Abstract
Recent studies of the transition to adulthood advocate taking a person-centered approach and modeling key transitional events simultaneously. This article advances this literature by focusing on precarious transitioning among at-risk youth and relating their transition experiences to criminal offending. I find evidence for three distinct “pathways” to adulthood. Those with juvenile convictions are equally likely to take one of two “precarious” routes to adulthood—an early family starter pathway or a stalled pathway. Importantly, early family starters are much less likely than stalled transitioners to offend as adults. The findings suggest the transition to adulthood represents a fork in the road for juvenile delinquents in which early family starting serves as an avenue out of continued offending.
Group-based trajectory modeling is now commonplace in the life course and developmental criminology literature. Traditional methods assume that the parameters of the offending trajectory are normally distributed in the population. Group-based modeling applications relax this normality assumption and capture a more complex reality (Blokland & Nieuwbeerta, 2005; Nagin, 2005).
After grouping respondents following like trajectories, it is common practice to distinguish the groups along other variables of interest, including transition-to-adulthood events. After identifying four trajectory groups in a sample of Dutch offenders, Blokland, Nagin, and Nieuwbeerta (2005) found that the “high-rate persisters” were the least likely to have been married and have children, and the most likely to have been unemployed. Blokland and Nieuwbeerta (2005) found “persistent offenders” to be least likely to change as a result of events such as marriage. Piquero, Brame, Mazerolle, and Haapanen (2002) similarly found full-time employment and marriage to operate differently across offender groups. Yet, after identifying four trajectory groups in a Boston sample, Laub, Nagin, and Sampson (1998) found marriage to inhibit criminal offending even after controlling for group membership.
In short, to investigate links between crime and the transition to adulthood, prevailing group-based modeling techniques cluster respondents on the basis of their criminal trajectories, then look retrospectively at the life-course characteristics that differentiate those on divergent paths. This article recognizes the value of group-based approaches for capturing diverse life experiences but turns the usual analysis on its head. Because transition events constitute the heart of the life course and have been found to be important catalysts in perpetuating and redirecting criminal trajectories, I group respondents on the basis of the transitional experiences themselves. I then evaluate whether the groups are useful in understanding criminal offending over the life course.
Drawing on data from the Cambridge Study in Delinquent Development, this article addresses three questions. First, is there evidence of transitional clusters that are distinct in terms of how members experience transitions out of school and into employment, independent residence, marriage, and parenting? Second, are juvenile offenders disproportionately likely to experience precarious clusters of transitions that include early school leaving and young parenting? Third, are the clusters related to adult criminal behavior in ways that suggest the entire transition-to-adulthood period (and not simply isolated events such as marriage and employment) is consequential to criminal desistance?
Literature Review
This article builds on a tradition of “person-centered” approaches in the sociological literature (Magnusson & Bergman, 1990). Person-centered methods have been contrasted with more prominent variable-centered approaches. The latter describe associations between variables and are well-suited to the task of understanding the relative contributions that predictor variables make to a given outcome. Person-centered approaches enable the researcher to understand how experiences across life domains come together in the individual and locate groups of individuals who share patterns of life-course characteristics (Singer, Ryff, Carr, & Magee, 1998). Proponents view person-centered techniques as a stronger methodological fit to life-course theories.
Macmillan and Eliason (2003) argue, for instance, that traditional variable-centered methods are ill-equipped to handle the interdependent and dynamic nature of the life course. They use data from the National Longitudinal Survey of Youth 1979 (2005) and describe the timing and order of social roles in four social institutions: education, employment, marriage, and parenting. They argue that modeling these roles simultaneously highlights the “general diversity that exists in the structure of the life course” (p. 546).
Osgood and colleagues (2005) apply latent class analysis to data from the Michigan Study of Adolescent Life Transitions and identify six groups of respondents based on their role statuses at age 24: fast starters, parents without children, educated partners, educated singles, working singles, and slow starters. They find that the path one takes reflects the social-class values and resources of one’s family and one’s own academic success at age 18.
