Stopping Out and Going Back: The Impact of Educational Attainment on Criminal Desistance Among Stopped-Out Offenders

Abstract

Education has been consistently studied as a source of crime prevention and control, but the relevance of returning and completing educational degrees among offenders who drop out, as an opportunity to further the process of desistance, has not received empirical attention. The current study addresses this gap in desistance research by examining the impact of educational return and specific degree attainment on desistance from crime using data from the National Longitudinal Survey of Youth 1997. Results indicate that reenrolling in educational pursuits can produce partial desistance effects as does specific degree attainment. The findings suggest a reconsideration of education as both a source of prevention and desistance and expands theoretical and practical discussion of desistance through educational pursuits.

Keywords

Education Life course criminology Desistance Stopping out

Desistance from crime has become a critical focus within theoretical and practical criminological research (Bersani & Doherty, 2018; Rocque, 2017). The process of reducing, and ultimately ceasing, criminal behavior can reflect stable long-term changes, such as identity transformation (Paternoster & Bushway, 2009), or short-term situational experiences, such as brief marriages (Bersani & Doherty, 2013). Desistance has been linked to specific social developments, including marriage, employment, and parenthood (Kerr, Capaldi, Owen, Wiesner, & Pears, 2011; Laub & Sampson, 2003), but has not yet been addressed in the context of educational reengagement.

While research efforts continue to elaborate on short- and long-term influences, and unpack exogenous and endogenous experiences, studies have not sufficiently incorporated concepts typically considered as crime prevention opportunities into elements that may facilitate desistance. Educational attainment, in criminological research, is largely conceptualized as an aspect of crime prevention (Hansen, 2003; Lochner, 2004; Machin, Marie, & Vujic, 2011), with the majority of policy efforts directed toward retaining at-risk students in adolescence to complete their degrees (Taheri & Welsh, 2016). While preventing school dropout is important, for social advancement and crime reduction, the fact remains that many individuals prematurely drop out of school (McFarland, Cui, Rathbun, & Holmes, 2018), which can have an impact on future likelihood of crime (Sweeten, Bushway, & Paternoster, 2009). This future likelihood may be mitigated by a return to, and completion of, formal education, but this question has not been examined in current desistance research. The current study addresses this question by examining dropout offenders within the National Longitudinal Survey of Youth 1997 (NLSY97), some of whom reenroll in educational pursuits and earn degrees, to assess whether those experiences influence subsequent criminal behavior.

Literature Review

Desistance

Desistance from crime involves the reduction, and ultimate cessation, of offending for those with criminal histories and has become a significant aspect of criminological research in recent years, spurring substantial debate over theoretical reasons to desist, unpacking internal motivations for change as well as external social experiences, and opening opportunity for significant policy changes to reduce crime (Bersani & Doherty, 2018; Bersani & Schellen, 2014; Giordano, Schroeder, & Cernkovich, 2007; Laub & Sampson, 2001; Paternoster & Bushway, 2009; Rocque, 2017; Stevens, 2012). In addition to ongoing scrutiny of internal and external actions and interactions facilitating desistance, debate continues over situational influences, such as the dissolution of a marriage (Bersani & Doherty, 2013), versus long-term stable changes, including cognitive transformations (Giordano, Cernkovich, & Rudolph, 2002), and changes in identity (Paternoster & Bushway, 2009).

Empirical support has consistently demonstrated desistance occurs as an ongoing process for offenders (Bersani & Doherty, 2018; Rocque, 2017; Sampson & Laub, 2003), but the consideration of short- versus long-term influences and exogenous and endogenous variables affecting change warrants additional attention in the context of educational experiences for offenders. Reenrolling in school, as an immediate short-term change for offenders who prematurely left, can establish new relationships with teachers who may supervise individual offenders as well as provide a change to one’s routine activities. Reenrolling and completing specific degrees can also produce long-term change through rebuilding specific social bonds to educational institutions, improving one’s knowledge base and comprehension of responsibility and mature judgment, and facilitate additional social bonds through career advancement (Vilorio, 2016) or marriage eligibility (Musick, Brand, & Davis, 2012). Desistance is a pervasive phenomenon and is largely considered normative, but the fact remains that significant variations in desistance occur by individual offenders (Bersani & Doherty, 2018), and the process of change can be enhanced by a number of social developments.

Before addressing theoretical developments in desistance research, and the potential contributions that educational involvement may provide, it is necessary to address the structural elements involved in the process of desistance. Desistance from crime has developed within life course criminology, which includes four core principles meant to explain different manifestations and developments of crime, including onset of offending behavior, escalation in offense trajectories through increased frequency or severity of behavior, persistence in offense trajectories, and transitions or turning points, such as decreased frequency through developments in romantic relationships or the tangible realization of a need for change, that facilitate desistance. The principles include acknowledging the relevance of historic time and place, the timing of events within lives, the importance of linked lives, and agency (Elder, 1998). These principles reflect the importance of studying crime in specific historic contexts (Bersani, Laub, & Nieuwbeerta, 2009), at distinct periods within an individual’s life (Moffitt, 1993), as influenced by relationships with others (Laub & Sampson, 2003), and through the incorporation of personal outlook and conceptualizations of choice and control (Giordano et al., 2002).

These principles are important in the current context as they provide guidance and emphasis on the complex interplay between education and offending. Within historic context, it is important to address that status dropout rates, defined as individuals 16 to 24 years old who do not hold high school degrees, declined from 1960 to 2016, but the problem of premature departure persists (McFarland et al., 2018). The timing of events is also relevant, given that adolescence is the developmental period where one is most likely to engage in risk-taking behavior, including dropping out and offending (Gottfredson & Hirschi, 1990). The desistance effect of linked lives is clearly articulated by research into marriage and employment (Laub & Sampson, 2003), but the likelihood of these social developments occurring at all can be increased by educational achievement (Musick et al., 2012; Vilorio, 2016), just as a spouse or employer can encourage educational pursuits. The concept of agency is of critical importance to understanding the causes of desistance, and while formal educational institutions are not likely the causal factors of change, the decision to return and complete a missing degree creates opportunities to address theoretical motivations to return and the possible influence education may have on social advancement.

Theoretical insights into desistance

Recent decades have seen a substantial rise in discussion, debate, and argument over the causes and correlates of desistance. Early efforts, specifically the theory of informal social controls, illustrated the importance of social experiences, such as marriage, employment, and military service, that created specific controls over an offender in adulthood (Laub & Sampson, 2003). While this research provided significant insight into opportunities for behavioral change, the assumption that offenders are passive actors subject to external controls has been challenged by more recent arguments. Theoretical arguments began to incorporate the perspectives, attitudes, and opinions of offenders as more active, agentic players facilitating their own desistance, including utilizing one’s past to create positive change (Maruna, 2001), or developing a sequence of cognitive transformations wherein the offender becomes aware of the need for change and actively pursues prosocial opportunities (Giordano et al., 2002).

The relevance of cognitive perceptions and transformations was further exemplified by the incorporation of emotional experiences and processes that influence cognitive abilities among offenders and can further behavioral change (Giordano et al., 2007). Paternoster and Bushway (2009) emphasized the active role individual offenders take in the process of desistance by focusing on evolving identities and specifically addressing the emotional importance of self-perception that instills a desire for change. These studies do not provide a comprehensive review of all theories applicable to desistance research, but they do illustrate the complex nature of behavioral change as well as the importance of addressing both external social experiences or controls that may facilitate desistance and internal processes motivating change.

