Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2021 Jul 13.
Published in final edited form as: Justice Q. 2019 Jul 12;37(4):739–760. doi: 10.1080/07418825.2019.1625428

The Effect of School Discipline on Offending across Time

Thomas J Mowen a, John J Brent b, John H Boman IV a
PMCID: PMC8277153  NIHMSID: NIHMS1706827  PMID: 34262239

Abstract

Despite decreases in offending and victimization in schools across the United States, many schools continue to use exclusionary discipline. Although school punishment has been tied to a variety of negative outcomes, the link between suspension and offending remains unclear. Using data from the National Longitudinal Survey of Youth 1997, this study examines the extent to which school punishment contributes to within-individual increases in offending across time and/or amplifies offending between-individuals. Results of a series of cross-lagged dynamic fixed-effects panel models reveal that school suspensions contribute to within-individual increases in offending. This relationship remains even when accounting for the effect of baseline levels of offending on future offending. Further, repeated suspensions amplify offending differences between-individuals.

Keywords: School punishment, discipline, deviance amplification, life-course

Introduction

Numerous scholars have recently documented that school discipline has become increasingly punitive as a consequence of the adoption of criminal justice personnel, logics, and strategies (Hirschfield, 2008; Kupchik, 2010; Nolan, 2011; Simon, 2007). The “criminalization of school discipline” (Hirschfield, 2008) places many students at risk of experiencing punishment (Kupchik, 2010), and about one-third of all students in the United States will receive at least one suspension by the time they graduate from high school (Shollenberger, 2015). Research has examined some negative outcomes of the use of exclusionary discipline including decreases in school performance (Rausch and Skiba, 2005), interruption of family relationships (Kupchik, 2016), and increased contact with the criminal justice system (Fabelo et al., 2011; Jaggers et al., 2016; Losen & Martinez, 2013). Yet, the relationship between school suspension and offending is much less clear. In a recent review, Hirschfield (2018, p. 162) referred to the relationship between school punishment and crime as “the least studied” area of research on school discipline.

Unraveling the link between school punishment and offending behaviors requires theoretical advancements that draw from multiple perspectives. First, prior research finds that formal sanctions can contribute to increased—not decreased—offending (Wiley & Esbensen, 2016). This process, referred to as “deviance amplification,” highlights that official interventions like arrest can confer stigmatizing labels to a person (see, generally, Paternoster & Iovanni, 1989) and contribute to increased offending through a labeling process (Bernburg & Krohn, 2003). Second, scholars have recently drawn from the life-course perspective (Laub & Sampson, 1993; Sampson & Laub, 1993) to understand how school suspensions relate to odds of arrest during adolescence (Mowen & Brent, 2016). Invoking the concept of a “turning point,” research also finds that youth who experience a suspension are placed at an increased risk of experiencing an arrest as they move through adolescence and into young adulthood (Mowen & Brent, 2016). As such, the concepts of deviance amplification (drawn largely from labeling theory) and turning points (drawn from the life-course perspective) offer highly complementary approaches to understanding the relationship between school suspension and offending behaviors. Thus, to advance our theoretical understanding of the link between suspension and offending, we conceptualize school suspensions as a key life event—a turning point—that may result in deviance amplification.

A recent and rapidly expanding group of studies advance evidence that suspensions may operate as a turning point toward deviance amplification. To date, at least three studies have established that students who are suspended report significantly higher levels of offending later in life than students who are not suspended (Hemphill et al., 2006; Rosenbaum, 2018; Wolf & Kupchik, 2017). While these important findings begin to shed light on the “the least studied” (Hirschfield, 2018) area on school discipline, they stop short of addressing two crucial, central questions. First, to what extent do school suspensions function as a turning point toward deviance amplification within-persons across time? And second, do suspensions amplify offending pathways between youth and young adults as they age? It is the purpose of this study to formulate an initial understanding of these two interrelated issues. To do so, we apply a developmentally-informed perspective drawing from the literature on deviance amplification, labeling, and cumulative disadvantage to examine the extent to which suspensions: (a) contribute to within-individual amplifications in offending across time and (b) present a “cumulative effect” on offending between-individuals through deviance amplification.

Outcomes of school discipline within the life-course context

Of all social institutions, schools are one of the most important for adolescents (Meyer, 1977). In addition to being key institutions of social control that influence adolescent development (Payne & Welch, 2016), schools are critical socializing agents that can produce either conforming or deviant behaviors (Gottfredson, 2001; Meighan & Harber, 2007; Payne & Welch, 2016). Prior research has highlighted that fair and consistent school discipline can promote prosocial outcomes for adolescents (Arum & Velez, 2012), including decreased bullying and victimization experiences (Gregory et al., 2010). On the other hand, researchers have tied the use of exclusionary and punitive disciplinary practices to increased grade retention (Losen & Martinez, 2013), decreased educational outcomes (Lyons & Drew, 2006), and increased perceptions of racial/ethnic inequality (Bracy, 2011). Scholars have also tied punitive practices to heightened levels of anxiety and stress (Kupchik, 2010), lower participation in extracurricular activities (Mowen & Manierre, 2017), decreased civic engagement (Kupchik & Catlaw, 2014), and deterioration in the broader school climate (for a review, see Welsh & Harding, 2015). Punitive school discipline has also been highlighted as a key mechanism for the “school-to-prison” pipeline—a process by which youth who are removed from school find themselves in contact with the criminal justice system (Hirschfield, 2008).

