Desistance from Crime during the Transition to Adulthood: The Influence of Parents, Peers, and Shifts in Identity

Abstract

Objectives:

Research on criminal continuity and change has traditionally focused on elements of the adult life course (e.g., marriage and employment); however, recent social and economic changes suggest the need to consider a broader range of factors. In addition, researchers have increasingly recognized the importance of identity changes in the desistance process.

Methods:

Using five waves of structured data from the Toledo Adolescent Relationships Study (TARS), we examined identity changes, shifts in involvement with delinquent peers, and variability in closeness with parents as influences on desistance. In-depth interviews with a subset of TARS respondents offered a person-centered lens on individual and social processes associated with variability in criminal behavior.

Results:

Findings indicated that identity changes were associated with declines in offending. In addition, changes in parental closeness and the extent of affiliation with antisocial peers contributed to patterns of offending, net of these subjectively experienced cognitive changes.

Conclusions:

Cognitive processes are important to desistance. However, they do not independently provide a path to sustained behavioral change. Social experiences, including changes in relationships/supports from parents and affiliation with delinquent peers, also figure into change processes. We discuss the implications of our findings for future research and programmatic efforts.

Marriage and employment have been central to theory and research focused on the process of desisting from crime (e.g., Farrington and West 1995; King, Massoglia, and MacMillan 2007; Sampson and Laub 1993; Uggen 2000). The emphasis on these important life-course transitions fits well with the notion of “asymmetrical causation,” or the idea that the causes of offending are likely to differ from the causes of crime cessation (Uggen and Piliavin 1998; see also Laub and Sampson 2001). Early findings supported this emphasis, as some research indicated that traditional risk factors associated with the onset of criminal behavior were not particularly strong predictors of desistance (Laub, Nagin, and Sampson 1998; Nagin, Farrington, and Moffitt 1995). Research instead has focused heavily on the ways in which marriage and employment provide offenders the elements of social control and stability necessary to support a change in life direction (e.g., Laub and Sampson 2003; Horney, Osgood, and Marshall 1995; Sampson, Laub, and Wimer 2006). Conversely, traditional risk factors (e.g., affiliation with delinquent peers, lack of attachment to parents, or antisocial family climate) have garnered relatively little attention as factors related to criminal persistence and desistance.

While marriage and employment are key adult experiences, our view is that an exclusive focus on these two domains does not result in a complete portrait of the adult life course or of the nature of desistance processes. One potentially significant complication is that, as compared to earlier generations, young adults are putting off marriage longer (often into their late 20s and beyond) or not getting married at all (Raley, Sweeney, and Wondra 2015). Employment experiences have also changed; the availability of stable blue-collar jobs has declined for those with only a high school education (Mortimer 2009; Schneider and Stevenson 1999), and frequent job changes have become increasingly normative (Metcalfe et al. 2003). Yet despite this changing social and economic landscape, declines in criminal behavior continue to occur, as revealed in longitudinal studies that have traced individuals from adolescence across the transition to adulthood (e.g., Farrington 1986; Shulman, Steinberg, and Piquero 2013). At the same time, researchers have highlighted variability within samples (see, e.g., Evans, Simons, and Simons 2016; Monahan et al. 2009), raising the possibility that previously observed general age trends and within-sample variability do not trace solely to marriage and employment circumstances.

Sampson and Laub’s theorizing about marriage and employment effects placed the emphasis on forces “external to the individual.” As interest in the area has developed, some researchers have focused attention on individual-level factors such as identity changes and other kinds of cognitive transformations that are not fundamentally dependent on these exogenous forces (Farrall and Bowling 1999; Giordano 2016; Giordano, Cernkovich, and Rudolph 2002; Maruna 2001; Paternoster and Bushway 2009). Although this focus on subjective changes offers a more agentic view of the desistance process, most of the research in this area has been based on qualitative research and retrospective accounts (Bersani and Doherty 2018; Giordano et al. 2002; Paternoster and Bushway 2009; but see Rocque, Posick, and Paternoster 2016). In addition, this line of research has the potential to foster a view of desistance as an individualistic process in which offenders simply decide that it is time to change, and subsequently things begin to fall into place.

The focus on subjective processes, including identity changes, has been a useful theoretical development, but as a standalone theory does not fully account for either (a) the difficulties some offenders face on the road to desistance (Halsey and Deegan 2015) or (b) the role of social experiences that can support or inhibit the process of making these significant life changes. In this analysis, we focus on two social domains—parents and peers—that figure heavily in studies of criminal “onset” but that have been relatively neglected in studies of adult desistance. Particularly within the contemporary context, parents and peers may continue to play a role as significant sources of support and influence, as more general research on the uncertain period of “emerging adulthood” has recently highlighted (Barry, Madsen, and DeGrace 2015; Fingerman et al. 2012; Young et al. 2015).

The current analysis draws on quantitative and qualitative data from the Toledo Adolescent Relationships Study (TARS), a prospective, longitudinal study of the lives of a diverse sample of teens first interviewed in 2001, and followed up four additional times across the adult transition period. The analyses will contribute beyond prior work that has focused primarily on transition events and subjective changes. Centering on relationships with parents and peers, we examine the degree to which time-varying measures of attachment to parents and peer delinquency are related to within-individual changes in self-reported delinquency/crime across the transition from adolescence to adulthood, taking into account marriage and employment circumstances. Next, we introduce measures of identity changes across the study period, focusing on the process of distancing from negative identities (troublemaker, partier), as well as the adoption of more positive self-views. This will allow us to determine the relative impact of each of these social and individual factors across a more complete roster of potentially important domains associated with behavior change. In addition, inclusion of the parent and peer variables along with indices of identity change will allow us to determine whether some of the effects of the changing character of relationships with parents and peers operate indirectly via their influence on identity change.

Detailed life history narratives elicited from a subset of respondents provide a window on how these social network and subjective factors may be linked within the life-course experiences of particular individuals in ways that either support or impede the process of enacting behavior changes. In contrast to many qualitative studies based on narrative accounts of change processes, searches of official records in the years after these interviews were completed provide objective data on whether respondents’ desistance narratives are backed up by a pattern of later avoidance of further involvement with the criminal justice system.

Background

A number of studies have examined associations between marriage, employment, and desistance from crime (e.g., Bushway and Apel 2012; Horney et al. 1995; King et al. 2007; Laub and Sampson 2003; Piquero, MacDonald, and Parker 2002; Sampson et al. 2006; Skardhamar and Savolainen 2014). Yet despite considerable empirical attention to these key turning points, findings provide a mixed portrait of the extent to which marriage and employment correspond to crime cessation. For example, King and colleagues (2007) found support for a marriage effect on crime for males, even accounting for their propensity to marry. However, using Norwegian registry data, Lyngstad and Skardhamar (2013) found that the propensity to commit crime declined prior to the transition to marriage, suggesting the role of other kinds of changes, including subjective ones. Studies of employment are even more mixed; Uggen (2000) found that employment programs did not reduce recidivism for younger individuals but had a beneficial effect among those aged 27 or older. These findings also convey that the older individuals may have differed from their younger counterparts in their general receptivity to programs that would be useful in the process of making changes in life direction. Also casting doubt on an automatic effect of securing employment, results of a meta-analysis indicate that employment programs alone were not effective in reducing recidivism (Visher, Winterfield, and Coggeshall 2005). In addition, a recent study by Skardhamar and Savolainen (2014) found that individual-level changes in offending often preceded the transition to stable employment.

