically divided into three categories: benefits to program participants (e.g., increased income from improved education), benefits to non-program participants (e.g., reduced costs to crime victims), and benefits to government/taxpayers (e.g., decreased remedial education costs, decreased costs to the justice system). The Canadian study of early intervention, discussed in this paper, is on Better Beginnings, Better Futures (BBBF; Peters et al., 2010). The costing perspective of the Canadian BBBF economic analysis was the government/taxpayers; Karoly et al. (1998) refer to this analysis as cost-savings analysis to differentiate it from the more traditional cost-benefit analysis. In this paper, we examine the social, health, educational and juvenile justice costs for each of our trajectories of delinquency.

Objectives of the study

The current study used data drawn from a longitudinal research study, Better Beginnings, Better Futures (Peters, Petrunka, & Arnold, 2003), which examined the long-term impacts of an early childhood prevention program. More specifically, the research project used a longitudinal sample of 842 at-risk youth from a multi-informant perspective (i.e., parents, teachers, self-reported youth ratings) to: (1) identify early trajectories of delinquency for both boys and girls at age 8 (Grade 3), age 11 (Grade 6), and age 14 (Grade 9); (2) examine risk and protective factors at the individual, family, peer, school, and community levels that may influence the likelihood that youth will engage in criminal behaviour in adolescence; (3) examine whether youth in the identified delinquency trajectories differ substantially in Grade 9 on emotional and behavioural problems, delinquency, experience of abuse, involvement with the criminal justice system, academic/school functioning, and health/health risk behaviours; and (4) estimate the costs to government associated with each delinquency trajectory on utilization of government resources in the criminal justice system, remedial education, health care and social services, and social assistance.

The BBBF data are the only existing Canadian data that include a large number of male and female youth living in neighbourhoods characterized by poverty. The dataset is also diverse in terms of ethnicity and other family demographic variables. The results are further instructive given this is the first early childhood prevention project in Canada to include an economic analysis of the estimated costs and savings to government. Thus, this research has the potential to provide empirically-based information for communities in Canada regarding identifying children and youth at risk of involvement in antisocial and delinquent behaviours, as well as for designing prevention and intervention programs that are community-based and that target empirically-based risk and protective factors associated with delinquency among children and youth.

Method

Participants

In the BBBF sample, the longitudinal research cohort was comprised of a focal cohort and a following cohort. Children in the focal cohort (n = 721) were born in 1989 and were recruited to the longitudinal study between Junior Kindergarten (JK) and Grade 3, mostly through the school system. Children in the following cohort (n = 238) were born in 1990, and were recruited to the longitudinal study when they were in Grade 3. For this study, there were 842 participants (396 girls and 446 boys), representing 88% of the original sample. These participants represent the longitudinal follow-up of the BBBF study and had data at ages 8 (Grade 3), 11 (Grade 6), and 14 (Grade 9).

Attrition was mainly due to two factors: (1) families relocated and the researchers were unable to contact them; and (2) families declined to be interviewed. As a test for attrition bias, we employed logistic regression to examine sociodemographic differences in children and families who dropped out of the research cohort between Grade 3 and 6 and between Grade 6 and 9, and families who completed all years of data collection. These analyses indicated no significant differences in sociodemographic variables between the retained and lost cases.

Approximately 30% of the households were headed by single parents, 34% of parents did not complete high school, 59% of families were living below Statistics Canada Low Income Cut Off line, and 19% were living in public housing. There were no significant gender differences on any of the demographic variables. Appendix A provides more information on the family demographics when the children were in Grade 3.

Measures Delinquency

Child delinquency measures were created using items from the National Longitudinal Survey of Children and Youth (NLSCY; Statistics Canada, 1995). Three different measures were created, one for parents, one for teachers, and one for the youth themselves. Ratings in Grade 3 were provided by teachers only, while ratings in Grade 6 were provided by parents, teachers and youth, and ratings in Grade 9 by parents and youth. Items for the parents and teacher versions and the Grade 6 youth version were rated on a three-point scale: 0 = never or not true, 1 = sometimes or somewhat true, and 2 = often or very true (e.g. "vandalizes", "steals", "destroys things", and "tells lies or cheats"). At Grade 9, the youth indicated whether or not in the past 12 months, they were part of a gang (0 = no, 1 = yes) and the remaining nine items were rated 0 = never, 1 = once or twice, 2 = three or four times, or 3 = five or more times (e.g., "stayed out all night without permission", "stolen something," "sold drugs", and "intentionally destroyed/damaged things"). Using principal component factor analyses, delinquency items from teachers, parents, and youth were combined separately at each of the three grades to create Grades 3, 6, and 9 delinquency scales: the Grade 3 delinquency scale was created by combining three teacher rated items; the Grade 6 measure had 13 items (6 parents, 5 teachers, and 2 youth); and the Grade 9 measure of delinquency included 16 items (6 parents and 10 youth). All three scales had high reliability.

Risk and Protective Factors

Risk and protective factors information about children, their families, and neighbourhoods was obtained by parent and child interviews, teacher questionnaires, and Canadian Education Quality and Accountability Office (EQAO) academic achievement test results when the children were in Grade 3. The specific details can be requested directly from the authors.

At the individual child level, we examined children's emotional and behavioural problems (anxiety, depression, hyperactivity, oppositional-defiant, passive victimization, physical aggression), number of serious injuries, social functioning (conflict management, cooperation, outgoing, self-concept, relationship with siblings, number of people important to child), and cognitive and academic functioning (Mathematics Performance on provincial standardized test, Achenbach Academic and Adaptive Functioning, WISC Block Design, Peabody Picture Vocabulary Test score, grade repetition, use of special education services).

At the family level, we examined sociodemographic factors (parent's education level, income, marital status, mobility, teenage parent), family functioning (hostile-ineffective parenting, consistent-effective parenting), substance use (high risk drinking and drug use), and parent's emotional functioning (depression, stress, social support).

At the peer level, we examined how well the child got along with his/her peers. At the school level, we examined parents' perceptions of the school and how involved the parents were at school. Finally, at the neighbourhood level, we asked parents to describe how satisfied they were with their home and neighbourhood, whether they lived in public housing, and how safe they felt from crime.

Grade 9 Outcomes

We examined 41 outcomes when youth were in Grade 9 along several domains obtained by parent and youth interviews, teacher questionnaires, and Canadian Education Quality and Accountability Office (EQAO) academic achievement test results at Grade 9. The specific details can be requested directly from the authors.

