Early longitudinal and birth cohort studies of delinquency discovered that not all delinquents are the same. Importantly, in the early 1970s, social researchers identified a group of “chronic delinquents,” a small group of offenders who accounted for the majority of recorded offenses. The existence of different groups, based on offending patterns, stimulated a great deal of research and debate. Unfortunately, the theoretical and practical implications of the chronic delinquent (also called the serious habitual offender or super-predator) could not be fully explored with the analytic methods of the time. Although powerful statistical techniques were available to focus on individual patterns of offending, it was not until recently that statistical techniques became available to aggregate individual patterns of offending into group-based patterns of offending. These group-based patterns of offending search for clusters or groups of individuals who share common histories of delinquent behavior. These common histories of delinquent behavior can then be displayed graphically with developmental trajectories, or lines of development that show offending rates over time. These graphical displays have confirmed the existence of different groups of offenders, each with unique histories of delinquent behavior.
Analytically, these methods recognize that delinquent behavior does not begin and evolve in the same manner for all offenders (e.g., some start early, others start late). Simultaneously, these methods recognize that some similarities in delinquent development do exist (e.g., those who start early all tend to subsequently offend at a high rate). Using these methods, we can empirically categorize offenders into groups. Each group has its own history of delinquent behavior or developmental trajectory. Within each developmental trajectory are offenders who share a similar pattern of offending. As Bobby L. Jones and Daniel S. Nagin recently argued in their 2007 Sociological Methods & Research, “charting and understanding developmental trajectories is among the most fundamental and empirically important research topics in the social and behavioral sciences” (pp. 542–543).
The importance of this research topic is confirmed by recent advances in theoretical criminology. In particular, recent developmental theories strongly support typologies of offending. Typological theories classify offenders into different groups (just as group-based modeling does) with each group having its own history of delinquent behavior. For example, some theories differentiate between life course persistent offenders who start offending early, offend at a high rate, and persist through the life course, and adolescent limited offenders who start offending late, offend at a low rate, and desist by the time they emerge into adulthood. Some theories further suggest that the causes of offending vary across groups. Peer pressure, for example, may be more relevant for adolescent limited offenders than for life course persistent offenders. Group-based modeling now provides the opportunity to fully explore these ideas and their implications on theory and practice.
As part of our research on disproportionate minority contact in the Alaska juvenile justice system, we recently utilized group-based modeling to further examine how juvenile offending patterns from age 10 to 17 vary by race. Previous Justice Center research clearly showed that minority youth were overrepresented in referrals to the Alaska Division of Juvenile Justice (DJJ). For example, in fiscal year 2005, 61 percent of youth referred to DJJ in Anchorage were minority youth, but only 34 percent of the youth population in Anchorage was minority. In Anchorage, the rate of referral to DJJ was 3.0 times higher for minority youth than for White youth. Similarly, in fiscal years 2005 and 2006, 47 percent of the youth referred to DJJ in Fairbanks were minority youth, but only 27 percent of the youth population in Fairbanks was minority. In Fairbanks, the rate of referral to DJJ was 2.5 times higher for minority youth than for White youth. To better understand the impact of race, we developed group-based models of offending for youth referred to DJJ in Anchorage and Fairbanks. To do so, we examined the offending histories of two separate cohorts. The first included any youth who had been referred to DJJ in Anchorage who was born in 1989. The second included any youth who had been referred to DJJ in Fairbanks who was born in 1988 or 1989. In this article, we present descriptive information on these two cohorts, present their developmental trajectories, and assess the extent to which the development of delinquent behavior varied by race.
The Anchorage cohort was built by selecting any youth who had ever been referred to DJJ in Anchorage and who was born in 1989. This cohort included 1,131 non-duplicated juveniles. The Fairbanks cohort was similarly built. It included 620 non-duplicated juveniles who were born in 1988 or 1989 and who were referred to DJJ in Fairbanks. For each juvenile in each cohort, we then measured the number of charges referred to DJJ at each age, from age 10 to 17. The 1,131 youth in the Anchorage cohort produced a total of 4,074 charges, for an average of 3.6 charges per youth. The number of charges for each youth ranged from one to 38. The 620 youth in the Fairbanks cohort produced a total of 2,624 charges, for an average of 4.2 charges per youth. The number of charges in the Fairbanks cohort ranged from one to 39. Additional details are shown in Table 1. It is important to note that 51 percent of the youth in the Anchorage cohort and 44 percent of the youth in the Fairbanks cohort had only one charge referred to DJJ from age 10 to 17. Only 10 percent of the youth in the Anchorage cohort were referred to DJJ for ten or more charges, but these youth accounted for 47 percent of the total number of charges. Similarly, only 12 percent of the youth in the Fairbanks cohort were referred to DJJ for ten or more charges, but these youth were responsible for 50 percent of the total number of charges. In both cohorts, a small group of youth (10 to 12%) was responsible for approximately half of the total number of referred charges. Table 2 shows the age at which the first charge was referred to DJJ. Over half (56%) of the youth in the Anchorage cohort and half (50%) of the youth in the Fairbanks cohort were referred to DJJ before age 15.