A Changing Social and Economic Landscape

Recent social, demographic, and economic changes add a further layer of complication to the basic notion of the “good marriage effect” and job stability as complete explanations for criminal desistance. For example, the average age at first marriage in the United States is 29.5 for men and 27.4 for women, as compared to the historically low median age at first marriage of the baby boom in the 1950s—roughly 20 for women and 22 for men (U.S. Census Bureau 2017). There are also striking differences in marriage by race/ethnicity and education level, such that members of race/ethnic minorities and individuals with lower levels of education marry later, are less likely to marry, and have lower levels of marital stability (Edin and Reed 2005; Lichteret al. 1992; Martin 2006; Raley and Sweeney 2009). Furthermore, cohabitation has surpassed marriage as the most common union experience in young adulthood, and transitions into cohabiting unions occur, on average, earlier than transitions to marriage (average age at cohabitation is 22; Hemez and Manning 2017; Manning, Brown, and Payne 2014).

Aside from changes in the likelihood and timing of marriage, changes in the labor market, including the disappearance of industrial and manufacturing jobs, have restructured the occupational opportunities of low-skilled men and women (Wilson 2011). These macro-level processes have also widened the wage gap between those with a high school education or less and college graduates (Danziger and Ratner 2010). This set of social and economic changes has contributed to delays in other traditional milestones associated with the transition to adulthood (e.g., completing education, parenthood, independent living; Settersten and Ray 2010). Thus, given the changing contexts of development, and research indicating that most individuals desist from crime during early adulthood (Farrington 2004), it is important to consider other potential factors underlying desistance processes.

Identity Changes

Desistance that occurs among individuals who have not yet married or who have uncertain job prospects draws attention to other dynamics, including subjective processes, that may be implicated in successful behavior change. This is also suggested by the recent research, as noted above, documenting that some declines in crime among those who eventually married occurred prior to the entry into marriage itself, or among the employed, prior to the individual’s employment (see Visher et al. 2005). In a series of recent articles, Bushway and Paternoster argued for a parsimonious approach, suggesting that identity change is key to understanding desistance from crime. The authors described a general process characterized by increasing disenchantment with the old way of life and with one’s identity as a criminal (drawing on the idea of a “crystallization of discontent”), and in turn the development of a new, more prosocial self (Bushway and Paternoster 2013, 2014; Paternoster and Bushway 2009).¹

This viewpoint is in many respects compatible with other work on subjective changes, including theorizing about the role of agency in the desistance process (Carlsson 2013; Graham and Bowling 1995; Maruna 2001), as well as the focus on cognitive transformations (Giordano et al. 2002). However, a symbolic interactionist view of human development and change highlights to a greater extent the role of social experiences in the genesis of cognitions, including views about the self (Giordano 2010). Mead described the distinctively human capacity to consider oneself as an object but emphasized that these self-conceptions are, as Berger (1963:98) noted, “socially bestowed, socially maintained, and socially transformed.” This is an important insight, as it is somewhat intuitive to consider self-conceptions and changes in them as individual level, deeply personal accomplishments. Here, we examine the role of parent and peer relationships as direct influences on continuity/change in criminal behavior as well as their indirect influence on the identity shifts that may catalyze and help to sustain these life-course changes (Giordano 2017; Mead 1938).

Relationships with Parents across the Transition to Adulthood

Research on the onset of antisocial behavior has historically focused on disruptions in parent–child relationships, following from attachment theory’s emphasis on the negative effects of a breakdown of early bonds (Hirschi 1969). Yet while many studies have examined links between problematic relationships with parents and juvenile delinquency, few studies have considered the continuing role of parents on young adults’ offending trajectories (but see Johnson et al. 2011). Nevertheless, the more general literature on the adult transition period provides growing support for the idea that the parent’s role is not confined to the periods of childhood and adolescence (Fingerman et al. 2012; Swartz and O’Brien 2016). This developing literature suggests that parents may continue to play an important role for young adults not only for tangibles such as housing but as a more general source of emotional support (see, e.g., Aquilino 2006; Johnson et al. 2011; Schulenberg and Zarrett 2006). This may be especially important, given the uncertainties of growing up within a contemporary context characterized by delays and instability in adult union formation and less than ideal economic prospects (Osgood et al. 2005). Thus, scholars focused on the period of emerging adulthood in particular have recently highlighted the degree to which the family of origin continues to “matter” as a source of support and influence (e.g., Furstenberg 2010).

In a prior qualitative study, Giordano et al. (2007) found that some offenders described an emotional mellowing process, whereby those who had formerly experienced difficulties in their relationships with parents over time and development became increasingly cognizant of their parents’ perspectives, less likely to rebel against them, and more likely to turn to them for emotional support. Yet this process is not inevitable. Thus, it is likely that variability will be observed in the character of relationships with parents, particularly for those with problem histories, and who may be struggling to develop a solid footing during this period of emerging adulthood. The current analysis will examine empirically the links between within-individual variability in perceived parental caring/support and self-reported criminal involvement across the period from adolescence to young adulthood.

Changes in Relationships with Peers

In contrast to family relationships, friendships are generally a relationship of choice, and studies indicate that most respondents tend to report positive attitudes toward their friends (Brown and Larson 2009). In light of this, research on peers and crime has tended to focus less on attachment processes than on the antisocial/prosocial characteristics of these companions (McGloin and Thomas 2019). Prior studies have traditionally emphasized that peer influence declines during the young adult period, as individuals become more heavily involved in romantic relationships and/or more focused on other core concerns associated with this period of the life course (e.g., economic considerations, children; Giordano, Cernkovich, and Holland 2003; Monahan, Steinberg, and Cauffman 2009). However, given the extended adolescent-to-adult transition period and delayed entry into marriage, recent research documents that peers often continue to be important in the lives of young adults (Demir and Özdemir 2010; Guan and Fuligni 2016). Nevertheless, research on peer effects on crime in adulthood is somewhat limited and often tied to investigations of the “good marriage effect.” Relying on National Youth Survey data, Warr (1998) early on found that the marriage—desistance link could be explained, in part, by decreases in peer socializing among married individuals. However, there is a need for research relying on a contemporary sample that examines variability in affiliation with antisocial peers among dating, cohabiting and married individuals, and the degree to which changes in these patterns of affiliation are related to changes in offending over time.

It is important to consider changes in the character of relationships with parents and peers within the framework of an investigation that also includes attention to identity changes. As noted above, Bushway and Paternoster (2012) argued that a change in identity is the most critical dynamic underlying crime cessation. Yet the authors suggested further that this type of change will lead the individual to make network changes (i.e., to stay away from friends who have been a negative influence). While we agree in general with this line of argument, our view is that identity change, peer delinquency, and support from parents are all distinct (and variable) phenomena. For example, some individuals who have distanced themselves from problematic identities, and even taken on new ones, may nevertheless retain delinquent friends. The current examination thus contributes beyond prior work by investigating whether these social relationships and identity processes contribute independently to variability in self-reported criminal behavior and providing a framework for understanding how social relationships figure into the change process.