To assess youth emotional and behavioural problems, ratings were collected from parents, teachers, and youth. The rating scales include emotional-anxiety disorder, physical aggression, oppositional-defiant, hyperactivity, and depression. To assess youth delinquency, parents completed a "youth trouble" scale, and youth were asked if they were part of a gang and the types of delinquent activities their friends engaged in. Youths' experiences with abuse were also assessed; youth were asked if they had been treated unfairly because of their gender, race, skin color, or religion and if they had been bullied or physically abused. Youth involvement with the criminal justice system was determined through a series of questions in the youth interview (had they ever been arrested, number of arrests, number of close friends arrested, ever been to court, and time in custody or other programs).

Youths' functioning in school was assessed through a series of questions asked of parents, teachers, and youths. Parents were asked if the child had repeated any grades or been suspended. Teachers were asked if the student had been suspended, received special education services, and current academic achievement. Students were asked how often they left/dropped out of school and how often they skipped class. Students' results on the standardized Ontario provincial mathematics achievement test at Grade 9 were also examined.

Finally, youth were asked a series of questions about their health and health risk activities. Specifically, youth were asked about their use of alcohol, tobacco, and illegal drugs, and had they ever been drunk. Youth were also asked to rate their stress level, indicate how often they had been seriously injured, were they sexually active, were they having unprotected sex, and had they ever been pregnant or gotten someone pregnant. Youth and parents were also asked to rate the youth's general health, and youth's body mass index was calculated based on their self-reported height and weight.

Estimated Costs of Government Resources Associated with Delinquency

We identified 12 measures in our data set that could be monetized to reflect children's and parents' utilization of government resources in health care and social services, remedial education, the criminal justice system, and social assistance (see Table 1 for summary). These measures were collected from children and their parents beginning when the children were in Junior Kindergarten (JK) up to and including Grade 9 (more specific details for how each of the 12 outcomes was monetized can be requested from the authors).

Statistical Analyses

For a complete description of the statistical analyses, please see Appendix B.

Results

Trajectories of Delinquency

According to the statistical tests, the six-group solution was the "best" model for the combined sample of girls and boys. Figure 1 depicts the distinct developmental trajectories of the six-class model for delinquency. Children in two of the trajectories had very low ratings of delinquency across time; we labelled these groups the lowest delinquency group and the second lowest delinquency group. Two other trajectories showed a similar pattern of delinquency ratings that was decreasing over time. In the moderate desisters group, children had moderate levels of delinquency at Grade 3 followed by low levels of delinquency at Grades 6 and 9. In the highest desisters group children had the highest level of reported delinquency behaviours at Grade 3, followed by a marked decrease in reported delinquency at Grades 6 and 9. The fifth trajectory group, labelled escalators had very low levels of reported delinquency at Grade 3 and increased markedly in their reported delinquency over time. By Grade 9, children in this trajectory had the second highest delinquency scores. The final group, labelled high delinquency, started with moderate levels of reported delinquency at Grade 3, marked by the highest reported levels of delinquency at Grades 6 and 9 of any of the trajectory groups.

Figure 1. Delinquency Trajectories of At-Risk Youth

Image Description

Figure 1 entitled Trajectories of Delinquency Scale depicts the distinct developmental trajectories of the six-class model for delinquency. The author is using different colors for the six-class model which are: Lowest Delinquency (orange), 2nd Lowest Delinquency (blue), Moderate Desisters (green), Highest Desisters (purple), Escalators (red) and High Delinquency (black). The measurement was established in Grade 3, 6 and 9. Children in two of the trajectories had very low ratings of delinquency across time; they were labeled as the lowest delinquency group (orange) and the second lowest delinquency group (blue). The figure also depicts two other trajectories showing a similar pattern of delinquency ratings that was decreasing over time. In the moderate desisters group (green), children had moderate levels of delinquency at Grade 3 followed by low levels of delinquency at Grades 6 and 9. In the highest desisters group (purple) children had the highest level of reported delinquency behaviours at Grade 3, followed by a marked decrease in reported delinquency at Grades 6 and 9. The fifth trajectory group, labeled escalators (red) had very low levels of reported delinquency at Grade 3 and increased markedly in their reported delinquency over time. By Grade 9, children in this trajectory had the second highest delinquency scores, just below the high delinquency group (black), which started with moderate levels of reported delinquency at Grade 3, but marked by the highest reported levels of delinquency at Grades 6 and 9 of any of the trajectory groups.

Table 2 depicts the percentages of children in each of the groups. Chi-squares tested for gender differences in the group membership of each trajectory group; a significant group by gender effect was found, that is there was a significant difference in the proportion of males compared to the proportion of females (p <.003). We then compared whether the proportion of males versus females differed for each of the six trajectory groups separately. There were significantly more females than males in the two low delinquency trajectory groups, p <.003 for both analyses. There were more males than females in the four remaining trajectory groups, but only the differences for the trajectory groups showing marked decreases in delinquency over time (the moderate and highest desisters) were significant (p <.003 for both analyses).

Grade 3 Risk and Protective Factors by Trajectories of Delinquency

We examined 31 risk factors and 17 protective factors at the individual, family, peer, school, and neighbourhood levels that may influence youth delinquent behaviours. Since Grade 3 is the earliest data point used to determining the trajectory groups, we selected Grade 3 risk and protective factors for this analysis to address whether these factors were associated with the different developmental trajectories of delinquency, and whether these factors were differentially associated for girls and boys.^{Footnote 4}

Of the 31 risk factors, 17 were found to be statistically significant at p <.001 with the full sample: 8 of the 10 individual child risk factors; 5 of the 12 family risk factors; 2 of the 6 school risk factors; the one peer risk factor; and 1 of the 2 neighbourhood risk factors (see Appendix C for more details). By Grade 3, there was evidence that children in the high delinquency, escalators, and the two desisters trajectory groups were experiencing many risk factors at the individual, family, school, and peer levels. For example, compared to the low delinquency groups, children from these four trajectory groups experienced more hyperactive, oppositional-defiant, and physically aggressive behaviours; family risk factors included single parenthood, less parental education, public housing, and hostile-ineffective parenting.

When examining pairwise comparisons for the 17 significant risk factors, the highest desisters group had the most frequent number of significant pairwise comparisons; in other words, this group of children experienced more risk than children in the other 5 trajectory groups. Specifically, they scored higher than the lowest delinquency group on all parent and teacher rating of child behaviour problems, their mothers had a lower education level, and they had poor sibling and peer relationships (see Table 3 for more details). That is, these children were experiencing more individual, family and peer problems.