Theoretical integration in desistance research is now on the rise, given the identified relevance of evolution in identity, the importance of rationality in cognitive transformations, arguments regarding age and maturity, and the elusive concept of agency in facilitating change (Bersani & Doherty, 2018). The influence of formal education on desistance from crime has not been examined, given the primary focus on education as a source of crime prevention (Reynolds, Temple, & Ou, 2010), but theoretical arguments would suggest that education positively influences desistance through both external social experiences, such as expanding employment opportunities (Musick et al., 2012), and internal developments, such as the possibility that educational efforts contribute to changes in cognitive processing (Giordano et al., 2002) or evolution in identity (Paternoster & Bushway, 2009).

The Importance of Education in Criminology

Within criminology, education is typically associated with structurally focused theoretical arguments. Existing research primarily focuses on education in adolescence to control antisocial behavior (Taheri & Welsh, 2016), and there is strong evidence for the preventive effect of educational attainment on crime in adulthood (Machin et al., 2011). However, educational pursuits serve as part of social developments and processes that offenders experience, even among those who drop out of school, and it is possible that reengaging with this process may contribute to other prosocial behaviors.

Education serves to prevent and control crime

The role of education in curtailing crime has consistently been viewed, both theoretically and empirically, as one of prevention (Machin et al., 2011). Existing research has largely fixated on the benefits of staying in, and completing, formal education through preventive efforts in adolescence (Taheri & Welsh, 2016). However, education does not universally prevent all crime (Groot & van den Brink, 2010), and individuals at risk of persistent offending are also at risk of dropping out of formal education prematurely (Hjalmarsson, 2008; Stearns & Glennie, 2006; Sweeten et al., 2009).

Research examining the interplay between education and crime focuses on adolescence, as it is a legally required period of educational involvement. It is also the age range where one is at highest risk of engaging in offending behavior (Gottfredson & Hirschi, 1990), and existing efforts aimed at educational expansion and enrichment, specifically in adolescence, have demonstrated significant effects on reducing criminal involvement (Taheri & Welsh, 2016). Graduating from high school has been shown to significantly reduce crime among male adolescents (Lochner, 1999), and results from a longitudinal assessment in Great Britain identified a specific impact of education on the likelihood of engaging in property crimes. Hansen (2003) examined age–crime curve patterns for different types of offending behavior between two groups, those who had dropped out of formal education prematurely and those who remained, and noted that the rate of decline for the educated group surpassed that of the dropout group. The results revealed a similar pattern for violent crime rates, with the decline in behavior occurring more rapidly for individuals who remained in school compared with those who dropped out early. As a further detail demonstrating the importance of education, this study identified a stable rate of property offenses for the dropout group beyond age 22, suggesting long-term significance to the impact of education, or lack thereof, on crime.

Additional research has elaborated on the relationship between formal educational attainment and specific types of offending behavior. An examination of official crime rates in Great Britain, in the context of changing compulsory education laws, suggested that education may have a causal impact on property crime rates but established no clear pattern of influence on violent crime rates (Machin et al., 2011). Cross-sectional data from self-reported interviews in the Netherlands revealed that more years spent in school negatively affected the likelihood of engagement in shoplifting, vandalism, and tax fraud (Groot & van den Brink, 2010).

Education may help facilitate desistance

Educational involvement is one aspect of social processes that adolescents experience and, as previously indicated, primarily serves as a protective influence against criminal behavior. However, this process can be interrupted if a student drops out before graduation. The reasons for dropping out early may be complex and multifaceted, incorporating prosocial opportunities and criminological risks (Jordan, Lara, & McPartland, 1996; McNeal, 1997; Stearns & Glennie, 2006). In addition, both dropping out and crime can be influenced by similar experiences, such as childhood homelessness (Aratani & Cooper, 2015), or other hardships, such as receiving public assistance in welfare and mental health services (Garcia et al., 2018). The complexity of how dropping out influences subsequent offending (Sweeten et al., 2009) further indicates the need to conceptualize educational experiences as social processes, wherein the return to and completion of a missing degree can facilitate other prosocial changes. However, before discussing the transition back to education and the implications on offending trajectories, it is important to unpack the complexity of measuring the dropout problem.

Dropping out, stopping out, and going back

The dropout problem has decreased over time, but still remains a substantial concern in the United States. In 2016, 4.8% of individuals in the 15 to 24 age range, or approximately 532,000 individuals, were identified as event dropout cases or those who left school between 10th and 12th grade without earning a degree. However, the total prevalence of adolescents and adults not enrolled in school or possessing a degree, referred to as status dropouts, was estimated at 6.1% for that year (McFarland et al., 2018). This discrepancy in defining dropout status reflects the difference between individuals who leave in the short term but reenroll later, termed stop-outs, and long-term stable dropouts. This creates the opportunity to study a return to education, and a reengagement in positive social processes, for those who leave. It is possible that recommitting to formal education, through reenrolling in school, completing General Education Development (GED) testing, or graduating from high school, may serve to further desistance through behavioral change, cognitive enhancement, or facilitation of changes to one’s identity.

Education creates opportunities for social advancements and growth over a number of contexts such as marriage, employment, and crime reduction. Educational trend data identify degree attainment as a significant factor increasing marriage likelihood for men and women in the United States in the second half of the 20th century (Musick et al., 2012; Torr, 2011). Education can strengthen the stability of marriage, a particularly relevant point given the relationship between divorce and returning to crime (Bersani & Doherty, 2013). One assessment identified that college-educated women are significantly more likely to have marriages that last 20 years or more when compared with women with high school degrees or less (78% vs 40%; Copen, Daniels, Vespa, & Mosher, 2012).

In addition to marriage, educational attainment strengthens many facets of employment. A comparison of degree attainment established a clear trend of increased level of attainment, covering those who dropped out to individuals with PhDs, associated with lower levels of unemployment and higher average earning rates (Vilorio, 2016). Racial, ethnic, and age comparisons have also established support for educational attainment improving earning rates (Day & Newburger, 2002).

However, different forms of educational involvement and completion may not equally produce growth and opportunity. Many consider GED testing as an alternative to typical education completion. Completing testing, thereby earning a GED certificate, provides opportunities that individuals without any degree or certification would not have (Maralani, 2011), but in comparison with high school graduates, certificate holders experience greater difficulty in obtaining and maintaining employment, earn less income over the course of adulthood, and are less likely to pursue further education (Tyler, 2003). Using information from the Data from the Health and Retirement Survey in 2002 (Health and Retirement Survey) to compare GED recipients with high school graduates, the GED group reported weaker physical health and cognitive assessments and a greater prevalence of depressive symptoms (Caputo, 2005). Given additional arguments linking GED completion with criminal justice involvement (e.g., incarceration experience; Tuck, 2012), it is possible that GED completion may provide less of a desistance effect among stopped-out offenders.

To date, empirical efforts have not adequately explored the education and crime relationship within desistance research. Given prior findings identifying a reduction in property offenses (Hansen, 2003), it is possible that educational attainment after one has dropped out may negatively affect specific forms of criminal conduct, but the overall desistance effect remains unstudied.

