Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2021 Apr 26.
Published in final edited form as: Youth Soc. 2018 Jun 14;52(3):377–402. doi: 10.1177/0044118x18781641

The Role of Violent and Nonviolent Delinquent Behavior in Educational Attainment

Jinho Kim 1
PMCID: PMC8075283  NIHMSID: NIHMS976244  PMID: 33907338

Abstract

Given large variations in the etiology and developmental trajectories of violent and nonviolent delinquency, this study examines whether educational outcomes of violent and nonviolent offenders might differ. In particular, this study attempts to remove environmental influences such as family background and neighborhood effects from the effects of delinquency because these factors are likely to differentially confound the effects of violent and nonviolent delinquency on educational attainment. By exploiting variation within sibling pairs, this study finds that the effects of engagement in violent delinquency on education is driven spuriously by shared family background, whereas the effects of nonviolent delinquency are quite robust to adjustment for family fixed effects. Moreover, relying on fixed effects estimates, this study finds that the effects of engagement in nonviolent delinquent activity on educational attainment occur in part through disruption of educational progress, rather than through institutional responses to student delinquency and social-psychological processes.

Keywords: Delinquency, Violent Behavior, Education

Criminal behavior among adolescents has received substantial recognition as a pressing social problem because it has long-term negative consequences on various life outcomes (T. E. Moffitt, Caspi, Harrington, & Milne, 2002; Nilsson & Estrada, 2011). Lower educational attainment is one of the most widely studied negative consequences of engagement in delinquent behavior. A number of past studies have suggested that adolescents who engaged in delinquent behaviors or entered the criminal justice system (i.e., arrest, conviction, charge, and incarceration) are more likely to have poor educational attainment: higher rates of school dropout and lower rates of college enrollment and graduation (Hannon, 2003; Hirschfield, 2009; Hjalmarsson, 2008; Kirk & Sampson, 2013; Monk-Turner, 1989; Nilsson & Estrada, 2011; Ou, Mersky, Reynolds, & Kohler, 2007; Siennick & Staff, 2008; Tanner, Davies, & O’Grady, 1999; Webbink, Koning, Vujic, & Martin, 2013). Although the deleterious effects of delinquency have been studied extensively and are linked to various educational outcomes, previous research is limited because of (1) limited knowledge on heterogeneous effects of offending types, (2) debates on whether effects are causal or spurious, and (3) inconclusiveness about the mechanisms at work between delinquency and education.First, surprisingly little is known about whether potential heterogeneous effects by types of offending exist. Despite the advantages of using a summary measure of delinquency, this method may be one of the primary reasons for some conflicting results from prior studies and mask meaningful differences in the relationship between different types of offending and educational attainment. For example, given relatively large variations in the etiology and developmental trajectories of delinquent activities (e.g., Moffitt, 1993), it is reasonable to speculate that engagement in different types of offending holds differential impacts on education. Specifically, following the criminal careers literature and the development criminology perspectives (see Piquero, Farrington, & Blumstein, 2007), this study assesses the extent to which violent and nonviolent delinquent behaviors affect educational attainment differently among a sample of adolescents who have committed major delinquent acts in the past (which cause measurable harm to others and/or merit attention by the police and judicial system).

Second, there is an ongoing debate about whether the “effect” of engagement in delinquency on educational outcomes is causal or spurious. While the majority of prior research has documented the adverse effects of delinquent behavior on education, existing evidence has been derived from empirical models that contain several methodological limitations. One of the most critical challenges in examining the effect of delinquency on educational attainment is that individuals select into delinquency based on unobserved as well as observed factors. If unobserved characteristics such as family background determine both delinquency and educational outcomes, the observed effects of delinquency on education can be spurious, and not causal. Distinguishing violent from nonviolent delinquency is also relevant to these confounding issues, in that differential contributions of family background characteristics on selection into violent and nonviolent behavior may differentially confound the effect on educational outcomes.

Third, although a number of prior studies have suggested several potential mechanisms that link delinquent behavior to education, including institutional responses to student criminality (disciplinary actions), social-psychological factors, and disruption of educational progress (e.g., Hjalmarsson 2008; Kirk and Sampson 2013; Siennick and Staff 2008), this evidence is far from conclusive. Related to the two issues raised above, previous studies are limited in that they rely on potentially biased estimates of the effect of delinquency, and do not consider the possibility of differential pathways to educational attainment on the basis of different types of offending.

In the present study, I draw upon the National Longitudinal Study of Adolescent to Adult Health (Add Health), a nationally representative sample with a large subsample of siblings as well as a rich array of self-reported delinquent behavior. The objective of this paper is twofold. First, this study examines whether the effects of violent and nonviolent criminal behavior on educational attainment differ among a sample of adolescents at risk for serious criminal offending. Importantly, using family fixed effects models to remove all measured and unmeasured family background characteristics shared by siblings, this study examines whether the relationship between delinquent behavior and educational attainment is spurious owing to unobserved family background characteristics. Second, using more credible estimates of the effect of delinquency (i.e., fixed effects estimates), this study tests several potential mechanisms through which violent and nonviolent delinquent behavior may influence educational attainment.

Delinquency and Educational Attainment

Scholars have long argued that delinquency impacts educational outcomes including academic achievement, high school dropout, and educational attainment. While the focus of previous research has been on establishing the relationship between criminal justice contact (e.g., arrest, charge, conviction, and incarceration) and education (e.g., Hirschfield 2009; Hjalmarsson 2008; Kirk and Sampson 2013; Monk-Turner 1989; Webbink et al. 2013), several studies have suggested that juvenile delinquency per se is related to educational outcomes. Tanner and colleagues (1999) found that both male and female adolescents who engaged in delinquent activities, especially skipping school and drug use, were less likely to receive a high school diploma and obtain a college degree. Siennick and Staff (2008) also showed that a higher delinquency score summarizing frequency of fighting, alcohol, drug use, and arrest is negatively associated with the probability of enrolling in a postsecondary institution and receiving a bachelor’s degree or a higher level of education.

Prior research has also documented heterogenous effects of delinquency on education. Hannon (2003) suggested that while those with a high delinquency score, defined by a variety scale of thirteen items related to a wide range of adolescent deviant activities, are more likely to exhibit low educational attainment. These harmful effects of delinquency are less salient for low-income youth because they tend to have additional structural barriers to educational attainment beyond delinquent behavior. Ward and Williams (2015) argued that the effect of delinquency on high school and college graduation is largely driven by early initiators, those who commit delinquent activity more intensely, and those whose delinquent activities involve income-generating acts (e.g., stealing, committing property crimes, selling drugs, etc.).

Despite theoretical arguments about and empirical evidence on the link between delinquency and educational attainment, there is an ongoing debate about whether the “effect” of engagement in delinquent behavior on educational outcomes is causal or spurious. The observed associations could reflect selection processes generated by (measured and unmeasured) common third factors that affect both the risk of engaging in delinquent behavior and educational attainment (Hirschfield, 2009; Kirk & Sampson, 2013; Ward & Williams, 2015). Therefore, if confounding selection factors are not properly controlled, they could inflate the association between delinquency and education.