The high delinquency group and the moderate desisters group also exhibited high levels of risk, especially when compared to the two lowest delinquency groups. For example, the high delinquency group was characterized by both parents and teachers as scoring high on hyperactivity, oppositional defiance, and physical aggression. They were more likely to come from a single family, live in public housing, experience hostile ineffective parenting, and have poor sibling and peer relationships than the lowest delinquency group. The high delinquency group had 11 significant risk factors in Grade 3, the escalator group had six, while the lowest delinquency group had none. Specifically, according to parents, the escalator group scored higher than the lowest delinquency group on hyperactivity, oppositional defiant behaviours and physical aggression. Compared to the lowest delinquency group, they were more likely to have a teenage mother, live in public housing and have poor sibling relationships. Thus, parents had identified these children as experiencing more problems, and they had many family risk factors.

Of the 17 protective factors, 7 were found to be significant at p <.003 with the full sample (see Appendix C), all in the individual child protective domain. The two lowest delinquency groups showed significantly higher levels of social skills (e.g., conflict management, helping/cooperation, outgoing/assertive) and adaptive functioning than children in the two desisters groups (see Table 4). Teachers also rated the high delinquency and escalators groups as showing more conflict management skills than the highest desisters group.

Of the 17 protective factors, 7 were found to be significant at p <.001 with the full sample (see Appendix C), all in the individual child protective domain. The two lowest delinquency groups showed significantly higher levels of social skills (e.g., conflict management, helping/cooperation, outgoing/assertive) and adaptive functioning than children in the two desisters groups (see Table 4). Teachers also rated the high delinquency and escalators groups as showing more conflict management skills than the highest desisters group.

When examining gender differences on the 31 risk and 17 protective factors, 5 risk and 2 protective factors were found to be significant at p <.001 (see Appendix C for full details). Specifically, we found that teachers rated girls as showing fewer hyperactive, depressive, oppositional-defiant, and physically aggressive behaviours. Teachers also rated girls as showing more conflict management and helping/cooperative behaviours. Parents rated girls as showing fewer hyperactive behaviours.

Grade 9 Outcomes by Trajectories of Delinquency

We examined the relationships between Grade 9 outcome variables and delinquency trajectories in a similar manner as we did for the Grade 3 risk and protective variables.^{Footnote 7} However, given the small sample sizes for some trajectory groups, we reclassified the 6 groups of trajectories into 4 groups by combining moderate desisters and highest desisters (and calling it desisters) and by combining the lowest and second lowest trajectories (calling it low delinquency); the other two groups, escalators and high delinquency, remained the same as before. For continuous variables, adjusted group means are reported and for dichotomous variables, odds ratios are reported.

Of the 41 Grade 9 outcomes examined, 31 were found to be significant at p <.003 with the full sample (see Appendix D for more details). To briefly summarize: in the Emotional and Behavioural Problems domain 7 of 10 outcomes were significant; in the Delinquency Problems domain all 3 independent measures of delinquent outcomes were significant; in the Experience of Abuse domain 1 of 3 outcomes was significant; in the Involvement with Criminal Justice System domain all 5 outcomes were significant; in the School Functioning domain 5 of 7 outcomes were significant; and in the Health and Health Risk Behaviours domain 10 of 13 outcomes were significant. These results indicate, as would be expected that by Grade 9, the high delinquency group and the escalators were already exhibiting significantly more problems than the youth in the other trajectory groups in all areas of their functioning (emotional and behavioural problems, criminal involvement, and engaging in unhealthy behaviours).

We then examined differences among our 4 trajectory groups on these 31 significant outcomes (see Table 5 for summary). The escalators and high delinquency groups differed significantly from the desisters and low delinquency groups on 26 of the 31 outcome measures. That is, the escalators and high delinquency groups exhibited more emotional and behavioural problems, engaged in more delinquent behaviours, were more likely to be involved in the criminal justice system, had poorer school functioning, and were more likely to be engaged in health risk behaviours compared to the other two trajectory groups. Some specific results merit emphasis. Compared to the youth in the low delinquency group, the youth in the high delinquency group were 25 times more likely to be part of gang in the past year, 33 times more likely to have ever been arrested, 91 times more likely to have ever gone to court, 13 times more likely to have been suspended from school in the past 3 years, 37 times more likely to have done hard drugs in the past year, and 20 times more likely to have had unprotected sex in their most recent sexual encounter. Additionally, compared to youth in the low delinquency group, the youth in the escalators group were 44 times more likely to be part of gang, 20 times more likely to have been arrested, 37 times more likely to have gone to court, 11 times more likely to have been suspended from school, 26 times more likely to have done hard drugs, and 15 times more likely to have had unprotected sex. The importance of these results is that they are separate indicators of involvement in crime than the items used to create the delinquent trajectories. Thus, using both self report and official data sources, there is converging evidence that these high risk youth are indeed high risk and engaging in high risk behaviours with significant consequences.

Estimated Costs Associated with Delinquency Trajectories

Most studies of juvenile delinquency over the past two decades have focused on older, serious and violent juvenile offenders. Younger delinquents have been ignored partly because their number is relatively small and their threat is not as immediate. However, whereas the number of very young offenders is small compared with older juveniles, child delinquents present unique challenges that need to be addressed. Intervening before minor offences become more serious and before the occasional offender becomes a chronic offender is important. Understanding the trajectories of delinquency at a young age and the risk and protective factors associated with those developmental trajectories can inform the development of early risk assessments and the development of prevention and intervention programs.

For each of the 6 trajectories of delinquency, we estimated an average cost/child/trajectory for each of the 12 monetizable government resources described in Table 1. For each child, we estimated the costs of utilizing the government resource by multiplying the unit cost available from a secondary source (e.g., $29.44 for an appointment with a family physician) by the occurrence of the event. All dollar figures that we report were discounted at a rate of 3 %. This discount rate falls within the range of rates commonly used and recommended in public-policy analysis (e.g., Karoly et al., 1998; Karoly et al., 2005; Reynolds et al., 2002). For a complete description of the statistical analyses, please see Appendix B.