Often incorporating notions of schooling and adolescence into its main tenets, the life-course perspective focuses on how events and processes—turning points—contribute to important changes over time. Turning points refer to events that “separate the past from the present” (Sampson & Laub 1993, p. 304) with a key emphasis on long-term outcomes (e.g. Moffitt, 1993). Despite the focus on the distant future, research routinely invokes the concept of a turning point to study important life outcomes over the shorter-term. To provide an example, research has invoked the concept of a turning point to explore how gang membership operates as a turning point in behavioral outcomes among youth (Melde & Esbensen, 2011) and how educational achievement serves as a turning point toward desistance for incarcerated youth within a one-year time period (Blomberg et al., 2012). Scholars have also examined residential change as a turning point toward desistance within a three-year time span (Kirk, 2012). Researchers have also conceptualized employment (Skardhamar & Savolainen, 2014) and programming (Benda et al., 2005) as turning points toward recidivism within five years, and Warr (1998) examined how marriage serves as a turning point away from offending within a three-year time span. In short, turning points can certainly promote long-term change over life histories (e.g. Laub & Sampson, 1993). However, they can also serve as catalysts that spark immediate changes which impact short-term outcomes as well.

In this study, we view school suspensions as important turning points in life which may promote changes in offending as youth move into emerging adulthood. In practical application, however, unraveling the extent to which suspensions are meaningful for offending behavior has proven to be difficult. Part of the reason for this lies in the observation that an important life event—like experiencing formal school sanctions—likely instigates within-person change while simultaneously establishing meaningful between-person differences. That is, suspensions may trigger fluctuations in one’s offending behaviors over time while also highlighting and reifying important differences in offending between individuals. And while researchers have demonstrated that suspensions function as turning points toward increased future contact with the criminal justice system (Mowen & Brent, 2016), research has failed to address the relationship between punishment and offending due—in large part—to the methodological challenges involving causal order (a point which we discuss momentarily). While the current study remedies this methodological limitation, we also move beyond prior research by placing a key focus on how suspensions carry implications for cumulative disadvantage, labeling, and deviance amplification in a longitudinal framework.

Cumulative disadvantage, labeling, and deviance amplification

The notion of cumulative disadvantage proposes that life events can be interrelated to one another (see Sampson & Laub, 1997). As negative events accumulate, their detrimental effects on behavior progressively accumulate and reinforce each other. Formal sanctions like arrest and incarceration have been shown to increasingly “add up” in a manner whereby disadvantage increasingly accumulates and ultimately results in elevated levels of offending (e.g. Liberman et al., 2014). From this perspective, school punishment—like other forms of formal intervention—can culminate in weakened pro-social bonds (Jenkins, 1997), a reduction in employment prospects (Sampson & Laub, 1997), and an increase in contact with criminal/delinquent associates and peers (Bernburg et al., 2006).

Researchers examining cumulative disadvantage often draw from labeling theory, a theory often intertwined with the societal reaction perspective (Sampson & Laub, 1997; see also, Liberman et al., 2014). Labeling theory asserts that official interventions like arrest and suspension can confer stigmatizing labels to a person (see Bernburg & Krohn, 2003). Negative labels attached to primary acts of deviance may manifest in deviant self-conceptions that produce secondary acts of deviance (see Becker, 1963; Lemert, 1951, Tannenbaum, 1938). Becker (1963, p. 31) proposes that being sanctioned and labeled is likely “one of the most crucial steps” in a process leading to a “stable pattern of deviant behavior.” And, as we previously highlighted, the developmental process through which sanctions increase future offending patterns is referred to as deviance amplification.

The concept of deviance amplification highlights the “ironic insight” that—rather than serving as a deterrent—official interventions and authorities may escalate subsequent levels of offending (Marx, 1981, p. 221; Farrington, 1977; Paternoster & Iovanni, 1989; Wilkins, 1965). Research in this area demonstrates that police contact, arrest, judicial interventions, and criminal justice involvement can reinforce, perpetuate, and increase levels of future offending (Bernburg & Krohn, 2003; Bernburg et al., 2006; Liberman et al., 2014; Wiley & Esbensen, 2016; Wiley et al., 2013). Evidence also suggests that official sanctions can indirectly increase offending by ‘knifing off’ conventional lifestyles, attenuating pro-social bonds, establishing connections with delinquent peer groups, provoking deviant self-conceptions and values, and creating feelings of social isolation and withdrawal (Bernburg & Krohn, 2003; Liberman et al., 2014; Link, 1982; Lopes et al., 2012; Paternoster & Iovanni, 1989; Sampson & Laub, 1997; Wiley et al., 2013). Overall, the deviance amplification framework proposes a dynamic process through which sanctions translate into heightened offending patterns over time.