Hypotheses

Hypothesis 1: Reenrolling in educational pursuits will positively influence desistance from crime among offenders who stopped out.

Hypothesis 2: Earning a GED certificate will not significantly influence desistance from crime among stopped-out offenders.

Hypothesis 3: Earning a high school diploma will positively influence desistance from crime among stopped-out offenders.

Method

Sample

This study utilized data from the NLSY97, initiated in 1997 using a sample representative of individuals living in the United States. The sample consisted of respondents born between 1980 and 1984, who were between the ages of 12 and 16 during the initial survey (Bureau of Labor Statistics, 1997-2011).

The NLSY97 included 8,984 youth in the initial survey. Respondents are interviewed annually, completing a self-administered survey that collects information on criminal activity, educational progress, employment, and marital status, among other topics. The NLSY97 was initially designed as a general population sample, which does not explicitly address individuals at risk of dropping out of school or developing offending trajectories. Therefore, the current study utilized a specific subset of the original NLSY97 to identify individuals for whom reenrollment and subsequent educational attainment could influence desistance from criminal behavior.

The first requirement involved identifying individuals who left the formal education system prematurely. The initial interview, and all subsequent interviews, asked respondents to report their current educational enrollment status. This allowed for the identification of respondents who reported no current enrollment, and who had not completed GED certification or graduated from high school, or earned subsequent degrees, as dropouts at each interview period. All cases of dropping out were summed and recoded dichotomously to identify all dropouts within the full NLSY97 sample.

The second requirement for study inclusion specified offense history prior to, or during, the year the respondent identified as having dropped out. Prematurely departing education does not necessarily mean that one has an offense history, but there is no current consensus in the life course literature on the minimum number, variety, or pattern of offenses one needs to demonstrate initially to be able to potentially desist later on. The current study identified offenders within each wave of dropouts through self-reports of different criminal activities. Specifically, at each interview respondents reported whether they had ever attacked someone with the intent to cause harm, sold illegal drugs, stolen something worth less than US$50, stolen something worth more than US$50, destroyed someone else’s property, or committed “other” property crimes. These variables were summed and recoded dichotomously for each wave of data collection to establish past offending prior to or concurrent with year identified as a dropout. This variable was also coded cumulatively so that an individual who committed an offense in 1997 but did not dropout until 1998, for example, could be included in the current study. The decision to use at least one prior offense to establish offender groups has been used in prior studies utilizing general population samples, specifically the National Longitudinal Study of Adolescent Health (Barnes & Beaver, 2012) and the NLSY97 (Forrest, 2014). These requirements identified a subsample of 2,104 dropout offenders, with 6,880 respondents excluded from analyses. Table 1 provides the descriptive details of this subsample.

Table 1.

Comparison of Dropout Offenders to Non-Dropout Respondents.

	Dropout offenders	NLSY97 remaining sample
	n = 2,104	n = 6,880
	M (SD)	M (SD)
Number of times attacked someone per year	0.26 (0.93)	0.09 (0.47)
Number of property crimes per year	0.74 (2.98)	0.35 (1.83)
Age—1997	14.24 (1.47)	14.33 (2.48)
Percent married	32.98	37.59
Percent parent	69.77	42.31
Average number of weeks worked per year	22.85 (13.44)	30.01 (14.00)
Percent African American	32.94	23.87
Percent male	56.65	49.52
Percent ever incarcerated	20.53	4.14
Child household size	4.78 (1.79)	4.48 (1.45)
Percent intact childhood home	29.71	54.80
Percent experienced hardship	8.89	3.60
Parent educational history	11.13 (3.86)	13.23 (3.82)

Note. NLSY97 = National Longitudinal Survey of Youth 1997.

Dropout offenders committed more violent (t = 8.23, p < .001) and property crimes (t = 5.71, p < .001) than non-dropout respondents, were younger at the beginning of the study (t = −2.26, p < .05), and were more likely to spend time incarcerated, χ²(1) = 589.37, p < .001. The subsample also experienced greater childhood hardship, χ²(1) = 97.61, p < .001, than non-dropouts, but were less likely to report their parents were married, χ²(1) = 405.95, p < .001, less likely to marry themselves, χ²(1) = 14.72, p < .001, and spent less time working (t = −21.16, p < .001). Finally, dropout offenders were more likely to be male, χ²(1) = 32.82, p < .001, have children themselves, χ²(1) = 486.33, p < .001, be African American, χ²(1) = 68.92, p < .001, grow up in larger households (t = 7.15, p < .001), and their parents spent fewer years in formal education (t = −21.94, p < .001) than non-dropouts.

Data Collection

Time-varying variables

The current study utilized two dependent variables to examine the specificity of educational attainment on desistance. During each interview, respondents were asked to report the number of times they engaged in different behaviors since the previous interview. Specifically, respondents revealed the number of times they attacked someone with the intent to cause serious harm, stole something worth less than US$50, stole something worth more than US$50, purposefully destroyed others’ property, or committed other property offenses. The four measures of property offenses, stealing items, property destruction, and other property offenses were summed for each interview period to create an overall property offense count variable. This approach has been used in previous desistance research to distinguish influences of change across different types of offending behavior (see Forrest, 2014). The violent crime count variable was utilized as reported. Table 2 provides further information on offending behavior over time, specifically the mean and standard deviation, as well as the number of observations, for each dependent variable across all identified waves of data collection.

Table 2.

Offenses by Waves of Data Collection.

Year	Attacked someone		Property crime count
Year	M (SD)	n	M (SD)	n
1999	0.67 (4.16)	2,100	1.68 (9.30)	2,092
2000	0.74 (5.13)	2,097	1.68 (10.48)	2,094
2001	0.42 (3.63)	2,101	1.49 (12.25)	2,097
2002	0.33 (2.19)	2,101	1.03 (8.80)	2,098
2003	0.48 (4.34)	2,102	1.22 (11.55)	2,101
2004	0.13 (1.74)	2,098	0.36 (5.21)	2,096
2005	0.15 (2.05)	2,100	0.36 (5.17)	2,095
2006	0.10 (1.15)	2,094	0.38 (5.32)	2,095
2007	0.12 (1.45)	2,100	0.45 (7.83)	2,095
2008	0.07 (0.73)	2,099	0.19 (2.50)	2,098
2009	0.07 (0.59)	2,101	0.34 (4.87)	2,096
2010	0.06 (0.59)	2,100	0.36 (5.94)	2,098
2011	0.04 (0.70)	2,098	0.19 (3.00)	2,094

To illustrate an influence of educational pursuits and attainment on subsequent offending, the current study utilized a staggered timeframe between educational status and offending. Specifically, each crime variable was measured in the wave immediately following that of educational status. The first wave of crime data was collected in 1999, as the first wave in 1997 included no means of distinguishing dropout offenders who reenrolled yet. Therefore, reenrollment and possible attainment (e.g., GED certification) were initially coded in the 1998 interviews, reflecting educational status between the first and second wave of data collection influencing criminal behavior occurring between the second and third. The final wave of data collection regarding crime was coded in 2011, as subsequent interviews did not collect all crime frequency variables.