Lingering Effects of Antisocial Family Climate

The analyses outlined above will allow us to investigate time-varying factors that have received less attention within the desistance literature relative to marriage and employment effects, and in particular to determine whether the character of relationships with parents and peers make a difference, net of marriage/employment and identity changes. However, recent findings relying on the current data set suggest that a comprehensive approach to family effects across the transition will include attention to antisocial family climate as well as the supportiveness or warmth of one’s relationships with parents (authors). Results showed that early exposure to parental substance use, violence, and other types of antisocial behavior was significantly related to the child’s own later patterns of similar behavior and indeed explained some of the effects of the experience of parental incarceration. Thus, it is important to consider the longer term effects of these family experiences within the context of the current study.

This nexus of family characteristics/background is hypothesized to have long-term or lingering effects for several reasons. First, a family climate that includes these experiences may have a direct impact, not only on initial forays into delinquent territory but also on the odds of continuing criminal behavior patterns. This is consistent with a social learning approach to the phenomena of intergenerational transmission but highlights that these dynamics may continue as the child matures into adulthood (Giordano 2010). Second, these experiences and the family instability and discord that often connect to them decrease the odds the child will be successful in areas that traditionally support crime cessation (i.e., academic success, full-time employment, relationship stability; Amato and Patterson 2017; Fomby and Cherlin 2007; Settersten, Furstenberg, and Rumbaut 2005). Third, antisocial family background may play a role as it “channels” and constrains the child’s own social network experiences, which may heighten the risk of continued involvement with delinquent peers. Finally, parents whose lives have included significant problems with substance abuse or other difficulties may not be in a position, relative to parents without these histories, to provide various forms of support to children during this transitional period. Although a comprehensive assessment of all of these pathways is beyond the scope of the current investigation, it is important to account for antisocial family background as a potentially significant factor that affects the patterns of self-reported crime that are observed across the study period.

The Current Study

The primary objective of the current analyses is to investigate the role of parents and peers as influences on individuals’ self-reported criminal behavior across the transition to adulthood, recognizing that transition events and subjective changes such as shifts in identity are also important to an understanding of behavior change. We rely on indices of parental attachment and measures of involvement with delinquent peers that were included at all waves of the study (respondents span ages 13–28 across the study period). Models also include two measures of identity that were included in the protocols at each wave. Recognizing that criminal behavior does not necessarily foster an overarching view of self as “criminal,” we focus on identities of “partier” and “troublemaker” that often connect to patterns of criminal behavior. In addition to assessing whether respondents’ distance themselves from such negative or problem identities, models include an index tapping a more positive identity. Positive identities can potentially center on many different content areas. Thus, we rely on disagreement with a general item tapping pride in one’s accomplishments. These multilevel models take into account relationship status (married, cohabiting, and single/dating), as well as whether respondents during this transitional period are either employed full time or completing their education (the variable is labeled “gainful activity”). Models also include a composite index of parent/family antisocial behavior, and controls for relevant sociodemographic, and other family background characteristics.

As a supplement to the quantitative analyses, we draw on in-depth interview data to illustrate key points. These data lend additional support to the quantitative results but provide a window on how these dimensions of change may fit together in the life-course experiences of particular individuals. The narrative accounts also suggest limitations of the quantitative analyses and thus highlight areas that may be fruitful as lines of additional research.

Data and Methods

This investigation drew on five waves of longitudinal data from the TARS, a stratified random sample of 1,321 adolescents and their parents/guardians. The TARS data were collected in the years 2001 (wave 1), 2002 (wave 2), 2004 (wave 3), 2006 (wave 4), and 2011 (wave 5). The sampling frame of the TARS study was devised by the National Opinion Research Center and encompassed 62 schools across seven school districts. The initial sample was drawn from enrollment records for 7th, 9th, and 11th grades, but school attendance was not a requirement for inclusion in the study. Interviews were conducted in-person, and data were entered directly into a laptop computer, first by the interviewer, and subsequently by respondents, who entered all personal responses directly. At the first wave, parents were also administered a paper and pencil questionnaire. The stratified random sample includes oversamples of African American and Hispanic adolescents. At the wave 5 interview, 1,021 respondents were retained (77 percent of the wave 1 sample).² Furthermore, attrition analyses revealed that subjects retained did not differ significantly on most dimensions, including baseline levels of offending, but were somewhat more likely to be female and to report a stepparent, single-parent, or “other” family structure. Additionally, African American respondents and those reporting low levels of parental education (less than high school) were less likely to be retained.

The current analyses relied on structured interviews conducted at waves 1–5. We excluded respondents who did not identify as African American, White, or Hispanic (n = 26), as well as the youngest (12 years) and oldest (29 years) observations, as there were too few to retain in our analyses (n = 38). We focused on a subsample of respondents who reported some delinquent involvement at the earlier waves of the study (i.e., respondents reporting no delinquency at waves 1–3 were excluded; n = 663), and thus the final analytic sample (n = 594) represented an 11-year accelerated cohort design across five periods with three overlapping cohorts (ages 13–28 years).

In-depth qualitative interviews were conducted with a subsample of approximately 100 TARS respondents at most waves of the study (1, 3, 4, and 5). After wave 1, individuals were chosen for in-depth interviews based on responses to the structured questions indicating some level of involvement in risky behaviors (waves 3 and 4 [the focus was on sexual risk]; wave 5 [previous reports of violent behavior]). Although these respondents were thus not initially chosen due to their criminal behavior profiles, subsequent analyses indicated that they reported more delinquency as adolescents relative to the sample as a whole and were more likely to have garnered arrests as juveniles and adults. The content of the interviews varied depending on specific objectives at each wave, but in connection with the waves 3–5 interviews, respondents were asked specifically about whether use of substances/partying had increased, decreased, or stayed the same since the time of the last interview. This straightforward question was often the basis for opening up the discussion to include a range of other problems, and respondents often brought up criminal involvement and legal difficulties throughout the process of narrating their life histories. Analyses of these data provided an initial basis for developing ideas about the desistance process during young adulthood that we subsequently investigated via a series of multilevel models and relying on the larger sample.

Here, we draw on the qualitative data to illustrate mechanisms of interest, and particularly to highlight some of the individual, social, and structural barriers that may be limiting to the individual’s potential to affect sustained behavior change. The analysis of the qualitative data followed an iterative process and included a period of discussion of cases as they were transcribed, followed by a period of open coding and memo writing. ATLAS-ti (Version 7) software was used to sort, code, and analyze the transcribed interviews. A coding scheme was developed based on the interview guide and summaries and was further refined as coding and analysis progressed (for more details on the general methods used in coding and analyzing these data, see Giordano, Copp, Longmore, and Manning 2015). Although we have analyzed all of the in-depth interviews extensively, in this investigation, we focused particularly on narrative accounts in which the respondent expressly mentioned a desire to change. The specific cases we draw upon illustrate how family and peer relationship experiences provide very different backdrops for moving forward in the direction of these desired changes.