A greater percentage of females was found in the Anchorage cohort than the Fairbanks cohort, with 41 percent of the Anchorage youth and 32 percent of the Fairbanks youth being female. The racial composition of both cohorts is shown in Table 3. Almost half (49%) of the youth referred in Anchorage were White and slightly over half (55%) of those referred in Fairbanks were White. Alaska Native youth were more prevalent in the Fairbanks cohort (30%) than in the Anchorage cohort (15%). Conversely, other minority youth were more prevalent in the Anchorage cohort than in the Fairbanks cohort—Black youth made up 10 percent of the Anchorage cohort and seven percent of the Fairbanks cohort, Asian youth made up seven percent of the Anchorage cohort and one percent of the Fairbanks cohort, and Pacific Islander youth made up four percent of the Anchorage cohort and zero percent of the Fairbanks cohort.
Group-based models were estimated for both the Anchorage and Fairbanks cohorts. In each cohort, we searched for groups of individuals who shared common histories of delinquent behavior. Delinquent behavior was measured by the annual rate of charges referred to DJJ while controlling for periods of detentions and institutionalizations, when individuals are not at-risk of offending. Additional statistical details are available on the Justice Center website (http://justice.uaa.alaska.edu). We found five distinct developmental trajectories for youth in the Anchorage cohort and three distinct developmental trajectories for the youth in the Fairbanks cohort. The different developmental trajectories are shown in Figures 1 and 2, and are summarized in Table 4. Again, these developmental trajectories show groups of offenders that share similar delinquent histories. In both Anchorage and Fairbanks, we found clear evidence of a low delinquency group and an early starter/desister group.
Youth in the low delinquency group had an offending rate that very slowly increased over time, never surpassing one referred charge per year. Just over half (51%) of the Anchorage youth and over three quarters (82%) of the Fairbanks youth were classified in the low delinquency group. Youth in the early starter/desister group started offending early, offended at a high rate, but began to desist by age 17. More specifically, the average offending rate for the early starter/desister group began to increase early (age 13), peaked at a high level at age 15, and then began to decline. In both Anchorage and Fairbanks, the increase in the average offending rate peaked at approximately four referred charges per year. The early starter/desister group included few youth (6% in Anchorage and 9% in Fairbanks). It is important to emphasize that although these youth began to offend at an early age and subsequently offended at a high rate, they also showed signs of desistance by age 17. This is an important finding because starting to offend at an early age is considered a strong risk factor for a long criminal career. Although the youth in this group entered the DJJ system early, they were able to show signs of desistance prior to adulthood. In part, this may have occurred because of the formal and informal services that they received. Understanding what reduced the delinquency of these youth is an important topic for future research.
In Fairbanks, we found an additional group of youth that showed signs of desistance by age 17. Contrary to the youth in the early starter/desister group, these youth did not start offending early. Youth in the late starter/desister group showed no increase in offending until age 15. More specifically, their average offending rate started to increase at age 15 and peaked at age 16 (at an average of 4.3 charges per year). At age 17, the average offending rate for youth in the late starter/desister group then dropped to 3.3 referred charges per year. Ten percent of the Fairbanks youth were classified as late starters/desisters. No comparable group of youth was found in Anchorage.
Three other groups were found in Anchorage. The moderate delinquency group included youth whose offending rate remained very low up to age 12, moderately increased at age 13 and 14, and then decreased. The average number of charges referred to DJJ for youth in the moderate delinquency group was 0.61 at age 13 and 0.53 at age 14. By age 17, these youth completely desisted, with an average of zero charges referred to DJJ. For these youth, delinquent involvement occurred primarily at age 13 and 14. Seventeen percent of the Anchorage youth belonged to this moderate delinquency group.
The other two groups found in Anchorage both showed signs of persistence rather than desistance. One group started offending early while the other started offending late (but neither showed any sign of desistance). The offending rate for youth in the early starter/persister group steadily increased from age 12 to 17, peaking at age 17. At age 17, youth in the early starter/persister group were referred to DJJ for an average of 4.2 charges. Five percent of the Anchorage youth were classified as early starters/persisters. Youth in the late starter/persister group began offending later (age 16) and offended at a lower rate. At age 17, youth in the late starter/persister group were referred to DJJ for an average of 2.1 charges (rather than the 4.2 for youth in the early starter/persister group). Seventeen percent of the Anchorage youth were late starters/persisters.
A limitation of these developmental trajectories is that they only provide a representation of the delinquent development that occurred for the youth in these two cohorts. Youth born today may have very different patterns of delinquent development. Nonetheless, these historical results are important because they allow us to now retrospectively assess the extent to which the delinquent development of these youth varied by race.