One of the key confounding factors should be environmental influences such as family background and neighborhood characteristics during childhood. For example, a large body of research suggests that family background characteristics such as genetic background, parental incarceration, and parenting styles and involvement are likely to jointly predict delinquency and educational achievement (Agan, 2011; Cernkovich & Giordano, 1987; Fagan & Wexler, 1987; Guo, Roettger, & Cai, 2008; Heckman, Pinto, & Savelyev, 2012; Laub & Sampson, 1988; Matsueda & Heimer, 1987; Patterson & Stouthamer-Loeber, 1984; Wells & Rankin, 1991). Although scholars have traditionally attempted to eliminate family-level confounders by adding a set of observed family-level controls such as parental education, family income, and family structure, merely controlling for family characteristics available in the data may not be sufficient to effectively address family-level heterogeneity, especially when these factors are unobserved.

Violent and Nonviolent Delinquency

Despite a growing body of evidence on the relationship between delinquency and educational attainment, but relatively little is known about whether this relationship differs by types of offending. In fact, criminological theories such as developmental criminology perspectives and the criminal careers literature have documented that different types of criminal offending exhibit unique structural progression or trajectories in terms of etiology, prevalence, frequency, duration, and seriousness (Burt & Neiderhiser, 2009; Lacourse et al., 2002; Loeber & Farrington, 1998; T. E. Moffitt, 1993; Piquero et al., 2007). However, prior scholarship has often relied on a summary measure of delinquent behavior as a representation of a general propensity toward delinquency, which may mask important heterogeneity in the relationship between delinquency and education. While delinquent activities may be categorized in several ways (e.g., Hannon, 2003; Ward and Williams, 2015), the etiology and developmental trajectories of violent and nonviolent delinquent behavior are known to be quite distinct.

The most well-known model of criminal offending trajectories is Moffitt’s (1993) developmental taxonomy. Moffitt (1993) argues that there are two types of adolescent offenders: life-course persistent (LCP) offenders and adolescence-limited (AL) offenders. LCP offenders are a group of offenders who continue to commit crimes during adolescence and beyond, whereas AL offenders are those who engage in delinquency only during adolescence. One of the major distinctions in patterns of offending between these two groups is that LCP offenders tend to commit violent crimes against persons, whereas AL offenders account primarily for such crimes such as theft, vandalism, public order, and substance abuse (mostly nonviolent crimes). AL offenders are likely to be considerably vulnerable to peer influence since mimicking peers’ behaviors allows adolescents to gain more privileges and thus increase access to some desirable resources during adolescence (i.e., adult life-styles, more power, and more sexual partners). Unlike AL offenders, criminal offending by LCP offenders is likely to be influenced by the factors that are determined during childhood such as “mild neuropsychological impairment, poor self-control, pathological interpersonal relationships, weak connections to other people, and a lifelong antisocial personality configuration” (Moffitt, 1993, p. 691). And, these factors are known to be exacerbated by exposure to criminogenic environments during childhood.

This argument on the differential etiology of violent and nonviolent delinquent behavior has conceptual and empirical implications for examining the effect of delinquency on educational attainment. Differential etiology and developmental trajectories of offending may indicate qualitatively distinct educational consequences of offending. Moffitt’s (1993) taxonomic theory of antisocial behavior suggests that the delinquent careers of AL offenders tend to focus on a group behavior (largely nonviolent delinquent behavior) and dissolve at the end of adolescence. Thus, their educational trajectories may not be as severely interrupted as LCP offenders. Alternatively, since those who engage in violent delinquency (i.e., LCP offenders) are more likely to be influenced by adverse family characteristics and environments, the harmful effects of engagement in violent delinquency on education may be attributable to underlying risk factors rather than to delinquency per se. The latter argument has important implications for empirical approaches that examine the effect of delinquency on education. While it seems likely that family background contributes to the significant relationship between delinquency and educational attainment, the effect of violent delinquency on education might be more severely confounded by shared family background than that of nonviolent delinquency because childhood family influences are a stronger predictor of violent delinquency than nonviolent delinquency. In other words, failure to account for family background characteristics may spuriously inflate the effect of violent delinquency to a larger extent than that of nonviolent delinquency.

Potential Mechanisms

Then, how does involvement in delinquency behavior hinder educational outcomes? Although far from conclusive, existing theories and empirical findings have suggested that three broad mechanisms, which are not necessarily mutually exclusive, could possibly help explain this linkage. Those mechanisms include (1) institutional responses to student criminality (disciplinary actions); (2) social-psychological factors (poor relationships with teachers and friends, and school detachment); and (3) disruption of educational progress (reduced academic efforts, educational aspirations, and academic achievement).

First, institutional reactions to delinquency may impede educational attainment (e.g., Kirk and Sampson, 2013). Engagement in delinquent behaviors, particularly in contexts with zero-tolerance policies, often leads to disciplinary actions (e.g., probation, out-of-school suspension, expulsion, etc.), which may adversely affect educational outcomes (Ladd, Birch, & Buhs, 1999; Newcomb, A.F., Bukowski, W.M., Pattee, 1993; Pianta & Stuhlman, 2004; Raffaele Mendez, 2003). Undergoing disciplinary actions may increase the likelihood of repeating grades through increased time out of school (missing classes), and consequently, failure to fulfill requirements for grade promotion (Hirschfield, 2009; UNESCO, 2012). Beyond high school education, disciplinary records could also interrupt an adolescent’s educational opportunities by creating barriers to college education. For example, colleges and universities commonly collect applicants’ high school disciplinary information and disciplinary records are known to play a critical role in obtaining college admissions as well as financial aid (Weissman, 2015).

Second, students who have engaged in delinquent behaviors are known to struggle with building social relationships in school settings. They are more likely to experience conflict and tension with teachers and other school staff members (Lewis, Romi, Katz, & Qui, 2008; Newcomb, A.F., Bukowski, W.M., Pattee, 1993; Sampson & Laub, 1997). Teachers may prefer and support students who do not have a delinquent record. Delinquent students often perceive adverse reactions by teachers and school staff members, which leads to disengagement from school and academic pursuits (Eccles et al., 1993; Freidenfelt Liljeberg, Eklund, Fritz, & af Klinteberg, 2011; Lewis, 2001). Exclusionary and punitive reactions by teachers and school authorities further alienate delinquent students from school resources and learning opportunities (Sampson & Laub, 1997). Although adolescents may engage in problem behavior (such as drinking and smoking) as a way of integrating themselves into groups and attaining a social status (Crosnoe, 2002; Schulenberg et al., 1999), engagement in major delinquent behavior may also jeopardize social bonds and connectedness to friends in school (Asher & Coie, 1991; Coie & Miller-Johnson, 2001). For example, delinquent adolescents tend to have social skill deficits, and are more likely to be rejected by their non-delinquent peers (Coie & Dodge, 1998; Rubin, Bukowski, & Parker, 1998). Moreover, they spend more time with other delinquent peers (who are more likely to approve their delinquent activities) and friends outside school who tend to form an identity in opposition to school and school authorities (Brendgen, Vitaro, & Bukowski, 1998; Flores-González, 2002). Such disrupted social relationships in school are related to school detachment, one of the proximate pathways linking delinquency to educational outcomes (e.g., Bernburg and Krohn, 2003; Hoffmann et al., 2013). Consequentially, delinquent students are likely to be uninterested in conventional activities within school (both academic and extracurricular) (Sabia, 2007), and this lack of participation in school activities may exacerbate their alienation from school (Ream & Rumberger, 2008).