Reenrolling in educational pursuits was measured by respondents reporting their current enrollment status. Respondents who reported no enrollment and who had not completed GED testing or graduated from high school were coded as 0, with those who reported being in school but not yet being a high school graduate coded as 1. Any respondent who indicated they were not enrolled but possessed a degree, such as a GED certificate, were coded as missing cases. Similarly, those who reported being enrolled in pursuit of higher educational degrees, such as an associate or bachelor’s degree, were also treated as missing values to avoid a confounding effect of postsecondary educational pursuits. Finally, coding enrollment status as a binary variable creates an additional complication in this context, as it does not distinguish between those who have not yet dropped out of school and those who reenrolled. For example, a respondent who reported being actively enrolled in 1998 who then dropped out in 1999 and reenrolled in 2000 would be coded longitudinally as 1, 0, and 1 again. To address this concern, I specified active enrollment status based on previous dropout status. For the model specifying reenrollment, respondents were only identified as actively pursuing education if they were identified as a dropout at a previous interview. This variable treated the hypothetical example just described as being unenrolled in 1998 and 1999, and reenrolled in 2000, to avoid this confounding effect.

Educational attainment was determined by respondent reports of the highest educational degree obtained. During each interview respondents were asked about the highest educational degree they had achieved prior to the start of the academic year. For example, in 1998 respondents reported their highest degree status before the start of the 1998-1999 academic year. Answers included no degree, GED certification, high school diploma, associate degree, bachelor’s degree, master’s degree, PhD, and other advanced degrees. Two key independent variables used in the current study included GED certification and high school graduation. For the former, respondents with no degrees were coded as 0 and those who identified as having completed GED certification were coded as 1. To isolate the effects of high school graduation, respondents with no degree were coded as 0 and those who reported having earned a high school diploma were coded as 1. In both cases, any subsequent advances in education (e.g., earning an Associate’s degree) were coded as missing values to isolate the specified effects. Figure 1 provides the distribution of degree completion status for sample respondents by the end of the current study period, the 2010 wave of data collection. For clarification, information was missing on this variable for 15.6% of subsample respondents.

Figure 1.

Degree completion by 2010.

Additional time-varying measurements included variables addressing previously established social controls. These included dichotomous self-reports on current marital status, parenthood, and a continuous measure of employment, defined as the number of weeks spent employed, either by oneself or in the employment of an individual or business, since the previous interview. These variables were identified and coded during the same year as educational status. These measurements identified respondents who were currently married and living with a spouse and those who had any biological children, either living with them or not. These measures reflect additional aspects of desistance (Kerr et al., 2011; Laub & Sampson, 2003). Respondent age in years was also identified at each interview and counted during the same timeframe as education, marriage, parenthood, and employment. During the initial interview in 1997, respondents were between 12 and 18 years of age. The last wave of data collected on age occurred in 2010, when respondents reported an age range of 25 to 31 years.

The NLSY97 provides data on offense histories, specifically monthly data on incarceration history prior to, and during, periods of data collection. An additional control measure was created, summarizing all affirmative responses to incarceration in the time since the previous interview, collected in the same timeframe as the other independent variables, and creating a dichotomous measurement of any incarceration experience by year.

Time-stable variables

All analyses included specific time-stable control measurements, beginning with dummy-coded indicators for male and African American respondents. Prior research has addressed several childhood and family factors that influence offending, including household size in childhood, intact household in childhood, defined as living with both parents in the same house, and hardship experiences in childhood, defined as any time spent living without running water or power or time spent in a homeless shelter. Prior desistance research has also incorporated control measures for parental educational history (Bersani & Doherty, 2013). The initial interview in 1997 identified these details, which were included in current models as a count of the number of individuals living in the respondent’s household, a dummy-coded response for living with both biological parents, a dummy-coded response for ever having lived through hardship, and a count-based measurement of the highest number of years any parent spent in education.

Analyses

All models were conducted using STATA 15, as it provides opportunities for descriptive reporting and allows for appropriate multilevel modeling. The current study utilized multilevel mixed-effects negative binomial regression models, with individuals at the second level and time, or specific interview waves, at Level 1 to examine the impact of different forms of educational status on overdispersed crime rates over 13 waves of data collection. The choice of multilevel negative binomial regression acknowledges the violation of the assumption of independence of measurements and provides an opportunity to study the effect of education within and between respondents (Raudenbush & Bryk, 2002). The choice of negative binomial modeling reflects the count-based distribution of offending behavior, specifically fitting an overdispersed distribution of crime frequency (see Tables 1 and 2).

Raudenbush and Bryk (2002) suggest modeling within-individual change over time by group-mean centering the time-varying covariates. This was done by calculating respondent difference from the specific mean within each time period, including educational enrollment, attainment, weeks worked, marriage, and parenthood status. This technique addresses the issue of any potentially biasing influence from time-stable differences in individual characteristics (Osgood, 2009).

The Level 1 equation, specifying within-individual change, follows this format:

\begin{array}{l} \log (λ_{i t}) = η_{i t} \\ η_{i t} = π_{0 i} + π_{1 i} {(age)}_{i t} + π_{2 i} {({age}^{2})}_{i t} + π_{3 i}, \dots, π_{k i} {(educational status)}_{i t} \end{array}

where η_it is the overdispersed count of criminal behavior for individual i at year t. The equation includes a quadratic function of age (age_it, ${age}_{i t}^{2}$ ) to address nonlinear developments in crime by age. Both age and age² were grand-mean centered to address collinearity.

The Level 2 equation, addressing changes between individuals, is modeled as follows:

\begin{array}{l} π_{0 i} = β_{00} + β_{01} {(controls)}_{i} + β_{02}, \dots, k {(\bar{educational status})}_{i} + r_{0 i} \\ π_{1 i} = β_{10} + r_{1 i} \\ π_{2 i} = β_{20} \\ π_{3 i \dots k i} = β_{30, \dots, k 0} \end{array}

where variations in crime count can be explained by aggregate measures of educational status, employment, marriage, parenthood, and other time-stable controls. This allows for variation in crime between individuals and by age. The error terms r_0i and r_1i address random effects for respondent and age, respectively, as desistance varies significantly between subjects and over time (Bersani & Doherty, 2018). All other variables were treated as having fixed effects.

Results

Reenrolling in School

Table 3 examines the effect of reenrolling in school on subsequent criminal behavior. Model 1 addresses assaultive behavior and identifies no within-individual changes in violent crime due to a respondent reenrolling in school. Each week spent employed decreases the log-count of attacks as does marriage, parenthood, and age. Respondents who reenroll in school see a decrease in the log-count of aggressive crimes by 0.16 when compared with respondents who did not (p < .1). Between-individual control measures for employment, child household size, and intact childhood households decrease the log-count of crime, while being a parent, being male, experiencing incarceration, and increases in one’s parents’ education increase criminal behavior. A likelihood ratio test comparing the random-effects model with a negative binomial model indicates the current model provides a significantly better fit to the data, χ²(2) = 1,152.30, p < .001.

Table 3.

Regressing Crime on Reenrollment.