Profiles of Group Membership
In Tables 5 and 6, we examine the demographic composition of each offending group in the Anchorage and Fairbanks cohorts. Summary statistics for each offending group are provided showing their composition by race and gender. Statistical tests were performed to examine if the percentage of youth within each racial and gender group varied significantly across offending groups (i.e., to determine if the percentage of White youth was the same in each offending group or whether White youth were over-represented in some offending groups). Results for the Anchorage cohort are presented in Table 5. The percentage of Black, Asian, and Pacific Islander youth did not vary significantly across offending groups. Conversely, the percentage of White, Native, and multiracial youth did vary significantly across offending groups. Too few multiracial youth were included in the Anchorage cohort to specifically locate the significant difference. For both White and Native youth, the differences occurred between the low delinquency group and the early starter/desister group and between the moderate delinquency group and the early starter/desister group. White youth were over-represented in the low and moderate delinquency groups, but were less likely to be found in the early starter/desister group. The opposite result was true for Native youth. They were over-represented in the early starter/desister group, and were less likely to be found in the low and moderate delinquency groups. Not surprisingly, males were more likely to be in the three high delinquency groups (early starter/desister, early starter/persister, and late starter/persister) while females were more likely to be in the two low delinquency groups (low delinquency and moderate delinquency).
Results for the Fairbanks cohort are presented in Table 6. Results are only presented for White, Native, and Black youth because the number of other minority youth in Fairbanks was too small (see Table 3). Youth who were not White, Native, or Black and youth whose race was unknown are included in the “other/unknown” category (N = 47). The percentage of Black youth and the percentage of youth in the “other/unknown” category did not vary significantly across offending groups. As in Anchorage, the percentage of White and Native youth did vary significantly across offending groups. White youth were over-represented and Native youth were under-represented in the low delinquency group. Conversely, White youth were under-represented in the early starter/desister group and in the late starter/desister group while Native youth were over-represented in these two groups. A significant gender difference was found between the low delinquency group and the early starter/desister group, with significantly more females in the low delinquency group.
Summary and Implications
Overall, three groups of youth were identified in Fairbanks. The first included youth who offended at very low rates (low delinquency group). The second and third groups both included youth who offended at high rates but began to desist by age 17, with some youth (early starters/desisters) starting to offend earlier than others (late starters/desisters). In Fairbanks, all groups showed signs of desistance by age 17. Additional research should examine why none of the Fairbanks youth belonged to an offending group that persisted.
Three additional groups were found in Anchorage. The moderate delinquency group showed low levels of offending from age 10 to 12, moderate levels of offending at age 13 and 14, and low levels of offending thereafter. The two other groups both showed signs of persistence, with one starting earlier (early starters/persister) than the other (late starter/persisters). The offending rate for the early starters/persisters began to increase at age 13 while the offending rate for the late starters/persisters began to increase at age 16. Neither group showed any signs of desistance by age 17. Additional research should examine why none of these youth began to desist by age 17.
White youth were over-represented in low delinquency trajectories while Native youth were over-represented in trajectories that showed higher rates of contact with DJJ. The difference between the low delinquency group and the early starter/desister group was found in both Anchorage and Fairbanks. This result is important because it shows that disproportionate minority contact (as defined by the difference in rates of charges referred to DJJ) was evident by age 13. At that age, Native youth were already disproportionately referred to DJJ relative to White youth. Interventions designed to reduce the disproportionate contact of Native youth must therefore begin early. In Fairbanks, the disproportionate contact of Native youth also occurred for youth who started to increase their offending rate at age 15 (i.e., for youth in the late starter/desister group). We did not find this result in Anchorage, but the Anchorage cohort did not have a late starter/desister group (all late starters persisted).
It is again important to emphasize that the majority of youth contacted by DJJ had very few referred charges from age 10 to 17. Descriptive statistics (Table 1) showed that 51 percent of the Anchorage youth and 44 percent of the Fairbanks youth had only one charge referred to DJJ from age 10 to 17. The group-based models showed that the most common trajectory was one of low delinquency (for 51% of Anchorage youth and 82% of Fairbanks youth), and this was particularly true for females. Unfortunately, Native youth were less likely to be found in the low delinquency trajectory. On a more positive note, Native youth were not disproportionately found in the two groups that persisted. Instead, Native youth were disproportionately found in a group that desisted (in both Anchorage and Fairbanks). This again suggests that some youth received the necessary formal and informal services to reduce their delinquent behavior, particularly in Fairbanks. In Anchorage, two groups showed no signs of desistance by age 17. The late starter/persister group is particularly problematic because it included 17 percent of youth and it was not identifiable until age 16. This leaves little time for successful interventions. Understanding what caused the changes in referral rates observed in Figures 1 and 2 is, as Jones and Nagin argued, one of the most fundamental and empirically important research topics. Understanding these developmental trajectories in more detail could have significant impacts on policy.
André B. Rosay is an Associate Professor and the Interim Director of the Justice Center. Ronald S. Everett is an Associate Professor with the Justice Center. This project was supported by Grant No. 2005-IJ-CX-0013 awarded by the National Institute of Justice, Office of Justice Programs, U.S. Department of Justice. Points of view in this document are those of the authors and do not necessarily represent the official position or policies of the U.S. Department of Justice. The authors thank the Alaska Division of Juvenile Justice.