Third, engagement in delinquency may undermine aspirational and behavioral investments in education, one of the important proximate determinants of educational attainment (Siennick & Staff, 2008). Delinquent youths tend to dedicate a great deal of time to unstructured activities with peers, and they may give less school effort than do other youths (Hirschi, 1969; Osgood, Wilson, O’Malley, Bachman, & Johnston, 1996). Moreover, a record of disciplinary actions and poor student-teacher relationships may lead to lower levels of motivation towards completing school work and low educational aspirations in general (Baker, Grant, & Morlock, 2008; Kohlberg et al., 2016). It also seems obvious that these two factors—declines in educational aspirations and lack of school effort—are linked to poor academic achievement, which contains consequences for educational attainment (e.g., Khattab, 2015). The Present Study.

The objective of this study is to examine potential heterogeneity in the effect of juvenile delinquency on educational attainment by types of offending (i.e., violent and nonviolent delinquent behavior). Using family fixed effects model, which offers the ability to account for confounding from all stable family-level background characteristics and environmental exposures shared by siblings, this study disentangles family background effects from the effects of violent and nonviolent delinquency on educational attainment. In particular, this study investigates whether the role of unobserved family heterogeneity in confounding the relationship between delinquency and education differs between violent and nonviolent delinquency.

In addition to separating the effects of delinquency from social background effects, this study further extends existing literature by uncovering potential mechanisms through which engagement in violent and nonviolent delinquent behavior influences educational attainment. Specifically, this study tests three broadly defined mechanisms: (1) institutional responses to student criminality (disciplinary actions); (2) social-psychological factors (poor relationships with teachers and friends, and school detachment); and (3) disruption of educational progress (reduced academic efforts, educational aspirations, and academic achievement).

Data and Methods

Data

The data used in this study come from the National Longitudinal Study of Adolescent to Adult Health (Add Health), a school-based, longitudinal study of the health-related behaviors of adolescents and their outcomes in young adulthood. Beginning with an in-school questionnaire administered to a nationally representative sample of students in grades 7 through 12 in 1994-1995 (Wave 1), the study follows up with a series of in-home interviews of students approximately 1 year (1996; Wave 2), 6 years (2001-2002; Wave 3), and 13 years later (2007-2008; Wave 4). This study uses Wave 1, Wave 2, and Wave 4. Other sources of data include questionnaires for parents, siblings, fellow students, and school administrators. By design, the Add Health survey included a sample stratified by region, urbanicity, school type, ethnic mix, and size.

Of the 20,745 individuals who completed the Wave 1 in home survey, approximately 15,701 were followed longitudinally at Wave 4. In the full sample, individuals with missing school identification numbers were dropped (n = 368), leaving a sample size of 15,333. A major benefit of Add Health is that it contains a sub-sample of siblings who were followed over time. This sample comprises approximately 5,400 individuals in Wave 1, nearly 60% of whom were followed (along with their co-sibling) longitudinally into Wave 4. This results in a final sample size of 2,982 for sibling samples. It should be noted that the sample size for the mechanism analysis is slightly smaller (n = 2,557) than the original sample because 12th graders (in Wave 1) who graduated from high school by Wave 2 were not interviewed in Wave 2 when all mechanism variables (except cumulative GPA) are measured. Despite this sample attrition, using Wave 2 measures as mediating variables is critical to establish temporal ordering of the variables. In addition, this study confirms that using this reduced sample did not affect the results of the main analysis.

Following recommended analytic practices (Allison, 2002), I employed multiple imputation to handle missing values in family income and mother’s education level measured at Wave 1 (about 20% missing data) and a number of mechanism variables measured at Wave 2 (about 5-12% missing data). Multiple imputation was implemented using the chain equations (ICE) procedure in STATA 15.1, and the effect estimates and standard errors reported in this paper are combined estimates from the 10 multiple imputation datasets. However, it is important to note that the missing data on family-level variables such as family income and mother’s education level does not affect the preferred specification of this study (i.e., family fixed effects models) because these models eliminate all characteristics shared by siblings (including family income and parental education) anyway.

Variables

The dependent variables of the study include years of schooling completed and bachelor’s degree completion at Wave 4. The focal predictor variable is the measure of delinquency, which is derived from responses to a number of delinquency items in the Wave 1 data. To assess if the effects of delinquency on educational attainment differ between nonviolent and violent delinquent behavior, the measure of delinquency includes two separate binary variables indicating whether a respondent engaged in nonviolent or violent delinquent behavior. In line with the delinquency literature (Guo et al., 2008; D. Haynie, 2003; D. L. Haynie, 2001), I divide 12 delinquency items into nonviolent and violent categories. The following four items were used to measure nonviolent delinquency: In the past 12 months, how often did you steal something worth less than $50; steal something worth $50 or more; burglarize; and sell drugs. The following eight items were used to measure violent delinquency: In the past 12 months, how often did you get into a serious physical fight; seriously injure another person; threaten to use a weapon on someone; get into a group fight; pull a knife or gun on someone; deliberately damage property; carry a handgun to school or work; and shoot or stab someone.

Several mechanism variables are used in the mediation analysis. Due to space limitations, a detailed description of the variables is presented in Table A1 in the supplementary files. Individual-level sociodemographic controls include gender, age, race/ethnicity, grade level, and an indicator of being the first-born child. In order to reduce further biases resulting from not accounting for cognitive and non-cognitive abilities that are known to determine both delinquency and education (Heckman, Stixrud, & Urzua, 2006), the following extra control variables are used: ability test scores (measured by the Peabody Picture Vocabulary Test) and a measure of low self-control. The items used to create the low self-control measure are listed in Table A2 in the supplementary files (Miller, Barnes, & Beaver, 2011). Family-level control variables include mother’s education, family income, and living in a rural setting. All these control variables come from Wave 1 data.

Analytic Strategy

In this study, I begin with ordinary least squares (OLS) regression using both full and family samples. Then, I add family fixed effects to reduce omitted variables at these levels. For the baseline empirical specifications, I estimate variations of the following model using OLS regression:

Yi=Delinquencyiα+Xiγ+εi (1)

where Yi is the outcome of interest, educational attainment measured at Wave 4, and the vector Xi represents a vector of standard socio-demographic characteristics, additional individual-level controls, and family-level controls. The estimates of Delinquencyi present a set of regression coefficients estimating the difference in the outcome variable between nonviolent and violent delinquent behaviors. In this OLS model, robust standard errors are clustered at the school level.

Eq. (1) cannot account for unobserved factors that might simultaneously affect delinquency and educational attainment. To address the potential biases that may stem from unobserved heterogeneity at the family level, I expand the baseline model to allow for the family fixed effects:

Yif=Delinquencyifβ+Zifδ+θf+εif (2)

where θf denotes a set of family dummies, and the vector Z represents individual level variables that vary between siblings (e.g., gender, age, birth order, etc.). I compare the coefficients of delinquency estimated by OLS and fixed effects models to assess whether baseline models are spuriously driven by omitted variable bias at the family and neighborhood levels. In fixed effects models, robust standard errors are clustered at the family level.