	Attacking someone			Property crimes
	n (ind/obs) = 2,096/23,510			n (ind/obs) = 2,096/23,473
	Coefficient	SE	Sig.	Coefficient	SE	Sig.
Intercept	−5.21	0.40	***	−5.41	0.54	***
Within-individual
Reenrolled	0.03	0.24		0.56	0.28	*
Weeks worked	−0.01	0.002	*	−0.003	0.003
Parent	−0.40	0.16	*	−0.24	0.20
Marriage	−0.53	0.22	*	−0.57	0.29	*
Age	−0.28	0.16	^†	−0.39	0.21	^†
Age²	−0.003	0.004		−0.003	0.01
Between-individual
Reenrolled	−0.16	0.09	^†	−0.24	0.11	*
Weeks worked	−0.001	0.0004	^†	−0.001	0.001	*
Parent	0.10	0.02	***	0.07	0.03	*
Marriage	−0.01	0.03		−0.06	0.04
African American	−0.27	0.16		−0.53	0.22	*
Male	1.32	0.16	***	1.60	0.21	***
Ever incarcerated	0.79	0.18	***	0.76	0.22	**
Child household size	−0.07	0.04	^†	−0.14	0.06	*
Childhood hardship	0.16	0.24		−0.05	0.33
Intact household	−0.49	0.17	**	−0.34	0.22
Parent education	0.06	0.02	**	0.08	0.03	**

Note. GED = General Education Development.

†

p < .1. *p < .05. **p < .01. ***p < .001.

Model 2 examines property crime rates and identifies an increase in the log-count of property offenses by 0.56 when offenders reenroll in education (p < .05). The within-individual controls for marriage and age decrease the count of property crimes one commits. Respondents who reenroll in school see a decrease in the log-count of property crimes by 0.24 when compared with respondents who did not (p < .05). Between-individual controls for time spent employed and race decrease the frequency of property offenses, while measures controlling parenthood, gender, incarceration, and parental education increase property crimes. A likelihood ratio test comparing the random-effects model with a negative binomial model indicates the current model provides a significantly better fit to the data, χ²(2) = 1,661.38, p < .001.

GED Certification

Table 4 examines the impact of obtaining GED certification on future crime. Model 1 focuses on assaultive behavior and identifies no within-individual changes in crime due to educational attainment. Within-individual controls for employment, parenthood, and marriage reduce crime frequency in this context. Respondents who completed GED certification saw an increase in the log-count of aggressive conduct by 0.04 (p < .1), relative to respondents who did not obtain GEDs. Between-respondent controls for parenthood, gender, incarceration, and parental education reveal increases in aggressive conduct, while measures for time spent working, race, child household size, and parent marital status reduce the frequency of aggressive behavior. A likelihood ratio test comparing the random-effects model with a negative binomial model indicates the current model provides a significantly better fit to the data, χ²(2) = 1,093.23, p < .001.

Table 4.

Regressing Crime on GED Certification.

	Attacking someone			Property crimes
	n (ind/obs) = 2,090/20,449			n (ind/obs) = 2,090/20,416
	Coefficient	SE	Sig.	Coefficient	SE	Sig.
Intercept	−5.16	0.41	***	−5.60	0.54	***
Within-individual
GED certificate	−0.12	0.19		−0.70	0.25	**
Weeks worked	−0.01	0.002	*	−0.002	0.003
Parent	−0.39	0.16	*	−0.30	0.21
Marriage	−0.48	0.23	*	−0.57	0.29	^†
Age	−0.18	0.17		−0.41	0.21	^†
Age²	−0.01	0.004		−0.002	0.01
Between-individual
GED certificate	0.04	0.02	^†	0.09	0.03	**
Weeks worked	−0.001	0.0005	*	−0.002	0.001	**
Parent	0.09	0.02	***	0.06	0.03	*
Marriage	−0.0003	0.03		−0.04	0.04
African American	−0.31	0.16	^†	−0.65	0.22	**
Male	1.33	0.16	***	1.64	0.21	***
Ever incarcerated	0.70	0.18	***	0.59	0.23	*
Child household size	−0.07	0.04	^†	−0.13	0.06	*
Childhood hardship	0.14	0.25		−0.04	0.33
Intact household	−0.52	0.17	**	−0.30	0.22
Parent education	0.06	0.02	**	0.09	0.03	**

Note. GED = General Education Development.

†

p < .1. *p < .05. **p < .01. ***p < .001.

Model 2 addresses property crimes and identifies a decrease in the log-count of property offending by 0.70 after offenders obtained GED certification (p < .01). Within-individual controls for marriage and age decrease the frequency of property crimes one commits. Respondents who completed GED certification saw an increase in the log-count of property crimes by 0.09 (p < .01), relative to respondents who did not obtain GEDs. Additional controls for employment, race, and child household size reveal decreases in the frequency of crime, while measures for parenthood, gender, incarceration, and parental education increase the frequency of property offenses. A likelihood ratio test comparing the random-effects model to a negative binomial model indicates the current model provides a significantly better fit to the data, χ²(2) = 1,466.76, p < .001.

High School Completion

Table 5 focuses on the impact of high school graduation on subsequent crimes. Model 1 focuses on assaultive behavior and identifies no within-individual changes in crime due to graduating from high school. Within-individual controls for employment, parenthood, and marriage decrease violent crime rates. Respondents who graduated from high school saw a decrease in the log-count of aggressive conduct by 0.07 (p < .05), relative to respondents who did not graduate. Between-respondent controls for parenthood, gender, incarceration, and parental education reveal increase in aggressive conduct, while measures for intact childhood home and child household size reduce the frequency of aggressive behavior. A likelihood ratio test comparing the random-effects model with a negative binomial model indicates the current model provides a significantly better fit to the data, χ²(2) = 1,163.99, p < .001.

Table 5.

Regressing Crime on High School Graduation.

	Attacking someone			Property crimes
	n (ind/obs) = 2,096/22,983			n (ind/obs) = 2,096/22,947
	Coefficient	SE	Sig.	Coefficient	SE	Sig.
Intercept	−5.28	0.40	***	−5.52	0.54	***
Within-individual
High school completion	−0.08	0.28		−0.57	0.33	^†
Weeks worked	−0.01	0.002	*	−0.002	0.003
Parent	−0.41	0.16	**	−0.26	0.20
Marriage	−0.54	0.22	*	−0.57	0.29	*
Age	−0.27	0.16		−0.40	0.21	^†
Age²	−0.003	0.004		−0.003	0.01
Between-individual
High school completion	−0.07	0.03	*	0.02	0.04
Weeks worked	−0.001	0.0005		−0.001	0.001	*
Parent	0.10	0.02	***	0.07	0.03	*
Marriage	−0.0004	0.03		−0.05	0.04
African American	−0.23	0.16		−0.56	0.22	*
Male	1.29	0.16	***	1.60	0.21	***
Ever incarcerated	0.73	0.18	***	0.71	0.22	**
Child household size	−0.07	0.04	^†	−0.14	0.06	*
Childhood hardship	0.15	0.25		−0.07	0.33
Intact household	−0.44	0.17	**	−0.34	0.22
Parent education	0.07	0.02	**	0.08	0.03	**

†

p < .1. *p < .05. **p < .01. ***p < .001.

Model 2 addresses property crimes and identifies a decrease in the log-count of property offending by 0.57 after offenders successfully graduated from high school (p < .1). Within-individual controls for marriage and age decrease the frequency of property crimes each respondent subsequently commits. Respondents who graduated from high school were no more or less likely to commit property offenses than those who did not. Additional between-respondent controls for employment, race, and child household size reveal decreases in the frequency of crime, while measures for parenthood, gender, incarceration, and parental education increase the frequency of property offenses. A likelihood ratio test comparing the random-effects model with a negative binomial model indicates the current model provides a significantly better fit to the data, χ²(2) = 1,642.74, p < .001.