Measures

Dependent variable.

Crime was a time-varying measure based on self-report inventories adapted from the inventory developed by Elliott and Ageton (1980) and was assessed at all five waves of the study. Respondents were asked how often they engaged in the following: “carried a hidden weapon other than a plain pocket knife,” “stolen (or tried to steal) something worth more than $50,” “attacked someone with the idea of seriously hurting him/ her,” “sold drugs,” “broken into a building or vehicle (or tried to break in) to steal something or just to look around,” and “used drugs to get high (not because you were sick)” (responses ranged from 1 “never” to 9 “more than once a day”). We created dichotomous variables indicating whether the respondent had engaged in these activities. Our final measure of crime was a six-item variety score measure, taken as the sum of these six items (α ranges from .65 [wave 5] to .73 [wave 1]).³

Time-varying independent variables.

Identity was measured at each of the five structured interviews, using the following prompt: “The following are ways people would describe each other. To what extent do you agree that other people would describe you as ____?” These reflected appraisals included things such as “flirty,” “funny,” “smart,” and “well liked.” We focused on two of these items: “troublemaker” and “partier” (responses ranged from 1 “strongly disagree” to 5 “strongly agree”); Matsueda 1992). We reverse coded the items, such that higher scores were indicative of movement away from a negative self-view, and our final measure, distancing from negative identity, was taken as the mean of these two items. Positive identity was taken from the following single item: “I feel I do not have much to be proud of” (reverse coded; responses ranged from 1 “strongly disagree” to 5 “strongly agree”).

To assess within-individual variation in association with delinquent peers, we used a six-item variety score measure. This measure, peer crime, was similar to that described above for respondents’ crime but assessed friends’ involvement in the various behaviors/activities (a ranges from .69 [wave 1] to .77 [wave 4]).

Parental closeness was assessed across the five waves of structured interviews using a series of questions in which respondents were asked how much they agreed with the following: “my parents often ask about what I am doing (e.g., in school, at work, with my friends),” “my parents give me the right amount of affection,” “my parents trust me,” “I feel close to my parents,” “I am closer to my parents than most kids my age,” “my parents sometimes put me down in front of other people” (reverse coded), “my parents seem to wish I were a different type of person” (reverse coded), “my parents push me pretty hard to do well,” and “I feel close to my parents.” Responses ranged from 1 “strongly disagree” to 5 “strongly agree,” and parental closeness was taken as the mean (a ranges from .74 [wave 1] to .82 [wave 5]).

We included time-varying indicators of relationship status to identify respondents who report being currently married, cohabiting, and dating/ single at each wave of structured interviews (1 = yes).

Given that there is considerable diversity in employment and education well into young adulthood, we assessed respondents’ involvement in gainful activity at each interview wave, which we defined as enrollment in school or full-time employment (Alvira-Hammond et al. 2014). Because few respondents were employed full-time during waves 1 and 2, we including any employment greater than 10 hr per week. Gainful activity was devised from the following questions: (1) “Do you currently have a job,” (2) “Is this job full time or part time?” and (3) “Did you attend school this past year.” Based on responses to these questions, we created an indicator variable, such that respondents who reported either full-time employment or school attendance were characterized as gainfully active (1 = yes).

Time-stable independent variables.

Parent antisocial lifestyle is measured using a 19-item scale that was developed based on prior analysis of qualitative data (Giordano et al. 2019). The items include the respondent’s retrospective report of family conflict including the accuracy of the following statements: “family members fought a lot,” “family members often criticized one another,” “family members sometimes got so angry that they threw things,” and “family members sometimes hit each other.” Respondents were also asked how often either one of their parents: “threw something at the other,” “pushed, shoved, or grabbed the other,” “slapped the other in the face or head with an open hand,” and “hit the other.” Additionally, during the first interview, parents were asked how often they “threatened to hit your child” and “pushed, grabbed, slapped, or hit your child.” They were also asked whether the following happened during their own teen years: “I was suspended or expelled from school,” “I got (someone) pregnant,” “I was arrested by the police,” “I drank alcohol,” and “I used drugs.” Finally, parents were asked how many times they had done the following in the past year: “used alcohol to get drunk,” “gone out to party with friends,” and “used drugs to get high (not because you were sick).” In addition to these items, we included an indicator of whether or not either parent had ever been incarcerated during the respondent’s lifetime. Items were first standardized, and our measure was taken as the mean across the standardized items (α = .80).

Models also include controls for sociodemographic and family background characteristics at the between-subjects level including gender, race/ethnicity including non-Hispanic White (reference category), non-Hispanic African American, and Hispanic. Family structure is measured using a series of dummy variables including two biological parent family (reference category), single parent, stepparent, and “other” family.

Analytic Strategy

Using a subsample of respondents who reported involvement in some level of delinquent activity as adolescents, we employed a series of hierarchical Poisson models to examine crime across the adolescent to young adult transition. As noted above, the TARS data use an accelerated cohort design in which three overlapping cohorts of respondents were followed over a roughly 11-year period, and interviews were conducted in years 1, 2, 4, 6, and 11. Given the variably spaced waves of data and the differing measurement occasions across participants, we used adolescent age as the metric of time in our analyses. This approach allowed us to estimate trajectories of crime over a longer period of time. Also, because we relied on respondents’ numeric age as the temporal predictor in our models, the spacing of waves does not affect model fit (Singer and Willet 2003).

In our analyses, crime was modeled as a function of age, identity, social network influences, parent antisocial lifestyle, and several controls. Given our interest in the influence of within-individual changes in identity and social network influences on within-individual changes in offending, we employed Allison’s (2005) between-within method, which enables us to use subjects as their own controls. This is accomplished by decomposing the time-varying variables into two parts, capturing within- and between-person variation. By subtracting person-specific means from all of the time-varying variables included in this investigation, the effects of each of our time-varying variables is determined entirely by “the portion of its variance that is independent from the other variables in the model” (i.e., within-individual change; Osgood 2010:380). This approach helps address a number of potential selection effects, as it permits the examination of within-individual change while controlling for sources of unobserved heterogeneity.

We begin by providing descriptive statistics for all study variables (Table 1). Next, we perform a test for equality of coefficients to determine whether the coefficients for the mean and deviation scores are the same. This is similar to the Hausman test and indicates whether the described fixed effects approach is preferred to a straightforward random effects model (Allison 2005). Results of this test indicated that the coefficients for the mean and deviation scores were significantly different, supporting our focus on the fixed-effects coefficients. We presented the results of the multilevel regression models in Table 2. Model 1 included the linear and quadratic effect of age, the centered and mean measures of marriage and gainful activity, and all time-stable predictors—including a measure of parent antisocial lifestyle. Model 2 introduced the between- and within-person measures of identity, as well as the measures of social network influences.

Table 1.

Means/Proportions and Standard Deviations of Analytic Sample by Wave.