Specifically, the key interest of this analysis is to examine whether and to what extent the inclusion of family fixed effects reduces the estimated coefficients for violent and nonviolent delinquency. This comparison will provide evidence for whether the effects of violent and nonviolent delinquency are spurious. If so, this study also explores how much of the spuriousness is due to shared family characteristics. Additionally, this study aims to uncover potential mechanisms that link delinquency to educational attainment. Relying on fixed effects estimates, the present study examines whether the inclusion of potential mechanism variables explains the observed effects of violent and nonviolent delinquency. Statistical analyses were conducted using STATA 15.1 (StataCorp, 2018).

Results

Summary statistics

Summary statistics for family and full samples are presented in Table 1. In the family sample, respondents completed over 13 years of schooling on average, and 31% of them obtained a bachelor’s degree by Wave 4 of the survey. There is a large portion of respondents who engaged in either violent or nonviolent delinquent behavior: 50% of respondents committed violent delinquent activity while about 25% of respondents engaged in nonviolent delinquent activity. Approximately 20% of students committed both violent and nonviolent delinquent behavior (not shown). Table A3 in the supplementary files presents summary statistics for potential mechanism variables.

Table 1.

Summary statistics, Add Health: family (N ≈ 2,982) and full sample (N ≈ 15,333)

Family sample
Full sample
Min Max
Mean SD Mean SD
Dependent variables
Years of schooling 13.30 1.95 13.38 1.92 8.00 19.00
Bachelor’s degree completion 0.31 0.46 0.32 0.47 0.00 1.00
Delinquency
Nonviolent delinquency 0.24 0.43 0.25 0.43 0.00 1.00
Violent delinquency 0.50 0.50 0.50 0.50 0.00 1.00
Nonviolent delinquent behavior
Stealing something worth $50 or more 0.05 0.22 0.05 0.22 0.00 1.00
Stealing something worth less than $50 0.20 0.40 0.20 0.40 0.00 1.00
Burglarizing 0.05 0.22 0.05 0.22 0.00 1.00
Selling drugs 0.07 0.26 0.07 0.26 0.00 1.00
Violent delinquent behavior
Getting into a serious physical fight 0.32 0.46 0.31 0.46 0.00 1.00
Seriously injuring another person 0.18 0.38 0.18 0.38 0.00 1.00
Being seriously injured in a physical fight 0.08 0.28 0.09 0.28 0.00 1.00
Threatening to use a weapon on someone 0.05 0.22 0.04 0.20 0.00 1.00
Getting into a group fight 0.20 0.40 0.19 0.40 0.00 1.00
Damaging property 0.19 0.39 0.18 0.38 0.00 1.00
Carrying a weapon 0.05 0.23 0.06 0.23 0.00 1.00
Shooting or stabbing someone 0.02 0.13 0.02 0.13 0.00 1.00
Pulling a knife or gun on someone 0.04 0.21 0.05 0.21 0.00 1.00
Control variables
Female 0.52 0.50 0.53 0.50 0.00 1.00
Age 16.08 1.72 16.11 1.72 12.00 21.00
White 0.59 0.49 0.55 0.50 0.00 1.00
Black 0.21 0.41 0.23 0.42 0.00 1.00
Hispanic 0.13 0.34 0.15 0.36 0.00 1.00
Other race/ethnicity 0.06 0.24 0.07 0.25 0.00 1.00
Grade = 7 0.14 0.35 0.13 0.34 0.00 1.00
Grade = 8 0.14 0.35 0.13 0.34 0.00 1.00
Grade = 9 0.19 0.39 0.18 0.38 0.00 1.00
Grade = 10 0.18 0.39 0.19 0.39 0.00 1.00
Grade = 11 0.17 0.38 0.18 0.39 0.00 1.00
Grade = 12 0.14 0.35 0.16 0.37 0.00 1.00
Grade = missing 0.03 0.16 0.02 0.14 0.00 1.00
First-born 0.37 0.48 0.51 0.50 0.00 1.00
Standardized PVT score −0.00 0.93 0.07 0.95 −5.71 3.06
Low self-control 0.06 0.98 0.01 0.99 −3.73 5.26
Mother’s education 13.08 2.37 13.21 2.38 0.00 17.00
Family income 0.45 0.52 0.47 0.51 0.00 9.99
Rural status 0.28 0.45 0.26 0.44 0.00 1.00
Criminal justice contact by age 18 0.12 0.32 0.12 0.33 0.00 1.00
Open in a new tab

Note. Summary statistics do not contain imputed values.

The fixed effects model would not work as intended if there was a great deal of concordance between siblings (i.e., they do not differ in observed characteristics). However, consistent with prior research using the same data (Fletcher, 2013; Kim, 2016), there is evidence of sufficiently discordant families. Table A4 in the supplementary files demonstrates that almost 60% of the siblings in the sample are discordant in terms of delinquency status (Column 1). In addition, more than 40% of the siblings were substantially discordant in terms of years of schooling (i.e., more than one standard deviation away from each other) (Column 2). Column 3 also suggests that a great deal of “unexplained” variation in the outcome measures remains even after measure and unmeasured family characteristics are controlled for.

Effects of violent and nonviolent delinquency on educational attainment

Table 2 presents OLS estimates of the effects of juvenile delinquency on years of schooling. All coefficients for control variables are omitted, but are available in the supplementary files (Tables B1-B3). Model 1 shows that those who committed either violent or nonviolent delinquent behavior are more likely to complete fewer years of schooling than those who did not engage in delinquent behavior. Results indicate that the effects of violent delinquency on schooling are greater than those of nonviolent delinquency (−0.393 vs. −0.157). In Model 2, using the family sample, I re-estimate the Model 1 specification. Results show that despite reduced sample size the relationship between violent and nonviolent delinquency and schooling remains considerably similar. These results are generally consistent with prior research that has argued that engagement in any delinquent behavior (whether violent or nonviolent) is negatively associated with educational attainment.

Table 2.

Regression of years of schooling on juvenile delinquency

Dependent variable (W4) Years of schooling Years of schooling Years of schooling Years of schooling
(1) (2) (3) (4)
Delinquency
Nonviolent delinquency −0.157*** −0.185** −0.207* −0.185*
(0.031) (0.069) (0.092) (0.091)
Violent delinquency −0.393*** −0.413*** −0.079 −0.074
(0.031) (0.064) (0.089) (0.088)
Criminal justice contact by age 18 −0.319*
(0.127)
Sample Full Family Family Family
N(Students) 15,333 2,982 2,982 2,982
N(Family clusters) – – 1,454 1,454
Family fixed effects No No Yes Yes
Individual-level controls Yes Yes Yesa Yesa
Family-level controls Yes Yes No No
Robust S.E. School School Family Family
Open in a new tab

Note.

a

Race/ethnicity is dropped because siblings have the same race/ethnicity. Individual-level control variables include gender, age, race/ethnicity, grade level, first-born child, standardized PVT score, and low self-control. Family-level control variables include mother’s education, family income, and rural status.