Detailed results for each of the 12 indicators of government resource utilization by trajectory group by gender can be found in Appendix E. Table 6 provides a summary of the government expenditures by general domain by trajectory group. To briefly summarize the results, government expenditures were highest in the Remedial Education domain (64% of costs), followed by Health Care and Social Services (29%), Social Assistance (6%), and Criminal Justice System (1%). The two lowest delinquency trajectories (82% of the sample) accounted for only 19.4% of the estimated government costs. In other words, approximately 80% of the estimated costs to government were from 18% of the sample. Specifically, we found that youth from the two desisters trajectory groups (13% of the sample) accounted for 40% of the estimated costs to government; and youth from the two most at-risk trajectories (escalators and high delinquency, 5% of the sample) accounted for 40.6% of the estimated costs to government. It is interesting to note that 80% of the estimated Criminal Justice costs were due to the high delinquency and escalators trajectory groups.

We also found that antisocial or delinquent girls cost society more money than antisocial or delinquent boys in all domains, with the exception of the Social Assistance domain. Specifically, summing across all 6 trajectory groups from ages 4 to 14, we estimated that girls cost $244,056 while boys cost $229,236. In addition, we estimated that girls' criminal justice costs were almost twice those of boys ($4,835 vs. $2,408).

Discussion

Given that over one-third of youth have been involved in some form of delinquency by the age of fourteen, and that childhood delinquency tends to predict violent behaviours throughout the course of a lifetime (Farrington, 1989), understanding the developmental pathways that lead to delinquency is a critical issue. The current study was designed to identify the delinquency trajectories of boys and girls living in disadvantaged communities in Ontario from ages 8 to 14, and examine the risk/protective factors, Grade 9 outcomes, as well as the estimated economic costs associated with each trajectory. Results indicated that children in the escalator group and the high delinquency group had significant negative outcomes by Grade 9 with respect to their behavioural, emotional, social, and risk-taking behaviour (e.g., drug use, unprotected sex), as well as high involvement in the criminal justice system. These problems also were costly to the government.

Developmental Trajectories of Delinquency

Our first objective was to examine the trajectories of delinquency in boys and girls from ages 8 (Grade 3) to 14 (Grade 9). Our results confirm the heterogeneity of the development of delinquency and are generally consistent with previous research. We found six groups of delinquency. As expected, two groups, lowest delinquency and second lowest delinquency, representing the majority of the youth (≈82% of the sample) reported consistently low levels of delinquency over time. Two other trajectories (highest desisters and moderate desisters) showed a similar pattern of delinquency ratings decreasing over time, representing the desisters (≈13% of the sample). Another group, the escalators (≈3.5% of the sample), had very low levels of reported delinquency at Grade 3 and increased over time. Finally, the high delinquency group started with moderate levels of reported delinquency at Grade 3 and had the highest reported levels of delinquency at Grades 6 and 9 of any of the trajectory groups. The high delinquency group represented approximately 1.5% of the sample. It may be that the low percentage of youth in the high delinquency group reflects the fact that we only have data up until the youth are in Grade 9, or approximately 14 years old. Thus, many youth may just be beginning to engage in delinquent acts. We hypothesize that with more longitudinal data points, the proportion of youth in the high delinquency group would increase and likely more closely resemble other research findings.

This study supported the trajectories of delinquency reported in other studies, but also identified some key differences. Similarities included: 1) that the majority of youth were involved in no or limited delinquent activities; 2) females were more likely than males to be uninvolved in delinquency (i.e., there were more females in the low delinquency and second lowest delinquency trajectory groups); 3) there was a group of individuals who desisted from involvement in delinquency; and 4) there was a trajectory of consistently high engagement in delinquent behaviour. The key differences from previous literature was the number of groups that had low levels of delinquency (i.e., there were two low and second lowest groups that engaged in minimal delinquent behaviours). Second, the shape of the high delinquency trajectory group was surprising, as there was a peak in delinquency in Grade 6. We expected that the peak would not be present, and if we had extended longitudinal data we would have expected to see it at around age 18. There are several possible interpretations to this early peak. First, no other study on delinquent trajectories has been conducted starting at such a young age. Second, the current study included girls which no other study of delinquent trajectories has done. Third, this study was based on community sampling, that is it was conducted in high risk, low socioeconomic status neighbourhoods. Lastly, it is possible that there are unique sample characteristics in the participants and the results may reflect this sampling. Nonetheless, more longitudinal research is required that begins as early as this research to validate findings.

Third, when we examined differences in the distributions of boys and girls within the diverse trajectory groups, we found that the escalators and high delinquency groups had equal proportional representativeness of males and females. That is, we found no gender differences in the distribution of boys and girls in the high delinquency group (2% of males and 1% of females), or in the escalators group (4% of males and 3% of females). Typically research reports that males are more likely to engage in delinquent behaviour than females, thus we expected to have more males than females in the high delinquency group. Notably, this pattern is inconsistent with the general developmental trend reported by Silverthorn and Frick (1999) who found that girls tend to experience a later onset of delinquency than boys, and the general finding that boys are more likely to be involved in high delinquent behaviour than girls. The discrepancy may arise because we have used a multi-informant approach, and have taken a person-oriented approach (as opposed to a group oriented approach), allowing us to examine heterogeneity within the development of delinquency. The small minority of at-risk girls in our sample demonstrated these problems as early as boys. Consistent with other research, we found that girls were overrepresented in the two low delinquency groups. However, we found there were significantly more males in the two desisters groups.

Risk and Protective Factors Associated with Delinquency Trajectories

Trajectories increase our understanding of delinquency development and identify behavioural patterns that emerge in individuals on a specific trajectory. Once these trajectories are identified, specific factors pertaining to the individual, peers, family, and community in general can be explored to determine which factors heighten the risk of delinquency (i.e., the chronic or increasing trajectories) or act as a protective factor against the involvement in delinquency (i.e., low, non-involved, or declining trajectories).

In this research we examined 31 risk factors and 17 protective factors at the individual, family, peer, school, and community level when the children were in Grade 3 (age 8) that may influence the likelihood that youth will engage in criminal behaviour in adolescence. Children at risk for delinquency (i.e., those in the high delinquency, escalators, and desisters trajectory groups) scored significantly higher on 17 of the 31 individual, family, peer, and neighbourhood risk factors. For example, children from these four trajectory groups experienced more hyperactive, oppositional-defiant, and physically aggressive behaviours; family risk factors included single parenthood, less parental education, public housing, and hostile-ineffective parenting. The most at-risk groups were experiencing problems in multiple domains, noted by multiple informants and assessments, yet they received limited interventions or support to address these problems. Thus, with comprehensive early assessments, early identification of at-risk children can occur early at school allowing the provision of extra services to prevent continuation of problematic and costly behaviours through adolescence.