Variables	Grand Mean	SD	Range	Wave 1	Wave 2	Wave 3	Wave 4	Wave 5
Dependent variable
Crime	0.943	1.24	0–6	0.924	1.009	1.178	0.879	0.592
Within-subjects
Age	18.95	3.84	13–28	15.35	16.52	18.34	20.50	25.47
Identity
Distancing from negative identity	3.206	0.91	1–5	3.08	2.99	3.08	3.31	3.69
Positive identity	3.897	1.04	1–5	3.88	3.79	3.84	3.96	4.05
Social network influences
Peer delinquency/crime	1.842	1.67	0–6	1.76	1.92	2.12	1.89	1.48
Parental closeness	3.882	0.63	1–5	3.86	3.79	3.91	3.91	3.99
Traditional predictors
Marriage	4.72%			0.00%	0.00%	0.72%	5.56%	16.95%
Cohabitation	12.22%			0.00%	0.00%	9.56%	21.78%	27.38%
Gainful activity	84.31%
Between-subjects
Parent antisocial lifestyle	0.062	0.45	−0.6–1.9
Female	47.22%
Race/ethnicity (White)
African American	25.21%
Hispanic	12.26%
Family structure (bio parents)
Single parent	25.06%
Stepparent	15.00%
Other family	13.29%

Source: Toledo Adolescent Relationships Study.

Note: n = 2,626.

Table 2.

Associations Between Identity Change, Social Network Influences, and Traditional Predictors and Trajectories of Offending.

	Model 1		Model 2
Variables	b (SE)		b (SE)
Fixed effects, composite model
Within-subjects
Initial status	−1.369	2.16	2.995	1.74
Age	0.419***	0.07	0.224***	0.07
Age2	−0.011***	0.00	−0.006***	0.00
Identity
Distancing from negative identity			−0.208***	0.04
Positive identity			−0.068**	0.02
Social network influences
Peer delinquency/crime			0.263***	0.02
Parental closeness			−0.123*	0.05
Traditional predictors
Relationship status (single/dating)
Marriage	−0.485**	0.17	−0.647*	0.17
Cohabitation	−0.065	0.08	0.070	0.08
Gainful activity	0.034	0.07	0.097	0.07
Between-subjects
Parent antisocial lifestyle	0.183*	0.07	−0.040	0.06
Female	−0.426***	0.06	−0.163**	0.05
Race/ethnicity (White)
African American	−0.190*	0.08	−0.024	0.06
Hispanic	0.136	0.10	0.067	0.07
Family structure (bio parents)
Single parent	0.163*	0.08	0.086	0.06
Stepparent	−0.038	0.10	−0.034	0.07
Other family	0.021	0.10	−0.002	0.08
Variance component, intercept
T	0.284 (0.03)		0.068 (0.02)

Source: Toledo Adolescent Relationships Study.

Note: n = 2,626.

^*p <.05.

^**p <.01.

^***p <.001.

Results

Quantitative Results

Descriptive statistics for all study variables (see Table 1) are based on the 2,626 person periods. The average level of crime was .943, which suggests a low level of offending among the majority of respondents. However, 54 percent of observations do reflect some degree of criminal involvement during the study period. The average age of respondents was approximately 19, and this ranged from 13 to 28. The average report of identity was 3.21, which corresponds to a response of “neither agree nor disagree” on respondents’ reflected appraisals as “partiers” and/or “troublemakers.” Reports of positive identity were 3.90, which suggest that respondents, on average, disagreed with the following: “I do not have much to be proud of.” Average peer delinquency was 1.842, suggesting that peer involvement in crime was slightly higher than respondents’ own criminal behavior. Parental closeness was 3.882, which suggests that respondents, on average, reported positive relationships with parents. Nearly 5 percent of observations reported being married, and roughly 12 percent reported cohabiting. Finally, more than four fifths of observations indicated gainful activity. It is important to note that the descriptives are presented in person-period format, and thus the prevalence of events including marriage and gainful activity cannot be directly inferred from these figures. In our sample, 17 percent of respondents reported being married, 41 percent reported cohabiting, and 100 percent of respondents reported gainful activity, at some point during the study period.⁴

Multivariate analyses addressed how respondents’ identity changes, in addition to shifts in the nature of their social relationships (parents and peers), were related to offending patterns (see Table 2). Model 1 included the linear and quadratic temporal predictors, between- and within-person components of our traditional predictors—marriage, cohabitation, and gainful activity—and time-stable factors. Results of this model indicated that the trajectory of crime is curvilinear across the study period, which is consistent with prior research on the age–crime curve (Farrington 1986; Sweeten, Piquero, and Steinberg 2013). That is, involvement in crime increases across the adolescent period and begins to decline during early adulthood. Consistent with prior research and theorizing, marriage was linked to declines in offending. However, transitions to cohabiting unions were not associated with within-individual changes in offending in these data. Furthermore, gainful activity was not associated with crime, controlling for other factors. It is likely that the lack of significance of the gainful activity index reflects the transitional stage of the life course, and more nuanced measures would more accurately capture the extent to which employment and educational considerations influence offending trajectories.

Controlling for time and our traditional predictors, the association between parent antisocial lifestyle and crime was significant and positive. This suggests that respondents exposed to parental antisociality and other related forms of adversity (i.e., parental incarceration) engage in higher levels of offending. In addition, female and African American respondents reported lower levels of delinquent/criminal involvement as compared to their male and White counterparts, while those raised in single-parent households reported higher levels of delinquent/criminal involvement relative to their peers who resided with two biological parents during adolescence.

Model 2 included indicators of identity and respondents’ social network influences. In this model, parental closeness and peer crime were significantly associated with crime, even controlling for shifts in the respondents’ identity over time and traditional predictors of criminal desistance. In particular, association with delinquent peers was linked to increases in crime, while parental closeness was related to declines in crime, across the study period. Furthermore, both components of identity, including distancing from the troublemaker/partier identity and resonance with a more positive identity (as indexed by pride in one’s accomplishments), were negatively associated with offending, controlling for other factors. This suggests that, net of time, changes in endorsement of the negative identity and the adoption of a positive identity correspond to declines in crime over time.

In addition, although the magnitude of the marriage coefficient declines following the addition of identity and social network factors to the model, the association between marriage and crime remains significant and negative in this full model (p < .05). In contrast, the link between parent antisocial lifestyle and crime was attenuated, and this was largely driven by the addition of peer delinquency/crime to the model. Finally, controlling for the full roster of study variables, the female coefficient remained significant and negative, suggesting that female respondents were less likely to engage in offending across the study period.⁵

In supplemental analyses, we considered the direct and indirect pathways between social network influences and crime using tests for multilevel mediation with bootstrapping.⁶ Results suggest significant direct linkages between the peer and parent factors and crime but also reveal a significant indirect pathway via identity processes. That is, mediation analyses focused on peer delinquency identified significant direct (.355, p < .001) and indirect (.030, p < .001) associations with crime, with similar findings for associations between parental closeness and crime (direct and indirect associations were −.264 and −.047, respectively; p < .001).⁷

Consistent with prior research, the findings reviewed above indicate that marriage is generally associated with reduced crime. Further, analyses document an impact of identity changes. Yet the other social factors assessed contributed further to variability in self-reported crime as observed from adolescence into the young adult years. In-depth interviews completed with a subset of the TARS respondents accord well with these results but illustrate how this broader set of factors may vary or coalesce in the life experiences of particular individuals. Below, we consider several respondents whose narratives differ in the social resources that may support or inhibit the process of making behavioral changes during this transitional period of the life course. We focus on three respondents who all express a desire to change. Yet the life history narratives we elicited provide a window on the broader social contexts within which these individual-level changes are taking place. These descriptions suggest, then, that as a stand-alone dynamic, identity shifts, and other types of cognitive transformations appear to be necessary but at times insufficient scaffolding for sustaining a pattern of sustained desistance.