Sandefur, Eggerling-Boeck, and Park (2005) note the strength of latent class analysis in locating interconnections among educational attainment, marriage, childbearing, work, and living independently. They differentiate patterns of “successful” and “unsuccessful” transitions to adulthood. They discuss how these patterns reflect changes in social and demographic behaviors over time by comparing results based on data from the High School and Beyond Survey and the National Educational Longitudinal Study.
Macmillan and Copher (2005, p. 859) similarly recognize that “the meaning of any given role is dependent upon the presence or absence of other roles.” They apply a latent class modeling strategy to understand role configurations among women of the National Longitudinal Survey of Youth (1979). Their analysis highlights the diverse contexts in which teen parenting occurs.
Despite these conceptual and substantive advances in the literature on the transition to adulthood, there remain important gaps. This project addresses two. First, extant research focuses overwhelmingly on the general population. We lack a holistic picture of the transition experiences of vulnerable youth who are especially likely to experience precarious transitions, including early school leaving, work instability, and young parenthood (Osgood, Foster, Flanagan, & Ruth, 2005). To begin to fill this gap, this article focuses on at-risk, working-class, urban youth. These boys are not only likely to experience precarious transitions but are also at risk of heightened criminal involvement. Thus, this article addresses a second gap in the literature by relating their transitional pathways to criminal offending. Researchers have seldom studied “explicit pathways through life as a mechanism that links origins and destinations” (Macmillan & Eliason, 2003, p. 548). I aim to accomplish this by identifying pathways to adulthood and exploring how these pathways link juvenile and adult offending.
Criminal offending is known to rise through the teenage years and come to a peak in late adolescence before declining sharply in early adulthood. Because the decrease in offending occurs during the same period in which many individuals enter adult roles, there is a potential for early delinquency to jeopardize successful entrée into adult roles. Juvenile delinquents may be especially “at risk” of experiencing troubled transitions. These transitions may, in turn, contribute to continued criminality as well as compromise educational, occupational, and psychological and other later life outcomes.
Criminological theories of cumulative continuity support this kind of cyclical development. Noting widespread desistance in late adolescence and early adulthood, life course and developmental criminologists seek to understand what drives a minority of criminals to persist beyond this norm, well into adulthood. Thornberry’s (2005) interactional theory proposes “a series of mutually reinforcing causal loops” wherein antisocial behavior leads to weakened social bonds, increased deviant peer relations, and academic failure, which leave delinquent youth ill-prepared for successful, noncriminal adult life. Similarly, Moffitt (1993) posits that life-course “snares” including teen pregnancy and interrupted education often follow early offending and contribute to criminal continuity. Sampson and Laub’s (1997) life-course theory of cumulative disadvantage posits that juvenile offending disrupts the social bonds necessary for a young person to make successful entry into adult roles, including marriage and employment. Without these forms of attachment to conventional society, individuals lack the social control that would otherwise inhibit criminal behavior.
Absent from these accounts, however, is a holistic picture of the transition to adulthood. Recent work suggests that studying transitions simultaneously would be helpful in understanding trajectories of offending among at-risk youth. In their analysis of criminal desistance, Giordano, Cernkovich, and Rudolph (2002) find that marriage and stable employment are more important when they occur together as a complete “respectability package” than when they occur alone. Hagan and Wheaton (2003) similarly propose that researchers give more attention to linkages between behaviors. They promote conceptualizing pathways that include “exits” from adolescent roles coupled with entrances into adulthood.
This article advances this literature by modeling the five key transitional events simultaneously. Doing so illustrates the typical transition to adulthood experiences of at-risk, working-class boys. I ask whether conceptualizing these transitional experiences as representing diverse pathways to adulthood provides a better summary than viewing them as representing a single, average pathway. I then examine how the pathways are differentially related to juvenile delinquency and criminal outcomes in adulthood.