+

p < 0.1,

*

p < 0.05,

**

p < 0.01,

***

p < 0.001

To address shared unobserved heterogeneity at the family level, I examine sibling comparisons by controlling for family fixed effects (Model 3). When all measured and unmeasured family-level heterogeneity is removed, the coefficient of the effect of violent delinquency only is reduced by more than 80% (Model 2 vs. Model 3). This substantial reduction makes the effect of engagement in violent delinquency on schooling statistically insignificant, suggesting that a substantial portion of the effect of violent delinquency is attributable to unobserved family background characteristics. In contrast, after controlling for family fixed effects, the coefficient for nonviolent delinquency rather slightly increases and remains statistically significant, suggesting that the effects of nonviolent delinquency is not confounded by stable family-level characteristics shared by siblings. These findings support the argument of this study that the effects of violent delinquency are more vulnerable to unobserved family heterogeneity than those of nonviolent delinquency.

Lastly, Model 4 shows that effects of engagement in delinquent behavior are robust to controls for criminal justice contact by age 18. These results suggest that the effect of delinquency on education is not confined to those who were involved with the criminal justice system. To interpret the findings, those who engaged in nonviolent delinquent behavior are likely to complete nearly 0.19 fewer years of schooling (i.e., one-tenth of a standard deviation) than non-offenders.

In Table 3, bachelor’s degree completion is used as an outcome. Results are extremely similar with the ones in Table 2. Models 1 and 2 show that results from the full and family samples are considerably similar. The effects of delinquency on bachelor’s degree completion seem to be largely driven by those who engaged in violent delinquent behavior (−0.093 vs. −0.034 for Model 1, −0.089 vs. −0.024 for Model 2). However, the inclusion of family fixed effects reduces the effect of violent delinquency by 82%, but not the effect of nonviolent delinquency (Model 3). Results suggest that the effect of engagement in violent delinquency is substantially confounded by family background characteristics. The effect of engagement in nonviolent delinquent behavior, however, is found to be robust to controls for shared family background (Model 3) as well as criminal justice contact (Model 4). To interpret the results, engagement in nonviolent delinquency reduces the probability of obtaining a bachelor’s degree by about 4 percentage points.

Table 3.

Regression of bachelor’s degree completion on juvenile delinquency

Dependent variable (W4) Bachelor’s degree completion Bachelor’s degree completion Bachelor’s degree completion Bachelor’s degree completion
(1) (2) (3) (4)
Delinquency
Nonviolent delinquency only −0.034*** −0.024 −0.045* −0.042+
(0.007) (0.018) (0.022) (0.022)
Violent delinquency only −0.093*** −0.089*** −0.016 −0.015
(0.008) (0.017) (0.021) (0.021)
Criminal justice contact by age 18 −0.051+
(0.029)
Sample Full Family Family Family
N(Students) 15,333 2,982 2,982 2,982
N(Family clusters) – – 1,454 1,454
Family fixed effects No No Yes Yes
Individual-level controls Yes Yes Yesa Yesa
Family-level controls Yes Yes No No
Robust S.E. School School Family Family
Open in a new tab

Note.

a

Race/ethnicity is dropped because siblings have the same race/ethnicity. Individual-level control variables include gender, age, race/ethnicity, grade level, first-born child, standardized PVT score, and low self-control. Family-level control variables include mother’s education, family income, and rural status.

+

p < 0.1,

*

p < 0.05,

**

p < 0.01,

***

p < 0.001

Uncovering potential mechanisms linking delinquency to education

In this section, I examine three potential mechanisms through which delinquency might affect educational attainment. For each mechanism, several relevant measures are investigated. In order to establish temporal ordering of the relationship, all mechanism variables are measured at Wave 2. All estimations within Table 4 use the family sample while controlling for family fixed effects. Model 1 presents the baseline results, wherein the coefficients of the estimated effects are slightly larger than the one in Model 4 of Table 2 possibly due to the sample attrition—i.e., the exclusion of the 12th graders who graduated from high school by Wave 2. First, I examine whether delinquent students have fewer years of schooling than non-offenders because they are more likely to receive disciplinary actions. Model 2 controls for whether respondents received suspension or expulsion. Results show that both disciplinary actions are not associated with years of schooling, and the inclusion of these variables does not significantly alter the estimated coefficient of nonviolent delinquency.

Table 4.

Potential mechanisms for association between juvenile delinquency and years of schooling completed

Dependent variable (W4) Years of schooling Years of schooling Years of schooling Years of schooling Years of schooling Years of schooling Years of schooling
(1) (2) (3) (4) (5) (6) (7)
Delinquency
Nonviolent delinquency −0.258* −0.254* −0.257* −0.270* −0.267* −0.208+ −0.233+
(0.120) (0.122) (0.120) (0.120) (0.120) (0.120) (0.119)
Violent delinquency −0.003 0.010 0.001 −0.014 0.002 0.055 0.034
(0.115) (0.116) (0.114) (0.114) (0.114) (0.112) (0.112)
Criminal justice contact by age 18 −0.326+ −0.302+ −0.314+ −0.322+ −0.330+ −0.261 −0.226
(0.171) (0.170) (0.172) (0.171) (0.169) (0.163) (0.164)
1. Education impeding stigma
Disciplinary action
Suspension −0.199
(0.190)
Expulsion −0.108
(0.397)
2. Social-psychological factors
Student-teacher relationship
Difficulty in getting along with teachers 0.014
(0.057)
Teachers care about you 0.043
(0.050)
Teachers treat students fairly 0.045
(0.045)
Relationship with friends
Difficulty in getting along with students −0.024
(0.054)
Friends care about you −0.046
(0.055)
Students are prejudiced 0.062
(0.045)
School attachment
Feel like part of school 0.092
(0.061)
Happy to be at school −0.025
(0.049)
Feel close to people at school 0.048
(0.060)
3. Human capital disruption
Educational effort and aspirations
Truancy −0.311**
(0.114)
Difficulty in paying attention in school −0.054
(0.050)
Difficulty in getting homework done −0.004
(0.044)
I want to go to college 0.263***
(0.047)
Academic achievement
Cumulative GPA 0.585***
(0.092)
Sample Family Family Family Family Family Family Family
N(Students) 2,235 2,235 2,235 2,235 2,235 2,235 2,235
N(Family clusters) 1,090 1,090 1,090 1,090 1,090 1,090 1,090
Family fixed effects Yes Yes Yes Yes Yes Yes Yes
Individual-level controls Yesa Yesa Yesa Yesa Yesa Yesa Yesa
Family-level controls No No No No No No No
Robust S.E. Family Family Family Family Family Family Family
Open in a new tab

Note.

a

. Race/ethnicity is dropped because siblings have the same race/ethnicity. Individual-level control variables include gender, age, race/ethnicity, grade level, first-born child, standardized PVT score, and low self-control. Family-level control variables include mother’s education, family income, and rural status.

+

p < 0.1,

*

p < 0.05,

**

p < 0.01,

***

p < 0.001

Second, the social-psychological pathway linking delinquency to education is investigated. This pathway consists of three seemingly separate sources of psychological distress in the school setting, including student-teacher relationships (Model 3), relationships with friends (Model 4), and school attachment (Model 5). While a large number of relevant measures is considered, no measure is found to be associated with years of schooling. Results from Models 3-5 imply that social-psychological factors are not the driving force that mediates the relationship between delinquency and educational attainment.