“Brand New Man”

Justin is a 23-year-old respondent who reported high levels of delinquency during adolescence, criminal justice system involvement as he entered his 20s, and more recent decreases in levels of self-reported crime. This respondent is currently married and employed as part of a maintenance crew at a big-box store, and his open-ended interview responses indicate a determination to continue in a positive direction. Thus, Justin’s narrative lines up in important respects with the previous focus on life-course transitions as well as with perspectives highlighting the role of identity changes. For example, in addition to recent raises at his job, Justin expressed pride that he no longer uses marijuana or alcohol, or gets in fights, and noted recent improvements in his relationship with his current wife. Justin’s discussion also incorporated a strong desire to distance himself from a range of negative identities:

I’m not like the drunks that was in there [jail]. Just go over to the neighborhood I grew up at and see all the people that you looked up to when you was little, they’re all smoking crack. All drunks. You know, let me get a quarter, let me get this, let me get that. I didn’t want to be that, you know.

Nobody’s gonna categorize me with these deadbeats [referring to men who do not pay child support] I’m not …. Because I had a rough childhood … we never had nothing. We were struggling. I don’t want my daughter to have the same thing.

Successful desisters’ frequent comments reflecting a desire to discard or avoid negative identities support the quantitative results, and in particular Paternoster and Bushway’s (2009) focus on this important aspect of identity change. Paternoster and Bushway and other researchers have suggested that in addition to moving away from negative identities, the individual will ideally be able to envision at least the “broad outlines of a satisfying conventional replacement self” (Giordano et al. 2002:1056; see also Paternoster and Bushway 2009). A key aspect of our argument, however, is that within the contemporary social and economic landscape, and given the disadvantages offenders often accumulate, the shift to the new identity is not a given, once the old one has been discarded. The desire to change may be sincere, but the new self may not be (a) automatic, easy to obtain, or to sustain and (b) often requires the buy-in of others who are integral to its successful enactment. Thus, while Justin focused heavily throughout the interview on his desire to be a good father, he did not live with his child and had not seen her in over a year:

Yea I don’t get to see her too much because we’re going through a custody battle, with, in Tennessee. And I don’t get along with the baby mama. And I don’t get along with the parents [her parents have custody]. Yea, they’re all dramatized people …. They live in a $300,000 house, bought with cash-drive around in Beamers and this and that but you know you can’t buy love. Right now … we’re waiting for them to get a court hearing. I’m going for custody.

The sometimes difficult, processual aspects of acquiring the new positive identity, and the role of others’ views become clearer as this respondent fills in details about his past. Justin initially attempted to avoid paying child support (“I didn’t for a little while. You know I ducked and dodged”), and his previous heavy drug and alcohol use, and patterns of aggressive behavior (“and then when I got in that fight down here at the trailer park I got that assault charge”) are potential factors in the grandparents’ current view that he should not have custody or periods of visitation. In this instance, then, the negative reflected appraisals (Sullivan 1953) of others present hurdles, even though Justin has recently changed his behavior and expresses a sincere desire to tackle this important aspect of his adult identity.

Further examination of the lengthy narrative also reveals considerable problems in Justin’s romantic relationships. In addition to the difficulties linked to his previous partner and her family, his current marriage is only recently improving (“It was like, okay she went and stayed at her parents’ house, I stayed here. We had our problems. It was kind of rough”). Thus, it does not appear that a stable romantic relationship accounts fully for the behavior changes Justin has made. It would be efficient to focus primarily on his own personal motivation and change in attitude. Yet support from the family of origin during this time of considerable uncertainty potentially adds to our understanding of this respondent’s choices and ability to sustain moves in a more prosocial direction:

I rely on them for strength, and you know not so much stability like money-wise or anything like that, but just you know help me and push me and stuff like that. And then I do the same for them when they have a problem. But I just depend on them more like we all need somebody to talk to or need somebody to be there for me, you know—they’re there. Just like when they’re having problems and they need somebody to talk to I’m there to talk to them. And then when the times get tough you know we’re there for each other. [Describes how he helped his father out financially when he was off work due to surgery]. It’s a give and take. He gave, I took it, and then when he needed it, I gave to him and he took it, you know.

Justin’s narrative highlights an even more fundamental role of these lifelong associations, as members of his family offered positive reflected appraisals about his future potential. These can be seen as a counterweight to the negative views of others who continued to focus on his problem behavior history (e.g., his ex and her family):

Both of them, my sisters I asked them how they feel about me. Well Buddy we love you, we want you to do better. We know you ain’t doing your potential, you know and I’m like, well if I’ve got two little sisters that know I can do better and they’re young kids, and they know I can do better, then let me prove to them that I can. You know, like I talked to my mom, my mom be like, boy you’re getting down. I know you can do better. I’ve seen you better than this. You know, and my dad, son I love you with all my heart I think you can do better. You know and then it’s like okay I got all these people that are willing to help me and give me this, you know this strongness, give me this push I might as well take the chance. And I did and look at me now. Eight months later, shiiiit. Brand new man.

The quotes above highlight the role of these family members not only as sources of support but in the redefinition process itself. Justin’s comments about friendships also reflect an effective coalescing of individual changes in perspective and shifts in the character of his social networks. This respondent recognized that maintaining a more prosocial lifestyle required limiting contact with previous companions, and his employment at the big-box store opened up opportunities for changing his patterns of affiliation:

I, I just usually kick it with people older than me, not my age, because they’re just so immature. You know they’re 23 years old and they’re all oh, I’m still young, I don’t have to settle down, this and that. So I think with older people they’re more on my level. They see eye to eye with me …. Like I’ve got a kid, you know a three year old daughter, so I’ve gotta be the man now, not wait until I’m 30, you know …. I don’t know.

In this instance, Justin made agentic moves to align his emerging sense of self with others who are further along in the “settling down” process. Yet viewed from a social learning perspective, this respondent is likely to be influenced further by his involvement with others who are fully enmeshed in an identity he is attempting to achieve (Emirbayer and Mische 1998; Giordano et al. 2002; Sutherland 1939). Consistent with this generally positive picture, searches of official records indicate that across the seven-year period after the interview, this respondent has had no further contact with the criminal justice system.