Method
The goal of identifying the typical pathways to adulthood taken by at-risk, working-class boys motivates the use of latent class cluster analysis (LCCA). In LCCA, a statistical model is postulated that includes a categorical latent variable, assumed to explain all of the association between the indicator variables (here, the five transitions). The categories are mutually exclusive and exhaustive. Model-based probabilities are used to identify the latent classes in number and kind, and to classify the cases into classes. These probabilities are estimated by maximum likelihood estimates, where each indicator (y) represents one of the five key transitions. This probability structure takes the following form:
where
A central advantage of LCCA over traditional cluster methods is the availability of formal diagnostic statistics to assist the researcher in model selection. Weighing these criteria, the most parsimonious model that provides adequate fit is typically preferred. Respondents are assigned membership to the class for which their posterior probability is the highest.
After identifying the typical pathways taken by at-risk boys, I ask whether the pathways are differentially related to juvenile and adult offending. In the first set of models, juvenile offending is used to predict transitional cluster membership using multinomial logistic regression. A number of individual and family controls are then included to better understand what other factors predict transitional cluster membership.
In the third and final stage of analysis, negative binomial regression is used to understand how one’s transitional pathway is related to later life offending. Negative binomial regression is used to model these data after a likelihood-ratio test for overdispersion indicated a low probability (p < .001) that the data were generated by a Poisson process. Controls that were significant in predicting transitional cluster membership in the second stage of analysis are then included to understand whether these same characteristics are responsible for the divergent criminal trajectories of the transitional clusters in adulthood.
Data
The Cambridge Study in Delinquent Development is a longitudinal study that began in 1961 when the respondents were 8 years old (Farrington, 1994; West & Farrington, 1973, 1977). A sample of 411 boys was selected from schools in an urban, working-class area of London. Most of the boys were white (97%). All were interviewed at ages 8, 10, 14, 16, and 18. At age 21, only delinquents who had ever been convicted were interviewed along with a similarly sized, but randomly selected sample of never-convicted youths. This effectively led to an oversample of ever-convicted youth. Of the 241 targeted, 218 (90.5%) were interviewed. Because several of the transitional behaviors of interest were measured at age 21, the analyses in this article are limited to these 218 youth. Fortunately, previous research has evaluated the extent to which the groups identified in group-based modeling strategies are robust to sample size. Although the groups vary with sample size, the number and descriptive picture of groups appears to stabilize in samples of 200 and greater (Sampson, Laub, & Eggleston, 2004). 1
Due to the distinctiveness of this sample, care must be taken when evaluating this study’s implications to boys in the general population. I conclude this article by discussing what the findings suggest about at-risk youth growing up in modern inner cities and outside of the British context. At the same time, this article answers a call from life-course researchers to identify contextual variation in the structure of the life course (Macmillan & Eliason, 2003). Despite the many studies that have examined the relationship between role transitions and crime, we know little about whether the relationships vary over time and place (Siennick & Osgood, 2008).
Measures
Education
There is no concept of “graduating” in the United Kingdom. The General Certificate of Education (GCE) examinations serve as the closest counterpart to the United States’ diplomas. This variable identifies those who, by their interview at 18, had taken the GCE or any other credentialing or occupational field examinations (coded 1). Students who leave school without passing any examinations are comparable with U.S. high-school dropouts in the sense that they leave with no credentials (coded 0; Kerckhoff, 1990).
Employment
The transition to stable full-time employment is considered to be a critical developmental task (Caspi, Wright, Moffitt, & Silva, 1998). If a respondent was unemployed for fewer than 4 weeks in the 2 years preceding his 21-year-old interview, he is considered to have achieved stable employment (coded 1). Time out of work is not counted if the respondent was in school full-time. Two cases are missing values on this indicator but are nevertheless included in the cluster analysis. Their missing values are handled in the likelihood function in which their log likelihood (LL) contribution is based on the other observed indicators.
Independent residence
In the 1970s, the median age at home leaving was 18.3 in the United States but 21.7 in Great Britain (Kerckhoff, 1990). Therefore, the present analysis measures early home leaving as having occurred by age 21.Those in college, the armed forces, or institutions are coded according to their situation during leaves or release.
Marriage
Those who had ever been married by the time they were interviewed at 21 are considered to have married young, regardless of whether the marriage ultimately ended in separation or divorce. This is an appropriate age at which to capture early marriages as the median age at first marriage in the 1970s was about 23.9 for British males (and 22.3 for U.S. males; Kerckhoff, 1990).