More specifically, youth assigned to the high delinquency group were already showing signs of problems in Grade 3. Parents and teachers rated them as higher than the low delinquency groups on hyperactivity, oppositional behaviour, and physical aggression. In addition, they were more likely to come from single parent homes, live in public housing and experience higher levels of hostile ineffective parenting and had poor quality peer and sibling relationships compared to the two low delinquency groups. Interestingly only the moderate desisters were viewed as more problematic with respect to their oppositional defiant behaviour and their physical aggression than the high delinquency group, according to teachers. With respect to school functioning, there were no differences on the PPVT test or on the likelihood that they received special educational services compared to the other groups. In fact this group had the lowest special education rates yet they had the lowest PPVT scores (although not significantly different from the other groups); they may not have been receiving the special services at school that they required.

The high delinquency group also did not score well on protective factors. They had significantly lower scores on conflict management and cooperative behaviours than the low delinquency and desisters groups, according to parents. This composition of risk and protective factors indicates that parents identified many behavioural and social problems in children in the high delinquency group. Interestingly, teachers rated the high delinquency group as showing more conflict management skills than the highest desisters group. This group was not viewed as the most problematic with respect to classroom behaviours, which may have minimized the extent of their problematic behaviours and limited the potential interventions they could have received.

Similarly, the youth in the escalators group were rated by their parents as showing the second worst problematic behaviours and had many family risk factors, such as more likely to live in a single parent home, lived in public housing, and had poor peer relationships. For both the high delinquency and the escalators groups, the issues at home may have played a role in their delinquent trajectories. Research has indicated that single parents may be less able to monitor their children than children living in two parent homes (Tremblay, Van Aken, & Koops, 2009). Similarly, they lived in social housing where there was a lack of monitoring and where they may have been more likely to associate with peers with similar problems, thus providing a peer group with similar problems to reinforce their aggressive and delinquent behaviour problems. Teachers did not perceive this escalators group as exhibiting many problematic behaviours compared to the other groups. This lack of concordance between parents and teachers may have contributed to them not being identified as having problems. It may be that the children were having fewer behavioural problems at school than at home, or it may be that the behavioural problems at school were not as extreme as those experienced at home. Furthermore, this discrepant finding between parent and teacher ratings may reflect the lack of services put into place to promote healthy behavioural and school functioning for these children, which may have inadvertently contributed to their ongoing problems. In any case, the disagreement between parents and teachers highlights the need to take parents' views into account in developing assessment and/or screening tools. Furthermore, these parents may need more services to help them address problematic behaviours early, at home. The combination of many risk and few protective factors, and little support in terms of educational assistance may have interacted and accumulated to maintain and increase their risk for delinquency over time.

The two desisters groups (highest desisters and moderate desisters) are an interesting contrast to the high delinquency and escalators groups. The desisters groups were viewed the most negatively by their teachers in Grade 3 (i.e., they had the highest score on all risk factors and the lowest overall protective scores as viewed by their teachers). Furthermore, the risk factors included both externalizing problems and internalizing problems. It may be in part this combination of depression, victimization, and externalizing problems that contributed to them being identified as such by their teachers. Parents also identified externalizing problems in these groups. Furthermore, the moderate desisters also have elevated risk with respect to their family environments (e.g., more likely to live in single parent homes, have a teenage mother, and a mother with lower education than the low groups). It may be that these families were receiving more social assistance due to their life circumstances than the high delinquency and escalators groups, which may have also been protective against future delinquency involvement. Furthermore, likely as a consequence of being identified by teachers as experiencing many behavioural problems and academic problems, they received the most special education services (43%) at school. These services may have acted as an effective early intervention for these students by promoting positive school functioning that in turn facilitated their desistance from delinquency and associated problematic behaviour.

There are several implications to these findings. For example, it supports the notion that developing an assessment/screening tool for risk measuring psychological, emotional, and behavioural functioning, as well as family and school functioning, can provide early identification of children who are at different levels of risk for future delinquency. In addition, providing interventions or strategies (such as special educational services) to those who are identified at-risk can prevent delinquency in the future. Among other things, the current research suggests that early investment in school services can make a measurable difference in delinquency trajectories by Grade 9. Without investment, the problematic and costly behaviours of at-risk youth are likely to continue through adolescence and potentially become more significant.

Grade 9 Outcomes Associated with Delinquency Trajectories

The third objective of this study was to examine whether youth in the identified delinquency trajectories differ substantially in Grade 9 on emotional and behavioural problems (e.g., emotional-anxiety disorder, depression, aggression, oppositional-defiant, hyperactivity-inattention), delinquency (e.g., association with delinquent friends, being part of a gang), experience of abuse (e.g., physical abuse, bullying, discrimination), involvement with the criminal justice system (e.g., arrests, court appearances, time spent in custody), academic/school functioning (e.g., achievement, use of special education services, student suspensions, grade repetition), and health/health risk behaviours (e.g., use of alcohol, tobacco, and illegal drugs, injuries, unprotected sexual activity, pregnancy). We found that early problems (i.e., emotional and behavioural, delinquency, academic) become even more significant by Grade 9. Our two most at-risk groups, the high delinquency and the escalators groups, had significantly more problems in all areas of functioning. They scored the highest on the majority of the emotional/behavioural (e.g., anxiety, hyperactivity, physical aggression); health (e.g., general health, use of tobacco/alcohol/drugs, sexual activity); criminal (e.g., arrests, court appearance, in custody), and school functioning (e.g., suspensions, special education, dropping out of school) domains.

An examination of some of the specific outcomes in Grade 9 for our two most at-risk groups (the high delinquency and the escalators) highlights that the pathway to delinquency is developmental and that early behaviours are indicative of significant problems by Grade 9. For example, even by Grade 9 these high risk groups were much more likely to be involved in gangs, to have been arrested, and to have a criminal record than the other groups. Furthermore, the escalators and high delinquency groups engaged in more risky health behaviours (e.g., consumption of hard drugs and involvement in unprotected sex behaviours). These behaviours are problematic not only in themselves but in their consequences (e.g., early pregnancy with potentially substance using parents).

Furthermore, the youth in the high delinquency and escalators groups were experiencing significant truancy, thus further limiting their long term employment and educational opportunities. In all of the domains examined, these at-risk youth were experiencing problems in Grade 9 that were much more severe than in Grade 3 and had much potentially significant longer term outcomes.

Estimated Economic Costs Associated with Delinquency Trajectories

The final objective of the present study was to estimate the costs to government associated with each delinquency trajectory on utilization of government resources in the criminal justice system, remedial education, health care and social services, and social assistance. The majority of the estimated costs associated with each of the trajectories was in the educational system - 64% of the costs were for remedial education. In contrast, the percentage of the estimated costs associated with the other domains was 29% for health care and social services, 6% for social assistance, and 1% for the criminal justice system.