Third, I investigate whether disruption of educational progress plays a mediating role in the relationship between delinquency and education. Two specific pathways include (1) a decrease in behavioral and aspirational investment in education and (2) a decrease in academic achievement. In Model 6, truancy and college aspirations are found to be strongly associated with years of schooling, and the inclusion of these variables attenuates the effect of engagement in nonviolent delinquent behavior by 20%. In addition, Model 7 shows that cumulative GPA is strongly associated with years of schooling, and the inclusion of the measure reduces the estimated coefficient of nonviolent delinquency by 10%. Inclusion of a set of mechanism variables that represent processes of disruption of educational progress (i.e., Models 6-7) is found to attenuate the effect of engagement in nonviolent delinquent behavior by almost 30% (results not shown).

Discussion

This study extends existing literature by examining whether the effects of delinquency on educational attainment differ between violent and nonviolent offending. In order to rule out the plausible alternative explanation that the effect of delinquency is spuriously driven by shared family characteristics, I use sibling comparisons to remove all measured and unmeasured family background characteristics. Then, using more credible estimates of the effect of violent and nonviolent delinquency, this study explores three broad mechanisms linking delinquency to educational attainment: institutional reactions, social-psychological factors, and disruption of educational progress.

The OLS estimates show that the negative effects of delinquency are much greater among those who engaged in violent delinquent behavior compared to those who engaged in nonviolent delinquent behavior. However, fixed effects estimates demonstrate that accounting for unobserved heterogeneity at the family level substantially reduces the effect of violent delinquency, but not the effect of nonviolent delinquency. The key finding of this study is that the effect of violent delinquent behavior is more vulnerable to unobserved family background characteristics than that of nonviolent delinquency. This is supported by existing evidence that violent offending is determined largely by early childhood factors (e.g., genetic factors and parent and family characteristics) (Barker & Maughan, 2009; Barnes, Beaver, & Boutwell, 2011; Chen & Jaffee, 2015).

Results suggest that the effect of engagement in violent delinquency on education is driven spuriously by unobserved family background, whereas the effect of nonviolent delinquency appears to be robust to adjustment for family fixed effects. Moreover, these results suggest that the effect of delinquency on educational attainment is more attributable to nonviolent delinquent behavior than violent delinquent behavior. It does not mean, however, that engaging in violent delinquent behavior is not harmful for educational attainment, but rather, it lends support to the notion that individuals who are most likely to engage in violent delinquency may come from certain types of backgrounds (e.g., families with low socioeconomic status or poor parenting), or have certain types of student profiles (e.g., low standardized test scores or negative personality traits) that result in poor educational attainment. Moreover, the current study found that these results are robust to controlling for criminal justice contact, providing further support for the argument that engagement in delinquency per se influences educational outcomes (e.g., Ward and Williams, 2015).

Results from the analysis on proposed channels through which delinquency impacts educational attainment suggest that the effect of engagement in delinquent activity on educational attainment occurs in part (about 30%) through disruption of educational progress (i.e., decline in educational effort, college aspirations, and academic progress), rather than through institutional responses to student delinquency (i.e., disciplinary actions) and social-psychological processes. In general, this finding is consistent with a couple of prior studies (Hjalmarsson, 2008; Kirk & Sampson, 2013) that supported disruption of educational progress as a key mechanism, but not social-psychological processes, though these studies investigated potential mechanisms underlying the link between criminal justice contact (e.g., arrest, incarceration) and education.

This study contains several limitations that should be acknowledged. First, although this study rules out important family background and childhood experiences shared by siblings, it is unable to account for confounding characteristics that are idiosyncratic to each sibling. Even if siblings share many things, there are family-level factors that they do not share. For example, parents may treat siblings differently. Siblings are not genetically identical unless they are identical twins. In school, they are in different grades and have similar but not identical teachers and friends. Therefore, estimates derived from family fixed effects models can still be biased due to individual heterogeneity. While this study adds further controls to account for important sources of individual heterogeneity, such as cognitive and non-cognitive abilities that are known to determine both delinquency and education (Heckman et al., 2006), it is impossible to completely remove the problem of selection on other observed and unobserved individual characteristics.

Second, there are some issues associated with using family-fixed effects models. Sibling comparison models could possibly reduce exogenous as well as endogenous variation. If endogenous variation is a large portion of the remaining between-sibling variation, the estimates obtained from the sibling models could suffer as much endogeneity inconsistency as the baseline model (Bound & Solon, 1999). Moreover, a careful interpretation of the results is required because the approach focuses only on the population of siblings who have discordant values for the variables of interest (R. Moffitt, 2005). For example, the effects of delinquency on educational attainment differ between adolescents with and without siblings. If this is the case, family fixed effects estimators will yield biased estimates of the population-level effects.

Despite these limitations, the current study lends support to the argument that engagement in delinquent behavior deteriorates educational attainment, which is consistent with the majority of prior research. However, the effects of delinquency on educational attainment are largely driven by those who engaged in nonviolent delinquent behavior. This article provides novel contributions to the study of education and criminology by revealing how the relationship between violent delinquency and education is spurious owing to unobserved family-level characteristics. Intriguingly, findings suggest that engagement in violent delinquent activity has no impact on educational attainment. These results may run counter to our expectations. However, this study argues that those who engaged in violent delinquent behavior are, of course, likely to show poor educational attainment, but not because they committed violent delinquent activity but because of underlying risk factors that are more closely related to family background.

In closing, I briefly discuss the theoretical implications that stem from this study. The sizable reduction in the coefficient of violent delinquency (but not that of nonviolent delinquency) in the family fixed effects models has notable theoretical implications. My findings are broadly congruent with a growing body of recent delinquency and violence literature that hints at the possibility that genetic and environmental influences affect the development of different criminal behaviors differently (e.g., violent vs. nonviolent crime) (Barnes, 2013; T. E. Moffitt, 1993; Zheng & Cleveland, 2015). In order to better understand whether and how engagement in delinquent behavior shapes various adolescent outcomes including educational attainment, researchers need to take into account the roles of genetic and environmental influences in the relationship. Specifically, it is crucial to expand our understanding of the degree to which genetic and environmental factors explain variance in the etiology of the different offending patterns (e.g., Barnes et al., 2011).

Finally, the findings of this study are relevant to policy. Distinguishing the effects of violent from those of nonviolent delinquency may be an important contribution because identifying who among delinquents is at greater risk of low educational attainment is critical in developing and targeting interventions. For example, while policy interventions aimed at preventing nonviolent delinquent behavior among adolescents will be highly effective for improving educational outcomes, in order to reduce the educational consequences of violent delinquency, interventions may have to focus on addressing more fundamental family predictors of violent delinquency (e.g., socioeconomic disadvantage, parenting style, marital disruption, parental incarceration, etc.). A further contribution of this study is to provide evidence on a number of important mechanisms through which delinquency affects educational attainment. Interventions to support delinquent students’ educational efforts and promote their college aspirations may help reduce the adverse effects of delinquency on educational attainment.