Parenthood
Those who had fathered a child by 21 are considered young parents. Just eight of these births occurred within marriages. Due to this low variability, the cluster analysis does not distinguish those who became fathers within marriage from those who became fathers outside of marriage. The article returns to issue of out-of-wedlock births descriptively, however.
Convictions
Criminal conviction data through age 50 were obtained by the original investigators through the Central Criminal Record Office in London. These data exclude convictions for traffic offenses, public drunkenness, and minor forms of assault. Pretransitional convictions include those accrued at ages 10, 11, 12, 13, and 14 to capture youthful offending that occurred prior to the minimum school leaving age of 15. Because the other transitions are assessed at 21, posttransitional convictions include those experienced between 22 and 50, inclusive. It is typical to adjust the number of offenses for time free in the community. In the present data, it is only known whether the respondent was not at risk for the entire age of interest. As a consequence, the count data may be biased downward for those who were only free in the community part of the year. Binary measures are less sensitive to this problem and yield results that are consistent with those presented below. Nine cases are missing posttransitional conviction data and are excluded from the associated analysis.
Controls
Several individual and family control variables are included in the analysis because of their theorized joint relationship with criminal and transitional behavior. IQ was tested at age 8 and 10 and is considered to be a stable individual characteristic. This article uses a measure of IQ that averages those two scores in the form of a 4-point scale representing the quartiles: low (90 or less), low average (91-98), high average (99-109), and high (110 or more). A number of measures from the interviews at age 14 are included to understand adolescent behavior just before the minimum school leaving age of 15, in other words, just prior to the transition-to-adulthood period. Low school commitment is a measure of whether the boys were “frequently truant,” as opposed to “occasionally” or “never” truant according to their teachers. At 14, the boys self-reported whether they had participated in various delinquency items: “never,” “once or twice,” “sometimes,” or “frequently” in their lifetime. To tap possible heterogeneity between transition groups that is not apparent in the official conviction data, this analysis includes a composite measure of self-reported delinquency, which identifies respondents who participated in any of the following six offenses more than “once or twice”: using a weapon in a fight, damaging property, stealing out of a motor vehicle, stealing a vehicle (or joyriding), breaking and entering, and shoplifting. Because adolescent relationships also play a role in shaping the transition-to-adulthood experiences and the delinquency of at-risk youth, I also include a measure of delinquent peers. The data include the boys’ self-reports, at 14, of the number of different delinquent acts committed by their friends: “9 or less,” “10 to 14,” “15 to 21,” “22 or more.” This measure indicates boys whose friends committed 22 or more delinquent acts.
The analysis also captures some important ways in which families influence adolescent criminal and transitional behavior. Parents’ education and work experiences may shape a boy’s educational, and employment expectations and desires. 2 The following variables were collected when the boys were 10, the oldest age at which they are available. Mother’s education and father’s education are dichotomous measures of whether each parent went to school beyond the minimum school leaving age in Britain at the time. Mother’s employment status identifies mothers who were working outside of the house. Father’s employment identifies fathers who were steadily employed. Given evidence that parents’ own misbehavior may influence their children, parental criminality is included and measures whether either parent was convicted of a crime by the time the boy was 10. Finally, a number of family conditions shape a boy’s financial flexibility, ability to remain living with his parents, and continue his education. Parental separation is a measure of whether the boy’s parents were no longer together for any reason other than death by the time the boy was 15. Family income is an assessment made by the interviewers on the basis of income, size of family, and level of spending, when the boy was 8. I include two dummy variables representing “comfortable income” and “inadequate income” classifications, where “adequate income” serves as the base category (see Table 1). 3
Descriptive Statistics.