As noted earlier, it was the desisters groups (highest desisters and moderate desisters) who received the most special educational services, and with respect to long term outcomes, this was a positive and preventative investment. A review of the specific estimated health care costs indicates that the escalators in particular had the highest costs associated with visiting their doctor, going to the emergency room, having serious injuries, and visiting with a nurse practitioner. These are reactionary costs (as opposed to preventative costs) in the sense that a significant event has happened. Furthermore, for girls in the high delinquency group, some costs were estimated as being much higher than for boys (e.g., number of serious injuries, and overnight stay in hospital. At-risk girls may be particularly vulnerable to medical problems associated with delinquency involvement compared to at-risk boys.

High risk girls were also more costly with respect to the criminal justice system. Compared to boys, the total estimated costs at age 14 (Grade 9) for girls was almost twice that for boys ($4,835 vs. $2,408). The data revealed that girls in the two high risk groups (high delinquency and escalators) were much more likely to have higher costs associated with each being arrested and court appearances. It appears that girls, once arrested, were also much more likely to enter the criminal justice system. Admittedly, our sample of girls was small and may not be representative, but it does reflect the developmental course and costs associated with a small sample of very high risk delinquent girls. The high risk boy groups also had the highest estimated costs, but not as high as those of the high risk girls. In summary, our findings suggest that girls cost the government more money than boys in all domains (except social assistance). Specifically, summing across the six trajectory groups, we estimated that, between the ages of 4 and 14, girls cost $244,056 while boys cost $229,236.

Furthermore, approximately 80% of the estimated costs to government were due to the two desisters trajectory groups (highest desisters and moderate desisters) and the youth from the two most at-risk trajectories (escalators and high delinquency) which represented 18% of the sample. Specifically, we found that youth from the two desisters trajectory groups (13% of the sample) accounted for 40% of the estimated costs to government (primarily driven by education costs, a preventative response); and youth from the two most at-risk trajectories (escalators and high delinquency; 5% of the sample) accounted for 40.6% of the estimated costs to government.

Additionally, 80% of the estimated criminal justice costs were due to the high delinquency and escalators groups. Even though the estimated Criminal Justice System costs to government were relatively low as of Grade 9 (only 1% of the overall costs), these two groups may just be getting started and the costs associated with these groups can only increase. Interestingly, the high delinquency and escalators groups accounted for 46% of the reactive costs (such as criminal justice system, health care and social services) compared to 32% for the two desisters groups and 22% for the two low delinquency groups; for the preventative costs (e.g., remedial education), the high delinquency and escalator groups accounted for 38% of the costs compared to 44% for the two desisters groups and 18% for the two low delinquency groups. The implication is that investing early in prevention costs such as remedial education may provide at-risk children and their families the opportunity to have more positive developmental outcomes and desist from delinquency involvement. As a consequence, investing in prevention can save the government money in the long run. The most at-risk groups did not receive sufficient early support and consequently the costs associated with them were reactive and costly.

Limitations

There are many strengths to the current research. The BBBF research sample comprised disadvantaged and at-risk communities; the communities were diverse (Francophone, Aboriginal, recent immigrants, and multicultural); the sample had both boys and girls; and the data allowed for economic analyses to be conducted. This is the first on a Canadian sample. Having said that, some limitations need to be noted. First, we were unable to examine the risk and outcome factors by trajectory for both boys and girls separately due to the low number of females in some of the trajectory groups. Second, some of the trajectories had a small sample size and hence the results may not be generalizable. For example, in the high delinquency group, the costs of delinquent behaviour in girls were high relative to boys. It may be that this is an atypical group that had many arrests, or in fact, it may be representative of an extreme group of high risk girls that to date have been neglected by research.

Conclusion

There are early indicators to the developmental pathways for delinquency. Risk and protective factors associated with more serious and escalating delinquency involvement become apparent as early as Grade 3, which could inform the implementation of an assessment/screening tool. Furthermore, the current research findings suggest that delinquency involvement does not just emerge, but develops over time, and without intervention, the problems accumulate and are serious and significant by as early as Grade 9. The increased likelihood of arrests, court appearances, and incarceration by Grade 9 for the high delinquency and escalator groups, indicate that the delinquent problems are significant and serious. Similarly, investment in prevention, such as educational support, can reduce delinquency involvement. The most at-risk groups for delinquency involvement (e.g., escalators and high delinquency) accounted for the majority of the reactive costs (e.g., criminal justice) and not the preventative costs (e.g., remedial education).

The present study also demonstrates that, although more research is needed to understand the developmental delinquency trajectories of girls, they appear to require more support than boys. Although our high risk sample of girls was limited, there are some preliminary indications from this research that they are at a heightened risk for problems, such as emotional problems, criminal activity and court system involvement, and the costs associated with their problems may be higher than for boys because they appear not only in the criminal justice system, but also in the health care system. Traditionally, we have estimated only the criminal justice costs. It may be that this venue does not reflect the full range of costs associated with female delinquency.

In summary, different developmental periods may have different risks and protective factors associated with delinquency. Thus, crime prevention needs to occur early in development and be ongoing. Our study indicates that there were more problematic externalizing behaviours in our high delinquency and escalators groups by Grade 3, as indicated by parents in particular and somewhat supported by teachers. Despite the problematic behaviours as reported by parents, teachers did not view them as displaying the most problematic externalizing behaviours in the class; instead, teachers rated the highest desisters group as having more oppositional-defiant and physical aggression problems than the high delinquency and escalator groups. This lack of identification may be one reason they did not receive extra support early. It may be that having problems identified early by others outside the family facilitate the identification and early intervention for children at risk for later serious delinquency. In addition to the behavioural problems, the family lives of the escalators and high delinquency group were also problematic. These children may have lacked opportunities to interact positively with other children and adults. They were living in homes characterized with higher levels of hostile and ineffective parenting and had poor peer and sibling relationships. They may have lacked a positive and supportive adult in their lives to champion them, model and reinforce positive behaviours and social relationships. Lastly, they likely lived in high risk neighbourhoods characterized by social housing and low socio-economic status that may have contributed to their delinquent trajectories. Furthermore, in these neighbourhoods, they may have had greater access to peers experiencing similar problems (as indicated by their associations with friends who were more likely to be delinquent and be arrested). Thus, there may be delinquency influence occurring within their peer groups. Therefore, crime prevention approaches need to target high risk families, living in high risk neighbourhoods, and provide family, school, and community support. This support needs to be ongoing to ensure that the behavioural problems demonstrated early in Grade 3 does not escalate and accumulate into serious delinquency and drug abuse by Grade 9.