“I’m a Mom Now, I’m Not Just Michelle”

Michelle, age 28, described herself as a “bad teen” and “that bad horrible off the hook girl”—descriptions that line up well with the “troublemaker” identity she reported on the wave 1 structured interview, early brushes with the system Driving under the influence (DUI, thefts), and initially high self-reported delinquency scores. This respondent’s narrative also included a description of a troubled family life during childhood and adolescence, and about a year spent in foster care. Subsequent responses on the structured interviews indicate declines in self-reported crime relative to her wave 1 score, lower scores on the “negative” identity index and increasing endorsement of the positive identity. This respondent also developed a strong change narrative, and as the above quote suggests, emphasized her responsibilities as a mother. She also discussed her work as a nurse’s assistant and plan to eventually obtain her nursing degree. Further, Michelle spoke positively of her relationship with Jacob, who was described as “the complete opposite” of her former (abusive) partner Dion, with whom she had cohabited for over five years. In spite of the extreme abuse that characterized her recent past, Michelle is thus proud of her accomplishments to date and ability to continue on a positive path:

I’m a trooper. I always make it happen … like I have a thing now that I tell myself every year. I have to do one thing better than what I did last year. Every year I have to upgrade and I have, every year I went from, I left Dion …. I left everything. And I did it all over, and every year I got a better apartment, and then a decent apartment, and then I had a bigger duplex and then I went from that to a small house.

Michelle’s adult circumstances and positive, efficacious attitude as expressed at the time of the wave 5 interview are factors that are generally associated with more positive outcomes. Yet we note that this respondent is not as well positioned as some other respondents (e.g., Justin, quoted above) in terms of family support and involvement with prosocial peers. Michelle’s narrative highlights that while adult romantic relationships generally develop as major preoccupations during this phase of the life course, the family often continues to play a role as young adults experience challenges and affect various transitions. Michelle indicated that her initial decision to move in with her boyfriend Dion was in part related to problems in her family life (see Hagan and Foster 2001). However, a key observation is that, for many individuals, the family does not recede into the background as a frozen in time early risk factor. For example, when her relationship with Dion became violent, Michelle was reluctant to return home (“That’s not somebody, that’s not somebody [you want to live with], like I don’t wanna live with my dad and like I didn’t have nobody else”). Her discussion of life with her partner Dion also provided a window on the nature of Michelle’s peer affiliations:

And he kept like doing it (throwing hot water on her) and I was standing in the snow … and so I just started walking and I had to spend the night at the dope man’s house. So that I had somewhere to go. And like my friend Caroline at the time lived there, she gave me like some clothes and stuff to change clothes and I just slept the night there and the next day like I went back.

This respondent’s descriptions of violent altercations with Dion point to another through line connecting early family life and these more recent circumstances. When asked why she fought back when Dion became abusive (e.g., she recounted how, during a particular fight, she hit him with an iron and knocked him out), Michelle said “I’ve always been fighting; I fought my dad since I was like 15, 16.” This effect of early, and at times, continuing socialization on the part of the family is also suggested by Michelle’s responses on the structured surveys. Her responses indicate that even as an adult (e.g., as reported at age 27 in connection with the wave 5 interview), her parents had hit her, and in turn, she reported yelling at, shoving, and hitting her parents.

Although wave 5 is the most recent round of interviews, a search of official records up to the present date fills in some details about Michelle’s long-term patterns of involvement in illegal activity. After the birth of her child—and consistent with the emphases in her narrative—no offenses were recorded for a period of about four years. More recently, however, Michelle was arrested on drug charges (in 2016 and again in 2017) and in 2018 for domestic violence. Details included in the most recent arrest warrant include allegations that Michelle had attacked her partner with a knife, strangled him, and pulled out his hair. Also of note, the victim was not the partner Michelle had described as a positive influence at the time of the wave 5 qualitative interview.

Grandma’s Boy

Anthony, age 25, spent most of his childhood and adolescence in the custody of his grandmother, and his narrative developed a contrast between his own home life and that of his siblings:

They didn’t, they wasn’t raised by my grandmother, they was raised by my mom. They lived a different lifestyle than I do, you know they sell drugs and stuff like that. Me I’m a grandma’s boy …. Yea my brothers are off the hook, though, especially Drew. (describes Drew’s violent behavior)

The portrait Anthony develops is backed up by an official records search indicating that Anthony’s mother had several arrests and periods of incarceration, all drug related, that occurred when this respondent was 4, 5, 14, and 15. Several of Anthony’s siblings had also garnered extensive arrest histories. It is potentially important to note as well that while Anthony considered himself fortunate to have been brought up by his grandmother, he nevertheless maintained frequent contact with his mother and all of his eight siblings. For example, when asked about his current friendships as a young adult, Anthony replied, “to be honest my family is my friends.”

This respondent also had frequent contact with his biological father, and at the time of the interview, the two were incarcerated in the same facility for a burglary they had carried out together. Anthony’s life history narrative thus well illustrates the idea of a long reach of “antisocial family climate” that continued to have effects as he has matured into adulthood. This respondent described how his father had asked if he wanted to “work a job” with him. He told the interviewer he believed that this was legitimate plumbing work, but instead the job involved stealing copper pipe from a construction site. Anthony’s background and current circumstances reflect an almost complete enmeshment in a family environment that is likely to make his desistance efforts more difficult to launch and/or to sustain. And while his grandmother might have been expected to provide support and to be a positive influence once he is released from prison, Anthony found out that his grandmother had died during the time in which he was incarcerated:

My family, my mom and them called up here and told the captain …. I fell apart. The most thing I was scared of was her dying while I was in prison. I asked could I go to the funeral … but they denied me … cause I have an escape on my record.

Anthony was scheduled to be released nine days after the interview, and as with all respondents, the interviewer asked about how he envisioned his future in the coming years:

I see it being pretty good. [I: Tell me what that means.] Well, I’m gonna have my own house. I’m gonna have me a way better job than what I have. Uh, just successful. Like I’m tired of this. I’m tired of coming here, I really am, like I should have learned from my first one. There gotta be an end to it and this is it. Just a better person, ya know. Staying out of trouble.

Although Anthony expressed a desire to change his behavior and start a new life, a records search of the years after his release show a new burglary conviction and subsequent two-year sentence that occurred approximately a year after this interview took place. The arrest warrant indicates further that this respondent had apparently not achieved his goal of living in his own home; at the time of his arrest, Anthony was living with his father. This example is also suggestive of the family’s role in channeling network affiliations: details in the record indicate that Anthony had committed the burglary with an individual who lived next door to his father. Further, during the wave 5 interview, Anthony had expressed negative attitudes toward men who hit women (reflecting on the behavior of his brothers who were described as highly abusive toward their partners). Thus, it is striking to note that this respondent was eventually arrested on two separate domestic violence charges in 2016 and again in 2017.

Discussion

Troubled relationships with parents and involvement with delinquent peers are generally acknowledged as significant predictors of delinquency, and researchers have also frequently observed “the concentration of offending in families” (Farrington, Barnes, and Lambert 1996). Nevertheless, these important parent and peer risk factors have largely been ignored in theorizing about and research on desistance processes. Instead, the focus has been on transition events, and more recently, on individual-level subjective changes (e.g., shifts in identity). Drawing on a large, diverse sample of respondents interviewed in adolescence and across the transition to adulthood, we found a positive association between marriage and crime, consistent with prior work, and findings also indicate that identity changes were linked to variability in self-reported crime. A key objective of the study, however, was to determine whether parent and peer relationships contributed to an understanding of variability in the patterns of self-reported criminal involvement, once these transition events and subjective changes had been taken into account.