Results
Table 2 displays the fit statistics for the models estimated. The evidence suggests that the null model should be rejected in favor of a multicluster model. The boys’ transitional experiences are not adequately represented by a single pathway to adulthood. The p values and bootstrap p values of models consisting of three, four, and five clusters are all greater than .05 indicating adequate model fit. Of these models, the three-cluster model, which has a bootstrap p value of .21, is the most parsimonious (17 parameters) and has the lowest Bayesian information criterion (BIC) statistic (1,191.00) and Akaike information criterion (AIC) statistic (1,133.46). Deviance-based tests comparing the deviance statistics (−2LL) of competing models confirm that the three-cluster model provides a better fit to the data than the two-cluster model (p < .001), but that a four-cluster model does not provide a better fit than the three-cluster model (p > .05; not shown). Finally, there are several significantly high bivariate residuals (BVRs) associated with the two-cluster solution, indicating associations between pairs of indicators left unexplained by the model (not shown). The BVRs associated with Model 3 are all acceptably low. The average posterior probabilities for the three clusters are .97, .99, and .82, which are well above .7, the recommended threshold for assignment accuracy in group-based modeling (Nagin, 2005). 4
Fit of Latent Class Models by Their Numbers of Clusters (n = 218).
Note: LL = log likelihood; BIC = Bayesian information criterion; AIC = Akaike information criterion.
Figure 1 displays the three clusters representing the three pathways to adulthood. The vertical axis represents the percentage of each cluster who reported experiencing the transition of interest. The transitions are shown on the x-axis.
Three pathways to adulthood.
I refer to the largest cluster as the “work- and education-minded” transitioners. Just more than half (51.4%) of the sample (112 respondents) had the highest posterior probability of belonging to this cluster. Those in this cluster were the most likely to have taken academic exams by age 18 (57.7% had), and all had achieved stable employment by 21. Yet, they were unlikely to have started their own homes and families by 21. Only 8.9% had transitioned to independent living by age 21. None had ever married or fathered a child by age 21.
Cluster 2 is the second largest cluster and represents 28.4% of the sample (62 respondents). I refer to this cluster as the “early family starters.” Just 32.3% of the early family starters had taken any credentialing exams by 18 years of age. Seventy-nine percent had transitioned smoothly into employment. They most stand out from the other clusters, however, with respect to their transitions to independent residence, marriage, and parenting. About 82.3% had moved out of the parental home by 21, 98.4% had married by 21, and 48.4% had become fathers by 21. (It is worth noting that 34.1% of their marriages were “forced” in the sense that a child was conceived prior to the marriage; 73.3% of those who had children by 21 did so out-of-wedlock.)
I call the smallest cluster the “stalled transitioners.” The stalled transitioner cluster represents about 20.2% of the respondents (44 respondents). Stalled transitioners were the most likely of the three clusters to have left school without credentials and to have experienced early unemployment. Just 30.2% of them took any credentialing exams and none of them transitioned to stable employment by 21. About 34.1% had transitioned to independent living circumstances by 21 and, although none had married, 11.4% had already become fathers by 21. In other words, all of those who had had children by 21 did so out-of-wedlock. I call them stalled because of their few transitions to new roles and responsibilities across these domains.
Together these clusters illustrate the diverse transitional experiences of at-risk, urban youth. Just more than half of the boys took what most would consider the ideal or “normative” pathway. They spent this early stage of the transition to adulthood reaching educational and employment milestones, while postponing entry into marriage, independent living, and parenting. By contrast, the other two pathways appear more precarious. Those on these pathways were more likely to take steps—including leaving school without credentials and becoming young parents—that might undermine long-term stability and success. The following section investigates whether early criminal offending predicts who will take these more precarious routes.
Table 3 displays the results of the multinomial logistic regression analyses of cluster membership on pretransitional delinquency. The first two comparisons use the work- and education-minded transitioners as the base category. The third comparison uses the early family starters as the base category so that all possible cluster comparisons could be displayed in Table 3. For each comparison, Model 1 (M1) displays the odds ratio associated with regressing cluster membership on pretransitional delinquency with no controls included.
Odds Ratios From Multinomial Logistic Regression Analysis of Cluster Membership on Pretransitional Convictions (n = 200).
Note: M1 = Model 1; M2 = Model 2; M3 = Model 3; M4 = Model 4. Standard errors are in italics.
p < .10. **p < .05. ***p < .01. ****p < .001.