Supplementary Material

supplementary files

References

  1. Agan A. Non-cognitive skills and crime. Unpublished Manuscript University of Chicago. 2011:1–23. [Google Scholar]
  2. Allison PD. Missing Data. Thousand Oaks, CA: Sage Publications; 2002. [Google Scholar]
  3. Asher SR, Coie JD. Peer Rejection in Childhood. New York, NY: Cambridge University Press; 1991. [Google Scholar]
  4. Baker Ja, Grant S, Morlock L. The teacher-student relationship as a developmental context for children with internalizing or externalizing behavior problems. School Psychology Quarterly. 2008;23(1):3–15. [Google Scholar]
  5. Barker ED, Maughan B. Differentiating early-onset persistent versus childhood-limited conduct problem youth. American Journal of Psychiatry. 2009;166(8):900–908. doi: 10.1176/appi.ajp.2009.08121770. [DOI] [PubMed] [Google Scholar]
  6. Barnes JC. Analyzing the Origins of Life-Course-Persistent Offending: A Consideration of Environmental and Genetic Influences. Criminal Justice and Behavior. 2013;40(5):519–540. [Google Scholar]
  7. Barnes JC, Beaver KM, Boutwell BB. Examining the Genetic Underpinnings to Moffitt’s Developmental Taxonomy: A Behavioral Genetic Analysis. Criminology. 2011;49(4):923–954. [Google Scholar]
  8. Bernburg JG, Krohn MD. Labeling, life chances, and adult crime: The direct and indirect effects of official intervention in adolescence on crime in early adulthood. Criminology. 2003;41(4):1287–1318. [Google Scholar]
  9. Bound J, Solon G. Double trouble: on the value of twins-based estimation of the return to schooling. Economics of Education Review. 1999;18:169–182. [Google Scholar]
  10. Brendgen M, Vitaro F, Bukowski WM. Affiliation with Delinquent Friends: Contributions of Parents, Self-Esteem, Delinquent Behavior, and Rejection by Peers. The Journal of Early Adolescence. 1998;18(3):244–265. [Google Scholar]
  11. Burt SA, Neiderhiser JM. Aggressive versus nonaggressive antisocial behavior: distinctive etiological moderation by age. Developmental Psychology. 2009;45(4):1164–1176. doi: 10.1037/a0016130. [DOI] [PMC free article] [PubMed] [Google Scholar]
  12. Cernkovich Sa, Giordano PC. Family Relationships and Delinquency. Criminology. 1987;25(2):295–319. [Google Scholar]
  13. Chen FR, Jaffee SR. The Heterogeneity in the Development of Homotypic and Heterotypic Antisocial Behavior. Journal of Developmental and Life-Course Criminology. 2015;1(3):269–288. [Google Scholar]
  14. Coie JD, Dodge KA. Aggression and antisocial behavior. In: Eisenberg N, editor. Handbook of child psychology: Social, emotional, and personality development. New York: Wiley; 1998. pp. 779–862. [Google Scholar]
  15. Coie JD, Miller-Johnson S. Peer factors and interventions. In: Loeber R, Farrington DP, editors. Child delinquents: Development, interventions and service needs. Thousand Oaks, CA: SAGE Publications; 2001. pp. 191–209. [Google Scholar]
  16. Crosnoe R. Academic and Health-Related Trajectories in Adolescence: The Intersection of Gender and Athletics. Journal of Health and Social Behavior. 2002;43(3):317. [PubMed] [Google Scholar]
  17. Eccles JS, Midgley C, Wigfield A, Buchanan CM, Reuman D, Flanagan C, Mac Iver D. Development during adolescence: The impact of stage-environment fit on young adolescents’ experiences in schools and in families. American Psychologist. 1993;48(2):90–101. doi: 10.1037//0003-066x.48.2.90. [DOI] [PubMed] [Google Scholar]
  18. Fagan J, Wexler S. Family Origins of Violent Delinquents. Criminology. 1987;25(3):643–670. [Google Scholar]
  19. Fletcher J. The effects of personality traits on adult labor market outcomes: Evidence from siblings. Journal of Economic Behavior & Organization. 2013;89:122–135. [Google Scholar]
  20. Flores-González N. School kids/street kids : identity development in Latino students. New York, NY: Teachers College Press; 2002. [Google Scholar]
  21. Freidenfelt Liljeberg J, Eklund JM, Fritz MV, af Klinteberg B. Poor school bonding and delinquency over time: Bidirectional effects and sex differences. Journal of Adolescence. 2011;34(1):1–9. doi: 10.1016/j.adolescence.2010.03.008. [DOI] [PubMed] [Google Scholar]
  22. Guo G, Roettger ME, Cai T. The Integration of Genetic Propensities into Social-Control Models of Delinquency and Violence among Male Youths. American Sociological Review. 2008;73(4):543–568. [Google Scholar]
  23. Hannon L. Poverty, Delinquency, and Educational Attainment: Cumulative Disadvantage or Disadvantage Saturation? Sociological Inquiry. 2003;73(4) [Google Scholar]
  24. Haynie D. Contexts of Risk? Explaining the Link between Girls’ Pubertal Development and Their Delinquency Involvement. Social Forces. 2003;82(1):355–397. [Google Scholar]
  25. Haynie DL. Delinquent Peers Revisited: Does Network Structure Matter? American Journal of Sociology. 2001;106(4):1013–1057. [Google Scholar]
  26. Heckman J, Pinto R, Savelyev P. Understanding the mechanisms through which an influential early childhood program boosted adult outcomes. American Economic Review. 2012;103(6):2052–2086. doi: 10.1257/aer.103.6.2052. [DOI] [PMC free article] [PubMed] [Google Scholar]
  27. Heckman J, Stixrud J, Urzua S. The effects of cognitive and noncognitive abilities on labor market outcomes and social behavior. Journal of Labor Economics. 2006 May;24:411–482. [Google Scholar]
  28. Hirschfield P. Another Way Out: The Impact of Juvenile Arrests on High School Dropout. Sociology of Education. 2009;82(4):368–393. [Google Scholar]
  29. Hirschi T. Causes of delinquency. Berkeley, CA: University of California; 1969. [Google Scholar]
  30. Hjalmarsson R. Criminal justice involvement and high school completion. Journal of Urban Economics. 2008;63(2):613–630. [Google Scholar]
  31. Hoffmann JP, Erickson LD, Spence KR. Modeling The Association Between Academic Achievement And Delinquency: An Application Of Interactional Theory. Criminology. 2013;51(3):629–660. [Google Scholar]
  32. Khattab N. Students’ aspirations, expectations and school achievement: what really matters? British Educational Research Journal. 2015;41(5):731–748. [Google Scholar]
  33. Kim J. Personality traits and body weight: Evidence using sibling comparisons. Social Science & Medicine. 2016;163:54–62. doi: 10.1016/j.socscimed.2016.06.054. [DOI] [PMC free article] [PubMed] [Google Scholar]
  34. Kirk DS, Sampson RJ. Juvenile Arrest and Collateral Educational Damage in the Transition to Adulthood. Sociology of Education. 2013;88(1):36–62. doi: 10.1177/0038040712448862. [DOI] [PMC free article] [PubMed] [Google Scholar]
  35. Kohlberg L, Hersh RH, Theory S, Practice I, Apr MD, Taylor P, Kohlberg L. Linked references are available on JSTOR for this article : Moral Development : A Review of the Theory. 2016;16(2):53–59. [Google Scholar]