Discussion

Reenrollment

Reenrolling in education produces reductions in different types of crimes for individuals who returned compared against those who did not. This recommitment to school may reduce unstructured time, thereby reducing opportunities for criminal behavior, and create numerous prosocial opportunities such as career advancement. This finding supports the first hypothesis and builds on previous research about reduction of type of offense into a broader desistance effect (Hansen, 2003). While this is a useful finding for the value of educational recommitment as a facilitator of change, there remain possible substantial baseline differences in dropout status, scale of offense history, and factors influencing the decision to return (Boman & Mowen, 2018).

The within-individual increase in property offending due to reenrollment, coupled with the lack of significance on assaultive behavior, suggests that educational engagement does not facilitate short-term prosocial growth and in fact may create immediate financial strains. However, the between-subject findings suggest broader long-term opportunities to develop prosocial career or personal trajectories influenced by recommitting to education. Reenrolling in education creates short-term changes in one’s routine activities and may reflect the initial stages of the reformation of an attenuated social bond or simply a delinquent youth acquiescing to parental pressure to return to school.

GED Certification

While the results provide inconsistent findings on the impact of GED completion on crime, they are somewhat supportive of the second hypothesis and concur with prior arguments suggesting a limited influence of certification on prosocial developments. Respondents who earned a GED committed fewer property crimes, but not violent crimes. This suggests that certification creates some opportunities for legitimate employment, but the fact remains that respondents who completed certification were more likely to commit violent and property crimes than offenders who did not earn this certification. If successful GED testing produces short-term employment changes, it is necessary to consider stability of employment, income changes, and qualitative aspects of work (i.e., whether respondents enjoy their jobs) before and after certification and incorporate long-term employment, relationship, and offending variables into the desistance argument (Tyler, 2003).

While offenders may drop out of school by choice, and even subsequently complete GED testing, it is important to acknowledge that many are mandatorily enrolled in educational programs due to placement in secure custody (Guiding Principles for Providing High-Quality Education in Juvenile Justice Secure Care Settings, 2014; Twomey, 2008). If certification testing is completed out of correctional requirement rather than choice, the agentic impact on desistance would not likely manifest.

High School Completion

As with the model examining GED certification, the assessment of high school completion reveals some inconsistencies but is generally supportive of the third hypothesis. Respondents who earned a high school diploma, having previously dropped out, were significantly less likely to commit property crimes afterward, but earning the degree did not alter their commission of violent crimes. Looking at the between-respondent comparisons identified an alternative effect to that of GED certification, where violent crime was less likely among graduates, while property crime rates did not differ between offenders who did and did not graduate. Past research identified an increase in violent crime after dropping out (Jarjoura, 1993), so returning and completing school would logically reduce this risk.

Graduation creates greater opportunities for social advancement, employment and even career growth (Day & Newburger, 2002; Musick et al., 2012), and given the findings on violent crime between those who graduated and those who did not, the long-term desistance influences warrant consideration.

Revisiting Theories of Desistance

These results suggest value in a return to education for those offenders who have dropped out, particularly given the limited prevalence of degree attainment among stopped-out offenders (see Figure 1). Unfortunately, the variables collected in the NLSY97 do not allow for explicitly testing current theories of desistance. Specifically, the data do not allow researchers to assess informal social control influences, such as spousal pressure on an offender to return and complete a degree, nor do they include variables examining cognitive processing or changes in identity. Nevertheless, the value of educational attainment, after initially dropping out, provides new opportunities to expand theoretical insight.

Current desistance research is shifting away from the presumed dichotomy of purely external versus internal influences of desistance (Bersani & Doherty, 2018). A comprehensive explanation of behavioral change must incorporate arguments of changing self-efficacy or control (Lebel, Burnett, Maruna, & Bushway, 2008; Liem, 2016), intentionally prosocial actions by the individual (Carlsson, 2016), and social experiences that either initiate or extend positive growth (Bersani & Doherty, 2018).

The decision to return and complete formal education could reflect internal agency, if a hypothetical dropped-out offender desires more stable employment or is actively attempting to develop a new prosocial identity. It is also distinctly possible that a hypothetical dropped-out offender could be pushed into going back to school by parents, friends, or a significant other and that degree completion could facilitate additional social controls. This study opens inquiry into blending external social control experiences with internal mechanisms of change by addressing the need, in future research, to study the reasons for dropping out, with the motivation to return plotted against time-specific developments in offending trajectories.

Laub and Sampson (1993) initially argued that transitioning out of crime involves gradual progression, which could be reflected in the return to education, obtaining the required degree, and pursuing career advancement opportunities. Going back to formal educational pursuits is reflective of a broader behavioral transition. This may not initiate the desistance process for stopped-out offenders, but is a contributing influence to intensify the reduction in frequency and severity of offending (Bersani & Doherty, 2018).

Considering Educational Policy

The purpose of the current study is to explore the potential relevance of education as a mechanism of desistance among offenders who prematurely dropped out of school. The current study adds to both theoretical and policy discussions of desistance by exploring the long-term impact of degree attainment on offending (Bersani & Doherty, 2018) and establishing significant between-individual effects for returning to education and earning a high school diploma, as well as identifying within-individual reductions in offending for those who completed GED testing or earned their high school diploma. It is likely that, among stopped-out offenders, educational attainment facilitates greater stability of employment and improvements in legitimate income (Musick et al., 2012). Therefore, policy consideration must address how to effectively facilitate a return to school for dropped-out youth, particularly those with offense histories, and develop opportunities for stopped-out youth and adults to consider specific career paths they might pursue as they return to formal education. This point is particularly relevant given the limited prevalence of degree completion status identified in Figure 1.

A related question to the topic of advancing employment opportunity and experience is whether educational attainment may also facilitate opportunities for physical relocation. If an offender lives in a high-crime neighborhood, how might degree completion and employment change facilitate new housing opportunities to further support prosocial change (Doherty & Bersani, 2016)?

Desistance research efforts have predominantly focused on the process of behavioral change, while policy arguments largely focus on desistance as an end state after which the individual commits no subsequent crimes. While both definitions remain useful for theory, research, and policy considerations, the current findings lend support to expanding the process definition, where a cessation of offending occurs gradually over time, further into policy discussions and efforts (Bersani & Doherty, 2018).

Future Research Opportunities

To dig further into the agency of desistance, future empirical work requires qualitative educational research, involving narrative interviews with stopped-out offenders who returned to school, to help articulate both external and internal factors of change (Bersani & Doherty, 2018). Desistance theories involving the evolution of identity and changes in criminological thought processes must be scrutinized during the time in which an offender has dropped out, after they have returned to school, and after the desired degree is obtained (Giordano et al., 2002; Paternoster & Bushway, 2009). Future efforts should also incorporate quantitative data on familial and peer influences on the ultimate decision to return to further elucidate social control arguments for desistance through education (Laub & Sampson, 2003) and attempt to provide further support for the need for theoretical integration in desistance work (Bersani & Doherty, 2018).