These results indicate that boys with early conviction records are especially likely to take one of the precarious routes to adulthood—the stalled transitioner pathway or the early family starter pathway. Each additional conviction between 10 and 14 increases the odds of following the early family starter pathway rather than the work and education pathway by a factor of 1.846 (p < .01) and increases the odds of following the stalled transitioner pathway rather than the work- and education-minded pathway by a factor of 2.171 (p < .01). Additional convictions do not make one precarious pathway a more likely route than the other, however. According to a Wald test for combining alternatives (not shown), the hypothesis that the two pathways are indistinguishable with respect to their relationship to pretransition convictions cannot be rejected (p > .10).
Models 2 to 4 investigate a number of individual- and family characteristics that may foreshadow early delinquency and transitional behavior. Individual controls are added in M2. Only “delinquent peers” surfaces as significant in understanding which transitional pathway a boy will take. “Delinquent peers” is therefore retained in M3 in which the family characteristics are added. The final model, M4, includes only those characteristics that were significant in M3: delinquent peers and parental criminality. 5
M4 results suggest that delinquent peers and parental criminality are not significantly helpful in distinguishing early family starters from work- and education-minded transitioners after controlling for pretransitional delinquency. These characteristics are helpful, however, in understanding who becomes a stalled transitioner rather than work and education focused. Controlling for juvenile convictions and parental criminality, boys with delinquent peers experience 4.724 times higher odds of becoming stalled transitioners rather than work- and education-minded transitioners (p < .001). With pretransitional convictions and delinquent peers held constant, having at least one parent with a criminal record is associated with 3.309 times higher odds of becoming stalled rather than work and education focused during the transition to adulthood. The coefficient associated with pretransitional convictions declines from 2.171 in M1 to 1.638 in M4. Although insignificant at α = .05, the associated p value is .057. The results associated with Models 2 and 3 indicate that this decline is mostly due to delinquent peers. This suggests that the relationship between pretransitional convictions and becoming a stalled transitioner rather than a work- and education-minded transitioner is either partly spurious due to delinquent peers or partly explained by delinquent peers. Parental criminality, on the other hand, is better understood as an additional characteristic helpful in distinguishing stalled transitioners from the work- and education-minded transitioners. Importantly, although the two precarious pathways cannot be distinguished on the basis of their pretransitional convictions, having delinquent peers increases the odds of becoming stalled rather than an early family starter by a factor of 3.512, and parental criminality increases the odds of becoming stalled rather than an early family starter by a factor of 2.508.
In short, compared with juvenile delinquents, nondelinquents begin the transition to adulthood with more security; they spend their late teen years and early 20s acquiring academic credentials and achieving steady employment. Juvenile delinquents are more likely to take one of two precarious routes to adulthood, either the early family starter pathway or the stalled pathway. The following section explores whether the pathways taken have implications for later life offending.
Table 4 displays the results of the negative binomial regression analysis of posttransitional convictions on cluster membership. The three negative binomial regression models described in Table 4 were each run twice so that all three cluster comparisons could be shown. The first two cluster comparisons include the work- and education-focused transitioners as the reference group. The third comparison includes early family starters as the reference group.
Incidence Rate Ratios From Negative Binomial Regression of Posttransitional Conviction Count on Transitional Cluster and Controls (n = 209).
Note: M1 = Model 1; M2 = Model 2; M3 = Model 3. Standard errors are in italics.
p < .10. **p < .05. ***p < .01. ****p < .001.
M1 displays the results of regressing posttransitional conviction count on transitional cluster with no controls added. These results reveal that stalled transitioners experience 5.462 times as many convictions between ages 22 and 50 as the work- and education-minded transitioners (p < .001). Although stalled and early family starters could not be distinguished on the basis of pretransitional offending, stalled transitioners are significantly more likely to accrue posttransitional convictions—4.456 times as many—than early family starters. In the posttransition period, early family starters look much more like the work- and education-focused transitioners. Indeed, there is no significant difference in the number of posttransitional convictions experienced by these two groups. These results suggest that the transition to adulthood represents an important fork in the road for juvenile delinquents. Although they are indistinguishable on the basis of early convictions, juvenile delinquents who take the early family route are less likely to offend in early adulthood than those who become stalled.