  36. Lacourse E, Côté SM, Nagin DS, Vitaro F, Brendgen M, Tremblay RE. A longitudinal-experimental approach to testing theories of antisocial behavior development. Development and Psychopathology. 2002;14(2002):909–924. doi: 10.1017/s0954579402004121. [DOI] [PubMed] [Google Scholar]
  37. Ladd GW, Birch SH, Buhs ES. Children’s social and scholastic lives in kindergarten: Related spheres of influence? Child Development. 1999;70(6):1373–1400. doi: 10.1111/1467-8624.00101. [DOI] [PubMed] [Google Scholar]
  38. Laub JH, Sampson RJ. Unraveling Families and Delinquency: A Reanalysis of The Gluecks’ Data. Criminology. 1988;26(3):355–380. [Google Scholar]
  39. Lewis R. Classroom discipline and student responsibility: the students’ view. Teaching and Teacher Education. 2001;17(3):307–319. [Google Scholar]
  40. Lewis R, Romi S, Katz YJ, Qui X. Students’ reaction to classroom discipline in Australia, Israel, and China. Teaching and Teacher Education. 2008;24(3):715–724. [Google Scholar]
  41. Loeber R, Farrington DP, editors. Serious and violent juvenile offenders: Risk factors and successful interventions. Thousand Oaks, CA: Sage Publications; 1998. [Google Scholar]
  42. Matsueda R, Heimer K. Race, family structure, and delinquency: A test of differential association and social control theories. American Sociological Review 1987 [Google Scholar]
  43. Miller H, Barnes J, Beaver K. Self-control and health outcomes in a nationally representative sample. American Journal of Health Behavior. 2011;35(1):15–27. doi: 10.5993/ajhb.35.1.2. [DOI] [PubMed] [Google Scholar]
  44. Moffitt R. Remarks on the Analysis of Causal Relationships in Population Research. Demography. 2005;42(1):91–108. doi: 10.1353/dem.2005.0006. [DOI] [PubMed] [Google Scholar]
  45. Moffitt TE. Adolescence-limited and life-course-persistent antisocial behavior: a developmental taxonomy. Psychological Review. 1993;100(4):674–701. [PubMed] [Google Scholar]
  46. Moffitt TE, Caspi A, Harrington H, Milne BJ. Males on the life-course-persistent and adolescence-limited antisocial pathways: Follow-up at age 26 years. Development and Psychopathology. 2002;14:179–207. doi: 10.1017/s0954579402001104. [DOI] [PubMed] [Google Scholar]
  47. Monk-Turner E. Effects of high school delinquency on educational attainment and adult occupational status. Sociological Perspectives. 1989;32(3):413–418. [Google Scholar]
  48. Newcomb AF, Bukowski WM, Pattee L. Children’s Peer Relations: a Meta-Analytic Review of Popular, Rejected, Neglected, Controversial, and Average Sociometric Status. Psychological Bulletin. 1993;113(1):99–128. doi: 10.1037/0033-2909.113.1.99. [DOI] [PubMed] [Google Scholar]
  49. Nilsson A, Estrada F. Established or excluded? A longitudinal study of criminality, work and family formation. European Journal of Criminology. 2011;8(3):229–245. [Google Scholar]
  50. Osgood DW, Wilson JK, O’Malley PM, Bachman JG, Johnston LD. Routine Activities and Individual Deviant Behavior. American Sociological Review. 1996;61(4):635–655. [Google Scholar]
  51. Ou SR, Mersky JP, Reynolds AJ, Kohler KM. Alterable Predictors of Educational Attainment, Income, and Crime: Findings from an Inner‐City Cohort. Social Service Review. 2007;81(1):85–128. [Google Scholar]
  52. Patterson G, Stouthamer-Loeber M. The Correlation of Family Management Practices and Delinquency. Child Development. 1984;55(4):1299–1307. [PubMed] [Google Scholar]
  53. Pianta RC, Stuhlman MW. Teacher-child relationships and children’s success in the first years of school. School Psychology Review. 2004;33(3):444–458. [Google Scholar]
  54. Piquero AR, Farrington DP, Blumstein A. Key issues in criminal career research: new analyses of the Cambridge Study in Delinquent Development Cambridge studies in criminology. New York, NY: Cambridge University Press; 2007. [Google Scholar]
  55. Raffaele Mendez LM. Predictors of suspension and negative school outcomes: A longitudinal investigation. New Directions for Youth Devlopment. 2003;(99):17–34. doi: 10.1002/yd.52. [DOI] [PubMed] [Google Scholar]
  56. Ream RK, Rumberger RW. Student Engagement, Peer Social Capital, and School Dropout Among Mexican American and Non-Latino White Students. Sociology of Education. 2008;81(2):109–139. [Google Scholar]
  57. Rubin KH, Bukowski WM, Parker JG. Peer Interactions, Relationships, and Groups. In: Eisenberg N, editor. Handbook of child psychology: Social, emotional, and personality development. New York: Wiley; 1998. pp. 619–700. [Google Scholar]
  58. Sabia JJ. Early Adolescent Sex and Diminished School Attachment: Selection or Spillovers? Southern Economic Journal. 2007;74(1):239–268. [Google Scholar]
  59. Sampson RJ, Laub JH. A Life-Course Theory of Cumulative Disvantage and the Stability of Delinquency. In: Thornberry TP, editor. Advances in criminological theory, volume 7: Developmental theories of crime and delinquency. New Brunswick, NJ: Transaction; 1997. pp. 133–161. [Google Scholar]
  60. Schulenberg J, Maggs JL, Dielman TE, Leech SL, Kloska DD, Shope JT, Laetz VB. On peer influences to get drunk: A panel study of young adolescents. Merrill-Palmer Quarterly (1982-) 1999:108–142. [Google Scholar]
  61. Siennick S, Staff J. Explaining the educational deficits of delinquent youths. Criminology. 2008:6207. [Google Scholar]
  62. StataCorp. Stata statistical software: Release 15.1. College Station (TX): StataCorp LP; 2018. [Google Scholar]
  63. Tanner J, Davies S, O’Grady B. Whatever Happened to Yesterday’s Rebels? Longitudinal Effects of Youth Delinquency on Education and Employment. Social Problems. 1999;46(2):250–274. [Google Scholar]
  64. UNESCO. Opportunities lost: The impact of grade repetition and early school leaving 2012 [Google Scholar]
  65. Ward S, Williams J. Does Juvenile Delinquency Reduce Educational Attainment? Journal of Empirical Legal Studies. 2015;12(4):716–756. [Google Scholar]
  66. Webbink D, Koning P, Vujic S, Martin NG. Why Are Criminals Less Educated than Non-Criminals? Evidence from a Cohort of Young Australian Twins. Journal of Law, Economics, and Organization. 2013;29(1):115–144. [Google Scholar]
  67. Weissman M. Education Suspended: The Use of High School Disciplinary Records in College Admissions. New York: 2015. [Google Scholar]
  68. Wells L, Rankin J. Families and delinquency: A meta-analysis of the impact of broken homes. Social Problems. 1991;38(1):71–93. [Google Scholar]
  69. Zheng Y, Cleveland HH. Differential genetic and environmental influences on developmental trajectories of antisocial behavior from adolescence to young adulthood. Journal of Adolescence. 2015;45:204–213. doi: 10.1016/j.adolescence.2015.10.006. [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

supplementary files
NIHMS976244-supplement-supplementary_files.docx (63KB, docx)

RESOURCES