In addition to ongoing theoretical debate, desistance research involves substantial methodological variation in conceptualizing and operationalizing behavioral change. The current study utilized a multilevel mixed-effects model, examining time at level 1 and individual at level 2, to document both within- and between-individual developments, which is consistent with prior desistance studies (Bersani & Doherty, 2013; Forrest, 2014). However, other studies utilize subgroup trajectory modeling to identify latent categories of offenders and examine different desistance influences (Laub, Nagin, & Sampson, 1998). This technique could be applied to the question of offenders who drop out and subsequently return to education, but an ongoing challenge for trajectory modeling in desistance research is how to distinguish de-escalating offense rates for those who are desisting but still active (e.g., chronic, high-rate offenders; Blokland & van Os, 2010; Bushway, Thornberry, & Krohn, 2003; Piquero, Farringon, & Blumstein, 2007; Sampson & Laub, 2003).

The current findings also revisit the principles of life course criminology by raising a question about the importance of the timing of events within one’s life. Specifically, does a return to education while still an adolescent provide more consistent prosocial change than if one drops out, ages into adulthood, then returns to earn a degree? This question can be explored by separating stopped-out offenders into those who returned before versus after age 18 and modeling offense trajectories over a multiyear follow up period.

Additional research questions might elaborate on race, gender, or possible socioeconomic differences in the return to education and whether that facilitates desistance. Differences in desistance experiences across race and ethnic status have been established (Piquero, 2015). Gender-specific developments have recently become a focal concern in desistance research (Bersani & Doherty, 2018; Broidy et al., 2015), but have not been sufficiently explored, particularly with regard to dropout status. Similarly, given the overlap between poverty and dropout risk (Aratani & Cooper, 2015; Garcia et al., 2018), socioeconomic status requires consideration when examining the crime-reducing effect of returning to, and completing, a desired educational degree.

Finally, the current study identified the impact of education on different types of crime, but specific measurement discrepancies in the NLSY97 prevent the inclusion of drug use in the current models. During each interview, respondents are asked to report recent cigarette, alcohol, and other drug use, but only within the last 30 days. This is an understandable measurement strategy, given the challenge of accurately recalling drug use over time, but unfortunately rules out the ability to directly compare drug use to other offending behaviors. Educational programs targeting substance use patterns produce inconsistent findings by different age groups (Onrust, Otten, Lammers, & Smit, 2016), which warrants much greater examination in the context of education reentry and degree completion.

Conclusion

Desistance from crime involves internal processes (e.g., cognitive transformations) and external controls (e.g., spousal influence), short-term experiences (Bersani & Doherty, 2013), and long-term developmental changes (Paternoster & Bushway, 2009). Debate continues on theory, definition, methodology, and policy impact, but efforts to study desistance have missed an important opportunity to incorporate education into the topic of behavioral change. The current study addresses this gap by examining dropped-out and stopped-out offenders, and examining the influence of returning to education, and earning either GED certification or a high school diploma, on different types of offending behavior. The importance of education in criminological work focuses primarily on preventing or controlling future crime (Hansen, 2003; Hjalmarsson, 2008; Lochner, 2004). While this is important, education can also facilitate desistance among those who dropped out of school.

Footnotes

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

The author(s) received no financial support for the research, authorship, and/or publication of this article.

ORCID iD

David Abeling-Judge

Author Biography

David Abeling-Judge is an assistant professor of Criminology and Criminal Justice Sciences at Barton College. His research interests include psychosocial influences of desistance from crime, life course criminology, and sociocultural developments in criminology. He earned a master of science in Forensic Mental Health Sciences from King’s College London and a PhD in Criminology and Justice Policy from Northeastern University.

References

Aratani

Cooper

J. L.

(2015). The effects of runaway-homeless episodes on high school dropout. Youth & Society, 47, 178-198.

Barnes

J. C.

Beaver

K. M.

(2012). Marriage and desistance from crime: A consideration of gene–environment correlation. Journal of Marriage and Family, 74, 19-33.

Bersani

B. E.

Doherty

E. E.

(2013). When the ties that bind unwind: Examining the enduring and situational processes of change behind the marriage effect. Criminology, 51, 399-433.

Bersani

B. E.

Doherty

E. E.

(2018). Desistance from offending in the twenty-first century. Annual Review of Criminology, 1, 311-334.

Bersani

B. E.

Laub

J. H.

Nieuwbeerta

(2009). Marriage and desistance from crime in the Netherlands: Do gender and socio-historical context matter? Journal of Quantitative Criminology, 25, 3-24.

Bersani

B. E.

Schellen

M. V.

(2014). The effectiveness of marriage as an intervention in the Netherlands. In Humphrey

J. A.

Cordella

(Eds.), Effective interventions in the lives of criminal offenders (pp. 101-120). New York, NY: Springer.

Blokland

A. A. J.

van Os

. (2010). Life span offending trajectories of convicted Dutch women. International Criminal Justice Review, 20, 169-187.

Boman

Mowen

T. J.

(2018). The role of turning points in establishing baseline differences between people in developmental and life-course criminology. Criminology, 56, 191-224.

Broidy

Stewart

Thompson

Chrzanowski

Allard

Dennison

S. M.

(2015). Life course offending pathways across gender and race/ethnicity. Journal of Developmental and Life-Course Criminology, 1, 118-149.

10.

Bureau of Labor Statistics, U.S. Department of Labor (1997-2011). National Longitudinal Survey of Youth 1997 cohort (rounds 1 - 15). Produced by the National Opinion Research Center, the University of Chicago and distributed by the Center for Human Resource Research, The Ohio State University. Columbus, OH.

11.

Bushway

Thornberry

T. P.

Krohn

M. D.

(2003). Desistance as a developmental process: A comparison of static and dynamic approaches. Journal of Quantitative Criminology, 19, 129-153.

12.

Caputo

R. K.

(2005). The GED as a signifier of later life health & economic well-being. Race, Gender & Class, 12, 81-93, 95-103.

13.

Carlsson

(2016). Human agency, criminal careers and desistance. In Shapland

Farrall

Bottoms

(Eds.), Global perspectives on desistance: Reviewing what we know and looking to the future (pp. 28-49). New York, NY: Taylor & Francis.

14.

Copen

C. E.

Daniels

Vespa

Mosher

W. D.

(2012). First marriages in the United States: Data from the 2006-2010 national survey of family growth. Hyattsville, MD: National Center for Health Statistics.

15.

Day

J. C.

Newburger

E. C.

(2002). The big payoff: Educational attainment and synthetic estimates of work-life earnings. Washington, D.C.: U.S. Department of Commerce, U.S. Census Bureau.

16.

Doherty

E. E.

Bersani

(2016). Understanding the mechanisms of desistance at the intersection of race, gender, and neighborhood context. Journal of Research in Crime & Delinquency, 53, 681-710.

17.

Elder

G. H.

(1998). The life course as developmental theory. Child Development, 69, 1-12.

18.

Forrest

(2014). Cohabitation, relationship quality, and desistance from crime. Journal of Marriage and Family, 76, 539-556.

19.

Garcia

A. R.

Metraux

Chen

C. C.

Park

J. M.

Culhane

D. P.

Furstenberg

F. F.

(2018). Patterns of multisystem service use and school dropout among seventh-, eighth-, and ninth-grade students. The Journal of Early Adolescence, 38, 1041-1073.