As discussed above and displayed in Table 3, the stalled transitioners were significantly more likely than the early family starters to have had delinquent peers and at least one parent with a criminal conviction. These differences represent heterogeneity among the juvenile offenders, which may help account for the stalled transitioners’ higher numbers of convictions in adulthood. Rather than interpreting the early family starter pathway as transformative, it may be that juvenile delinquents who have criminally active friends and parents become entrenched in contexts conducive to crime, whereas those with more prosocial relationships self-select into early family starting.
M2 therefore includes delinquent peers and parental convictions. Because having delinquent peers proved insignificant, the final model (M3) includes only parental criminality. The results indicate that controlling for transitional pathway, those boys who had at least one parent with a criminal record experience nearly twice (1.994) as many convictions in adulthood (p < .05). The coefficient associated with stalled rather than work- and education-focused transitioning declines from 5.462 to 4.392. The coefficient associated with taking the stalled pathway rather than the early family starter pathway declines from 4.456 to 4.166. Importantly, the coefficients displayed in the final model remain large and statistically significant. Thus, even after controlling for early heterogeneity, the relationship between transitional behavior and later offending remains quite strong.
Summary
In general, the work- and education-minded cluster represents the most “normative” or ideal pathway to adulthood. The majority of respondents in this cluster followed traditional expectations about leaving school with credentials and entering the workforce smoothly without major periods of early unemployment. At 21, it appeared that these respondents were postponing family formation until these educational and occupational transitions were achieved. Very few of them experienced early transitions into parenthood or marriage, and perhaps because of that, they were also unlikely to have moved out of their parents’ homes early. Not surprisingly, those on this pathway to adulthood were less criminally active than those on the other two pathways.
Boys with criminal convictions by age 14 were significantly more likely to take one of two precarious pathways: the stalled transitioner route or the early family starter route. The majority of these boys left school without any credentials to show for their time, but the pathways diverged as the boys continued their transitions to adulthood. A higher proportion of early family starters transitioned smoothly into employment, and experienced early transitions out of their parents’ home and into marriage and fatherhood. Not only were the stalled transitioners less likely than the early family starters to experience these transitions, but none had achieved stable employment by 21. The 11% who became young fathers did so outside of marriage. Importantly, although they could not be distinguished on the basis of early convictions, the criminal behavior of the early family starters became more subdued over time compared with the stalled transitioners. In adulthood, the criminal behavior of early family starters was on par with those on the more traditional pathway, which emphasized employment and education before starting a family. These relationships remained after controlling for a number of childhood and family characteristics, suggesting a transformative quality of the early family starting pathway.
Discussion and Conclusion
It is well known that criminal offending peaks in late adolescence, just as young people are beginning to make important choices about their life courses. It is also known that coming from working-class, urban environments places youth at risk of criminal participation and precarious transitioning. Yet, work on the relationship between crime and the transition to adulthood has been limited. Recognizing the value of group-based trajectory modeling for illustrating diverse offending patterns, this article has taken a related approach to locate the typical transition patterns of at-risk youth. The pathways identified reinforce the need to take a comprehensive view of the transition to adulthood to understand crime over the life course.
Developmental and life-course criminologists have long been interested in understanding criminal persistence. In their research on the role of cognitive transformation in the desistance process, Giordano and colleagues (2002) found that respondents who had not married or found stable work were the most likely to be persistent in their offending. The patterns highlighted in this article suggest, however, that it is not simply the lack of marriage and employment that contribute to criminal persistence but also the occurrence of precarious transitions such as early exits from school and into parenthood. This is consistent with Moffitt’s (1993) argument that life-course “snares” including precariously early transitions out of school and into parenting roles serve to entrap young offenders into criminal offending. However, these events, too, must be viewed in context as their criminogenic effect may be counteracted or ameliorated by early entry into adult responsibilities, including marriage and work.