Although we have made a great deal of progress in understanding individual differences in antisocial behaviour and linking these to interventions, much work remains to be done. Research that continues to monitor the development of these trajectories could be informative as youth transition into early adulthood. The mental and physical health and other needs of children at-risk for delinquency involvement should not be ignored. An examination of the youth who desist from delinquency provide strong support for the value of investing early in children to prevent negative long term outcomes. Even modestly successful prevention and intervention investments, such as in education, yielded significant benefits, including decreasing future expenditure associated with delinquency, improving well-being and safety of families, children, and youth in a community, and reducing crime and delinquency.

References

Andrews, D. A., & J. Bonta. 1998. The Psychology of Criminal Conduct (2nd ed.). Cincinnati, OH: Anderson.

Angus, D. E., J. E. Cloutier, T. Albert, D. Chenard, A. Shariatmadar, W. Pickett, et al. 1998. The Economic Burden of Unintentional Injury in Canada. Toronto, ON: SmartRisk Foundation.

Barnett, W. S, & L. N. Masse. 2007. "Comparative benefit-cost analysis of the Abecedarian program and its policy implications". Economics of Education Review, 26:113-125.

Boyd, J.W., W. S. Barnett, E. Bodrova, D. J. Leong, D. Gomby, K. B. Robin, & J. T. Hustedt. 2005. Promoting Children's Social and Emotional Development through Preschool. New Brunswick, NJ: NIEER.

Browne, G., A. Gafni, & J. Roberts. 2002. Approach to the Measurement of Costs (expenditures) when Evaluating Health and Social Programs. (Working Paper Series 01-03). Hamilton, Ontario: McMaster University, System-Linked Research Unit on Health and Social Service Utilization.

Cohen, M.A. 1998. "The monetary value of saving a high risk youth". Journal of Quantitative Criminology, 14:5-33.

Cohen, M.A., & A. R. Piquero. 2009. "New evidence on the monetary value of saving a high risk youth". Journal of Quantitative Criminology, 25:25-49.

Farrington, D. P. 1989. "Early predictors of adolescent aggression and adult violence". Violence & Victims, 4(2): 79-100.

Hepworth, P. 2000. "Jack's troubled career: The costs to society of a young person in trouble". Prevention Newsletter, 2:10-11. Ottawa: Justice Canada, National Crime Prevention Centre.

Hoeve, M., A. Blokland, J. S. Dubas, R. Loeber, J. Gerris, & P. H. Van Der Laan. 2008. "Trajectories of delinquency and parenting styles". Journal of Abnormal Child Psychology, 36(2):223-235.

Jones, B., D. S. Nagin, & K. Roeder. 2001. "A SAS procedure based on mixture models for estimating developmental trajectories". Sociological Methods and Research, 29:374-393.

Karoly, L., P. Greenwood, S. Everingham, J. Houbé, M. Kilburn, C. Rydell, et al. 1998. Investing in Our Children: What We Know and Don't Know about the Costs and Benefits of Early Childhood Interventions. Santa Monica, CA: RAND Corporation.

Karoly, L. A., M. R. Kilburn, & J. S. Cannon. 2005. Early Childhood Interventions: Proven Results, Future Promises. Santa Monica, CA: RAND Corporation.

Kass, R. E., & A. E. Raftery. 1995. "Bayes factor". Journal of the American Statistical Association, 90:773-795.

Lacourse, E., S. Côté, D. S. Nagin, F. Vitaro, M. Brendgen, & R. E. Tremblay. 2002. "A longitudinal-experimental approach to testing theories of antisocial behaviour development". Development and Psychopathology, 14:909-924.

Lacourse, E., D. Nagin, R. E. Tremblay, F. Vitaro, & M. Claes. 2003. "Developmental trajectories of boys' delinquent group membership and facilitation of violent behaviours during adolescence". Development and Psychopathology, 15:183-197.

Lerner, R. M. 1996. "Relative plasticity, integration, temporality, and diversity in human development: A developmental contextual perspective about theory, process, and method". Developmental Psychology, 32:781-786.

Moffitt, T.E. 2001. "Childhood predictors differentiate life-course persistent and adolescence-limited antisocial pathways among males and females". Development and Psychopathology, 13:355-375.

Moffitt, T.E., A. Caspi, M. Rutter, & P. A. Silva. 2001. Sex Differences in Antisocial Behaviour: Conduct Disorder, Delinquency, and Violence in the Dunedin Longitudinal Study. Cambridge: Cambridge University Press.

Mrazek, P. J., & C. H. Brown. 2002. "An evidenced-based literature review regarding outcomes in psychosocial prevention and early prevention in young children". In C. C. Russell (Ed.), The State of Knowledge about Prevention/Early Intervention (pp. 42-144). Toronto, ON: Invest in Kids Foundation.

Nagin, D.S. 1999. "Analyzing developmental trajectories: a semi-parametric, group-based approach". Psychological Methods, 4:139-157.

Nagin, D.S. 2005. Group-Based Modeling of Development. Landon: Harvard University Press.

National Council on Welfare. 2004. Welfare Incomes 2003. Ottawa: NCW.

Nores, M., C. R. Belfield, W. S. Barnett & L. Schweinhart. 2005. "Updating the economic impacts of the High/Scope Perry Preschool Program". Educational Evaluation and Policy Analysis, 27(3):245-261.

Odgers, C.L., T. E. Moffitt, J. M. Broadbent, N. Dickson, R. J. Hancox, H. Harrington, R. Poulton, M. R. Sears, W. M. Thompson, & A. Caspi. 2008. "Female and male antisocial trajectories: From childhood origins to adult outcomes". Development and Psychopathology, 20: 673-716.

Ontario Ministry of Community and Social Services. 2003. Ontario Disability Support Program. Retrieved January 25, 2008.

Peters, R. DeV., A. J. Bradshaw, K. Petrunka, G. Nelson, Y. Herry, W. M. Craig, et al. 2010. The ‘Better Beginnings, Better Futures' Ecological, Community-Based Early Childhood Prevention Project: Findings from Grade 3 to Grade 9. Manuscript submitted for publication.