20.

Giordano

P. C.

Cernkovich

S. A.

Rudolph

J. L.

(2002). Gender, crime, and desistance: Toward a theory of cognitive transformation. American Journal of Sociology, 107, 990-1064.

21.

Giordano

P. C.

Schroeder

R. D.

Cernkovich

S. A.

(2007). Emotions and crime over the life course: A neo-meadian perspective on criminal continuity and change. The American Journal of Sociology, 112, 1603-1661.

22.

Gottfredson

M. R.

Hirschi

(1990). A general theory of crime. Stanford, CA: Stanford University Press.

23.

Groot

van den Brink

H. M.

(2010). The effects of education on crime. Applied Economics, 42, 279-289.

24.

Guiding principles for providing high-quality education in juvenile justice secure care settings. (2014). Washington, DC: U.S. Departments of Education and Justice.

25.

Hansen

(2003). Education and the crime-age profile. British Journal of Criminology, 43, 141-168.

26.

Health and Retirement Survey (2002). Public use dataset. Produced and distributed by the University of Michigan with funding from the National Institute on Aging (Grant #: NIA U01AG009740). Ann Arbor, MI.

27.

Hjalmarsson

(2008). Criminal justice involvement and high school completion. Journal of Urban Economics, 63, 613-630.

28.

Jarjoura

G. R.

(1993). Does dropping out of school enhance delinquent involvement? Results from a large-scale national probability sample. Criminology, 31, 149-172.

29.

Jordan

Lara

McPartland

(1996). Exploring the causes of early dropout among race, ethnic, and gender groups. Youth & Society, 28, 62-94.

30.

Kerr

D. C. R.

Capaldi

D. M.

Owen

L. D.

Wiesner

Pears

K. C.

(2011). Changes in at-risk American men’s crime and substance use trajectories following fatherhood. Journal of Marriage and Family, 73, 1101-1116.

31.

Laub

J. H.

Nagin

D. S.

Sampson

R. J.

(1998). Trajectories of change in criminal offending: Good marriages and the desistance process. American Sociological Review, 63, 225-238.

32.

Laub

J. H.

Sampson

R. J.

(1993). Turning points in the life course: Why change matters to the study of crime. Criminology, 31, 301-325.

33.

Laub

J. H.

Sampson

R. J.

(2001). Understanding desistance from crime. Crime and Justice, 28, 1-69.

34.

Laub

J. H.

Sampson

R. J.

(2003). Shared beginnings, divergent lives: Delinquent boys to age 70. Cambridge, MA: Harvard University Press.

35.

Lebel

T. P.

Burnett

Maruna

Bushway

(2008). The “chicken and egg” of subjective and social factors in desistance from crime. European Journal of Criminology, 5, 131-159.

36.

Liem

(2016). After life imprisonment: Reentry in the era of mass incarceration. New York: New York University Press.

37.

Lochner

(1999). Education, work, and crime: Theory and evidence (Working Paper No. 465). Rochester, NY: Rochester Center for Economic Research. Retrieved from https://ssrn.com/abstract=208291

38.

Lochner

(2004). Education, work, and crime: A human capital approach. International Economic Review, 45, 811-843.

39.

Machin

Marie

Vujic

(2011). The crime reducing effect of education. The Economic Journal, 121, 463-484.

40.

Maralani

(2011). From GED to college: Age trajectories of nontraditional educational paths. American Educational Research Journal, 48, 1058-1090.

41.

Maruna

(2001). Making good: How ex-convicts reform and rebuild their lives. Washington, DC: American Psychological Association.

42.

McFarland

Cui

Rathbun

Holmes

(2018). Trends in high school dropout and completion rates in the United States: 2018. Washington, DC: National Center for Education Statistics, Institute of Education Sciences.

43.

McNeal

(1997). Are students being pulled out of high school? The effect of adolescent employment on dropping out. Sociology of Education, 70, 206-220.

44.

Moffitt

T. E.

(1993). Adolescence-limited and life-course-persistent antisocial behavior: A developmental taxonomy. Psychological Review, 100, 674-701.

45.

Musick

Brand

J. E.

Davis

(2012). Variation in the relationship between education and marriage: Marriage market mismatch? Journal of Marriage & Family, 74, 53-69.

46.

Onrust

S. A.

Otten

Lammers

Smit

(2016). School-based programmes to reduce and prevent substance use in different age groups: What works for whom? Systematic review and meta-regression analysis. Clinical Psychology Review, 44, 45-59.

47.

Osgood

D. W.

(2009). Statistical models for life events and criminal behavior. In Piquero

Weisburd

(Eds.), Handbook of quantitative criminology (pp. 375-396). New York, NY: Springer.

48.

Paternoster

Bushway

(2009). Desistance and the “feared self”: Toward an identity theory of criminal desistance. Journal of Criminal Law & Criminology, 99, 1103-1156.

49.

Piquero

(2015). Understanding race/ethnicity differences in offending across the life course: Gaps and opportunities. Journal of Developmental and Life-Course Criminology, 1, 21-32.

50.

Piquero

Farringon

D. P.

Blumstein

(2007). Key issues in criminal career research: New analyses of the Cambridge study in delinquent development. Cambridge, UK: Cambridge University Press.

51.

Raudenbush

S. W.

Bryk

A. S.

(2002). Hierarchical linear models: Applications and data analysis methods (2nd ed.). Thousand Oaks, CA: SAGE.

52.

Reynolds

A. J.

Temple

J. A.

S. R.

(2010). Preschool education, educational attainment, and crime prevention: Contributions of cognitive and non-cognitive skills. Children and Youth Services Review, 32, 1054-1063.

53.

Rocque

(2017). Desistance from crime: New advances in theory and research. New York, NY: Palgrave Macmillan.

54.

Sampson

R. J.

Laub

J. H.

(2003). Life-course desisters? Trajectories of crime among delinquent boys followed to age 70. Criminology, 41, 555-592.

55.

Stearns

Glennie

E. J.

(2006). When and why dropouts leave high school. Youth & Society, 38, 29-57.

56.

Stevens

(2012). “I am the person now I was always meant to be”: Identity reconstruction and narrative reframing in therapeutic community prisons. Criminology & Criminal Justice, 12, 527-547.

57.

Sweeten

Bushway

S. D.

Paternoster

(2009). Does dropping out of school mean dropping into delinquency? Criminology, 47, 47-91.

58.

Taheri

S. A.

Welsh

B. C.

(2016). After-school programs for delinquency prevention: A systematic review and meta-analysis. Youth Violence and Juvenile Justice, 14, 272-290.

59.

Torr

B. M.

(2011). The changing relationship between education and marriage in the United States, 1940-2000. Journal of Family History, 36, 483-503.

60.

Tuck

(2012). Repatriating the GED: Urban youth and the alternative to a high school diploma. The High School Journal, 95, 4-18.

61.

Twomey

(2008). The right to education in juvenile detention under state constitutions. Virginia Law Review, 94, 765-811.

62.

Tyler

J. H.

(2003). Economic benefits of the GED: Lessons from recent research. Review of Educational Research, 73, 369-403.

63.

Vilorio

(2016). Education matters. Washington, DC: U.S. Bureau of Labor Statistics.