Sampson and Laub (e.g., 1993) have emphasized the role of marriage and work in promoting desistance from crime. The patterns described in this article not only reinforce this notion, and suggest that these events are consequential even when they occur precariously early or are precipitated by early exits from school and early parenting. This is striking given the capacity for early exits to undermine the quality of employment and marriage. Indeed, early school leavers experience lower pay, fewer promotions, less job security, and are more vulnerable to economic shifts and evolving job markets (Wilson, 1996). Likewise, marriages that begin at an early age or in response to a pregnancy are more likely to be unstable and ultimately dissolve (Furstenberg, Brooks-Gunn, & Morgan, 1991).
Without disputing these and other long-term consequences of precarious transitioning, this article suggests that early family formation may have a positive side for juvenile delinquents who experience it, serving as an avenue out of continued offending. This is fortunate given that the transitional patterns identified among this sample of at-risk youth suggest that young offenders are unlikely to experience what Giordano has called a “complete respectability package” consisting of marriage and work without also experiencing early hiccups, including early school leaving and young parenting. It is likely that early marriage and early work stability not only play an important role in desistance but also offsetting early precariousness in the transition to adulthood itself. At the same time, early transitions out of school, out of parental household, and into parenting may play an important role in precipitating marriage and employment. The clustering of these events among those who desist from crime suggests it would be useful to more broadly conceptualize them as representing an interrelated “recovery package.”
This article is limited in its ability to solidify the transformative quality of the early family starter pathway. The most elaborate statistical tools for establishing causality fall outside of this person-centered analysis. Future analysis might also investigate the mechanisms by which transition experiences influence criminal behavior. It may be that periods of rolelessness allow time for “unstructured socializing activities,” which are conducive to crime (Osgood, Wilson, O’Malley, Bachman, & Johnston, 1996). During periods of rolelessness, stalled transitioners may lack the social bonds useful in discouraging misbehavior (Laub & Sampson, 1994). Early marriages, even if they are unstable in the long run, may provide important social control during the critical period of early adulthood.
It is also important to consider how the distinctiveness of this sample shaped the pathways identified. By design, the data used in this article overrepresent delinquent males because the investigators interviewed only a random subsample of the never-convicted youth at 21. Had they all been included, one would expect that the proportion of boys on the work and education pathway would have been higher, as this was the pathway most typical of nondelinquents. Because they had not been convicted, however, one would not expect their inclusion to substantially alter the core findings about the relationship between the pathways and crime. Moreover, given the highly contextualized nature of life-course transitions, they are largely contingent upon historical time and place. Had this study been set in a working-class community in today’s United States or United Kingdom, the emergent pathways would likely reflect important historical changes in the social and economic climate. Given the rising importance of education, those who leave school without credentials now face greater hardships in a job market in which opportunities for unskilled and semiskilled workers have become increasingly scarce (Wilson, 1996). The transition to stable, full-time employment may be more difficult in the United States than in the United Kingdom where the school-to-work transition is aided by employer-sponsored training programs (Kerckhoff, 1990). The age of marriage has also been on the rise since the 1970s making early marriages much rarer. Moving out of the parental home has become less tied to family initiation and there have been numerous changes in cultural norms regarding premarital sex, out-of-wedlock births, and father responsibility. In today’s working-class communities, early parenthood no longer “forces” marriage and independent living to the degree it once did.
For these reasons, young people who begin their pathways precariously by leaving school early and becoming young parents are now probably less likely to combine these transitions with early transitions into steady employment, marriage and independent living, which might otherwise aid their “recovery.” This may mean that youth with criminal backgrounds are more likely today than in the past to begin their transitions to adulthood precariously and become “stuck” on criminal trajectories. Although these hypotheses must be left for future research, it is clear that a full understanding of the relationship between crime and the transition to adulthood will be aided by its study across a variety of historical and geographic contexts.
Footnotes
Acknowledgements
The author thanks Robert Sampson, John Laub, Mary Waters, Chris Winship, David Farrington, Audrey Thomas, Chris Wimer, Laura Tach, and anonymous reviewers for their comments.
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
This research was supported by a Graduate Society Fellowship for Dissertation Completion from the Graduate School of Arts and Sciences, Harvard University.