Peters, R. DeV., K. Petrunka, & R. Arnold. 2003. "The Better Beginnings, Better Futures Project: A universal, comprehensive, community-based prevention approach for primary school children and their families". Journal of Clinical Child and Adolescent Psychology, 32, 215-227.

Reynolds, A. J., J. A. Temple, D. L. Robertson, & E. A. Mann. 2002. "Age 21 cost-benefit analysis of the Title I Chicago Child-Parent Centers". Educational Evaluation and Policy Analysis, 24(4), 267-303.

Rutter, M. 1986. "Child psychiatry: The interface between clinical and developmental research". Psychological Medicine, 16, 151-169.

Savoie, J. 2006. "Youth self-reported delinquency". Toronto. Juristat, 27(6). Ottawa: Statistics Canada.

Schonberg, M.A. & D. S. ShawS. 2007. "Risk factors for boy's conduct problems in poor and lower-middle-class neighborhoods". Journal of Abnormal Child Psychology, 35, 759-772.

Serbin, L.A., D. M. Stack, N. De Genna, N. Grunzeweig, C.E. Temcheff, A.E. Schwartzman, & J. Ledingham. 2004. "When aggressive girls become mothers: Problems in parenting, health, and development across two generations". In M. Putallaz & K. Bierman (Eds.), Aggression, antisocial behavior, and violence among girls: Duke series in child development and public policy (pp. 262-285). New York, NY:Guilford Press.

Silverthorn, P. & P. J. Frick. 1999. "Developmental pathways to antisocial behavior: The delayed-onset pathway in girls". Development and Psychopathology, 11, 101-126

Statistics Canada. 1995. National Longitudinal Survey of Children and Youth: Overview of survey instruments from 1994-95 data collection cycle I. Statistics Canada Catalogue no. 89F0077XIE. Ottawa, ON: Statistics Canada. Retrieved December 4, 2008

Tremblay, R. E., M. A. G. Van Aken, & W. Koops. (Eds.) 2009. Development and Prevention of Behaviour Problems: From Genes to Social policy. New York, NY : Psychology Press. 280 p.

Waddell, C., J. M. Hua, O. M. Garland, R. D. Peters, & K. McEwan. 2007. "Preventing mental disorders in children: A systematic review to inform policy-making". Canadian Journal of Public Health, 98, 166-173.

Wiesner, M., &M. Windle. 2006. "Young adult substance use and depression as a consequence of delinquency trajectories during middle adolescence". Journal of Research on Adolescence, 16, 239-64.

Wolke, D., S. Woods, L. Bloomfield, & L. Karstadt. 2000. "The association between direct and relational bullying behaviour problems among primary school children". Journal of Child Psychology and Psychiatry, 41:989-1002.

Appendix

Appendix B. Description of statistical analyses

To identify the trajectories of delinquency we used the semi-parametric group-based trajectory approach (Jones et al., 2001; Nagin, 1999; Nagin, 2005). In this modeling, the dependent variable was the total standardized delinquency scale score at Grades 3, 6, and 9. The censored normal distribution was used to model the trajectories to account for the censoring at the lower and upper bounds of the delinquency scale. A polynomial relationship was used to link age to delinquency behaviour. We compared models with different numbers of groups using a Bayesian information criterion (BIC) (Kass & Raftery, 1995). A large BIC value corresponds to a good model with a large log-likelihood value and not too many parameters. We tested competing models of 2, 3, 4, 5, and 6 groups of delinquency to determine the "best" model based on BIC criterion; we found that the BIC values for 2-, 3-, 4-, 5-, and 6-group models were, respectively, -886.8, -881.2, -894.4, -851.8, and -838.2. Application of the maximum BIC for model selection indicated that the six-group solution was the "best" model for the combined sample of girls and boys.

To examine trajectory group differences on the risk and protective factors, as well as the outcomes, we employed a combination of analysis of variance (ANOVA) and logistic regression depending on the type of outcome variable in question (i.e., ANOVA for continuous variables, and logistic regression for binary variables) to compare the means or proportions of variance of the variable. Omnibus F or chi-square tests were reported to indicate the significance of overall relationship, and Bonferroni tests were carried out to examine pairwise comparisons.

To estimate costs associated with each trajectory of delinquency, we estimated an average cost/ child/ trajectory for each of the 12 monetizable government resources described in Table 1. For each child, we estimated the costs of utilizing the government resource by multiplying the unit cost available from a secondary source (e.g., $29.44 for an appointment with a family physician) by the occurrence of the event. All dollar figures that we report were discounted at a rate of 3 %. This discount rate falls within the range of rates commonly used and recommended in public-policy analysis (e.g., Karoly et al., 1998; Karoly et al., 2005; Reynolds et al., 2002). All missing values, including the values of the missing grades (such as Grades 4, 5, 7 and 8 when no data collection took place), were interpolated, given that there were at least 60% data points present. Each grade specific cost figure was then combined and reclassified into three major groups, JK to Grade 3 (ages 4 to 8), Grade 4 to Grade 6 (ages 9 to 11), and Grade 7 to Grade 9 (ages 12 to 14), and presented by delinquency group trajectories and child's gender. We used the following equation to estimate the average cost for each of the 12 measures of utilization of government resources for each grade. The cost values are based on the value (v) of each outcome as outlined in Table 1 (e.g., $29.44 for a visit to a family physician), multiplied by frequency of occurrence (o) of that outcome for each child for that year.

Image Description

This picture depicts an equation used by author to estimate the average cost for each of the 12 measures of utilization of government resources for each grade.

Mathematically speaking, the formula is written as VO (in capitals) equals the Greek symbol sigma (meaning “sum”) of the compound v (in lower-case, with a subscript lower-case i) multiplied by o (in lower-case, with a subscript lower-case i) divided by n (in lower-case). Above the sigma symbol is the lower case letter n, and below it is the equation i (in lower-case) equals one.

The definitions are as follows:

VO = Average cost for an outcome measure in a grade;

i = number of children (1, ..., n);

n = sample size;

v = value of outcomes ($);

o = occurrences of the outcome.

Conceptually, the equation means that the average cost for an outcome measure in a grade is equal to the sum of the value of outcomes for each child (in dollars) multiplied by each child's occurrences of the outcome, over the sample size, from the first child to the last child in the sample.

The cost values are based on the value (identified by a Capital V) of each outcome as outlined in Table 1 (e.g., $29.44 for a visit to a family physician), multiplied by frequency of occurrence (identified by a Capital O) of that outcome for each child for that year.

Appendix C

Appendix D

Appendix E

Date modified:

2022-08-09