Results showed that increases in parental support/closeness and decreased involvement with delinquent peers were significantly associated with observed declines in self-reported crime across the study period, net of marriage and employment status, and the identity changes. Reductions in the associations between parents, peers, and crime upon introduction of the identity variables provide suggestive evidence that some of the consequences of these network characteristics may be indirect, via their influence on continuity and change in identity. Analyses also indicated a significant impact of early antisocial family climate, although again, this was no longer significant once levels of peer delinquency were introduced into the models. The latter finding provides evidence that one of the long-term consequences of growing up in a family characterized by violence, drug use, and system involvement may be the ways in which these family experiences channel and constrain the individual’s later social ties (i.e., increase the risk of affiliation with delinquent peers).

In-depth interviews conducted with a subset of respondents provided a window on how these social and individual factors vary in the life experiences of particular respondents in ways that support or impede the desistance process. We focused on three individuals who had all developed change narratives but who differed in the nature of their family and peer relationships. The life history narratives support the idea that a change in perspective is a key driver of the desistance process, and analyses further suggest that this often encompasses a desire to distance from negative identities and to move toward a more prosocial lifestyle. Yet the narratives also illustrate that social relationships make a difference early on and over the long haul, potentially increasing or decreasing the likelihood of affecting a pattern of sustained behavior change. The theoretical emphasis and findings are consistent with the more general notion that human development occurs in large measure via social relationships (Sullivan 1953). Our view is that this perspective also provides a useful backdrop for understanding the process of making these types of life changes.

Consideration of relationships with parents and peers may be factored into future research examining the subjective processes that undergird desistance and studies focused on other relationships such as marriage. This will allow us to develop a more comprehensive picture of the full complement of and relative salience of individual and social factors that may be associated with sustained behavior change. Several limitations of the current study suggest areas that could be pursued in future quantitative and qualitative research on desistance. First, the measures of identity were limited and clearly did not tap all of the domains around which change can occur. One complication in working to craft more adequate measures is that certain negative identities may well have applicability for many offenders (since the “troublemaking” itself may become a master status; Lemert 1967). However, positive, hoped for identities appeared to vary considerably. This is consistent with Giordano et al.’s (2002) discussion of the idea of different “hooks for change,” but it would be useful to develop structured indices that reflect a range of these positive identity hooks. A comprehensive treatment of individual-level changes would also benefit from additional exploration of the idea that some types of cognitive transformations that serve as underpinnings of desistance may involve areas that do not connect directly to major adult roles such as spouse or parent (see, e.g., Giordano et al. 2007; Mulvey et al. 2010).

Second, our measures of social relationships, identities, and offending were contemporaneous, and thus we cannot establish the temporal ordering of these factors within waves. In light of prior research in the developmental and desistance literatures, we do interpret the direction of effects as social network and identity influences on offending, and not vice versa. Given the lack of attention to social forces in the desistance literature, and the potential for the consequences of social relationships (and identity shifts) for crime to be rather immediate, we suggest that the approach employed in these analyses marks an important first step in establishing these linkages. However, future research should further unpack the nature of these associations using other methodological approaches including lagged models.

Another limitation is that in contrast to the other focal variables assessed, the measure of antisocial family context was not time varying and was only measured at wave 1. Thus, repeated measures of antisocial family climate across the full study period would provide a more adequate portrait of the nature and impact of these aspects of family background. Research is also needed that examines more explicit linkages between family factors (e.g., examining how antisocial family climate affects the family’s support function during the transition period). Further, while our intention was to foreground family and peer influences on desistance during this phase of the life course, clearly romantic partners become increasingly salient across the period. Thus, research is needed on the specific mechanisms through which romantic partners (including but not limited to marital partners), influence identity, and other types of cognitive transformations, as well as potentially offering the kinds of emotional and tangible support that increase the odds of successful desistance. Person-centered strategies that take into account variability in the individual’s full portfolio of social resources and liabilities would contribute beyond the variable-centered approach we adopted in the current investigation.

It would also be useful to explore some of these areas through the use of qualitative methods. For example, we found that affiliation with delinquent peers contributed to an understanding of variation in self-reported crime, net of the identity changes that are considered central to the desistance process. Thus, research is needed that explores further the observed variability in the nature of social ties during this phase of the life course. For example, about 25 percent of the individuals who changed their identities over the period did not “knife off” delinquent friends. Thus, it is important to understand how some individuals with problem histories have been able to establish new, more prosocial friendships, while others continue their involvement with companions they fully realize are not a good influence. As Harding, Morenoff, and Wyse (2019) noted, in a description of results of a study of the process of offender reintegration, others have not developed such ties but instead have adapted a pattern of isolation as a different strategy to avoid potentially negative influences.

Recognizing the above study limitations, the findings nevertheless have potential implications for the design of more effective programmatic efforts. In moving beyond the focus on key transition events, including “the good marriage effect,” researchers initially developed a view of desistance as a rather personal journey. This conception fits well with the criminal justice system and other self-help programs’ emphasis on taking personal responsibility. While the individual’s own role and use of agency are integral to the process, wide variations exist in the character of social ties that may facilitate or impede the individual’s ability to change direction. For example, the idea of cognitive restructuring (e.g., as exemplified in programs such as “thinking for a change”; Golden, Gatchel, and Cahill 2006) fits with a more individualistic conception of the process. Yet the results of this investigation highlight that net of identity transformations, the character of peer and family relationships “matter” in predicting the odds of desistance. Programs are needed that incorporate issues related to family and peer relationships and how these align with or inhibit efforts to change. For example, risk and needs assessments that are used with prison populations frequently collect information on inmates’ family and peer relationships. Yet correctional programming may not fully incorporate this information into the programmatic emphases or does so in a limited way. It would thus be important to know not only whether or not an individual has “social support” but to address the character of that support across the full network of affiliations, including whether these others are themselves generally prosocial in their orientations.

Biography

Jennifer E. Copp is an assistant professor in the College of Criminology and Criminal Justice at Florida State University. Her research examines the influence of intimate and familial relationships on developmental outcomes across a variety of domains, including crime and other problem behaviors. Her recent work considers how incarceration, and other forms of system contact, influence future behavior, family life, and child well-being.

Peggy C. Giordano is a distinguished research professor of sociology at Bowling Green State University. Her research centers on basic social network processes, and the ways in which close relationships influence delinquency and criminal behavior over the life course. She is especially interested in family dynamics associated with the intergenerational transmission of crime and other problem outcomes, as well as the impact of extra-familial influences such as peers and romantic partners.

Monica A. Longmore is a professor of sociology at Bowling Green State University. Her interests include parenting and family dynamics, social psychological processes, and ways in which these family circumstances and subjective understandings affect adolescent and young adult behavior and well-being. Her research also focuses on precursors and consequences of intimate partner violence.

Wendy D. Manning is a distinguished research professor of sociology at Bowling Green State University, director of the Center for Family and Demographic Research, and co-director of the National Center for Family and Marriage Research. Her research focuses on relationships that exist outside the boundaries of marriage, including cohabitation, and effects of varied family forms on relationship dynamics, stability, and parenting practices.