Given the developmental nature of these insights, many propose that the notion of deviance amplification is consistent with a “stepping-stone perspective” and has an “obvious affinity to a life-course, developmental framework” (Sampson & Laub, 1997, p. 9; see also Loeber & Le Blanc, 1990). Recalling the work of Becker (1963) and Lemert (1951), scholars have established that heightened levels of offending early on in life are likely driven by processes of labeling and sanctioning. Therefore, we borrow from the life-course and labeling frameworks to conceptualize school suspension as a type of sanction that contributes to amplified levels of deviance across time. In line with the aforementioned idea that sanctions can produce cumulative effects, we also assert that as individuals experience more suspensions, they will also exhibit higher levels of subsequent offending over time.

While research has yet to fully disentangle this process in the context of suspensions and delinquent behavior, Liberman et al. (2014) studied a related process by examining how arrest contributed to increased offending and arrest via secondary deviance. Using three waves of data encompassing school-aged youth from the Project on Human Development in Chicago Networks, the authors found that youth who were arrested reported significantly greater levels of offending than those not arrested. Perhaps most important to the current study, Liberman et al. (2014, p. 35) suggested that “labeling effects should be strongest for the first labeling event, and each repeated labeling event should have a smaller marginal effect.” In the context of their study, this means that the first arrest should relate to the largest increase in offending, the second arrest to a smaller increase in offending, and the third arrest to an even weaker increase. Despite this, an alternative interpretation drawing from the concept of cumulative disadvantage would suggest that the behavioral effect of labeling could intensify as the events compound upon each other. Thus, instead of the labeling event weakening with each event, cumulative disadvantage would strongly suggest that subsequent labeling events should be seen as having a potentially stronger effect on behavior.

Along with other work that has demonstrated that school suspensions can place individuals at increased odds of contact with the criminal justice system later in life (Mowen & Brent, 2016), the notion of cumulative disadvantage sets a precedent through which to view school discipline as a force that impacts outcomes over time. However, prior work has failed to realize the close linkage between the concept of cumulative disadvantage and offending across time. The failure to fully realize the implications and meaning of cumulative disadvantage on offending in the context of school discipline is, unfortunately, rather common. We believe a key reason for this stems from data and methodological limitations preventing an analysis on the effect of suspension on between-person differences and within-individual changes while simultaneously accounting for levels of offending—a point to which we now turn.

Key limitations to existing research

We believe that three interrelated theoretical and methodological limitations have made it considerably difficult to directly model the effects of school discipline on delinquent conduct. First, there is a “chicken and egg” issue in that individuals who are committing delinquency are also likely to be suspended, resulting in difficulty disentangling causal ordering. That is, individuals who are suspended are also likely to be more deviant than those who are not suspended. As a result, suspension should be related to delinquency (see a similar discussion by Hirschfield, 2018, p. 151–152). For example, while Wolf and Kupchik (2017) demonstrate an association between suspension in wave one (1994–1995) of the National Longitudinal Survey of Adolescent to Adult Health and crime in wave 4 (2007–2008), the control for prior offending behavior also came from wave one, thus raising issues of causal order between suspension and baseline offending. Although we discuss this point in greater detail in our analytic strategy section, the method we use allows us to validly and directly model the effects of prior offending behaviors on future offending to assess the joint impact of delinquency and suspensions on within-person changes in offending across time (Allison, 2015).

Second, and related to the prior point, the true nature of the effect of suspension on within-person changes in offending has been difficult to capture. Prior work typically examines suspension as a predictor in one wave and offending as an outcome in one other wave, thus producing between-person estimates (e.g. Wolf & Kupchik, 2017). And though a useful methodological tool, propensity scores tend only to assess between-person differences in offending (e.g. Liberman et al., 2014; Rosenbaum, 2018). This carries implications for how suspension is seen in a developmental and life-course perspective. By focusing on between-person differences, research may have failed to capture the importance of suspension as a turning point that triggers within-person changes over time. Our work overcomes this issue by examining both between-person differences as well as within-individual changes over time.

Third, prior research has failed to consider the cumulative effect of suspensions on offending. This is a notable limitation given that the research on cumulative disadvantage and deviance amplification consistently and strongly suggests that negative experiences may accumulate, or add up, in their effects on offending (e.g. Wiley & Esbensen, 2016), a point which raises attention to the goals of this study.

Current study

Drawing from the life-course perspective, this study uses data from the National Longitudinal Survey of Youth 1997 and cross-lagged dynamic panel data models to examine the role, cumulative effect, and impact of suspension on offending. We have two specific goals. First, we examine whether school suspensions represent a turning point that contributes to within-individual increases in offending across time. Based on prior research, we expect (Hypothesis 1) that school suspensions will contribute to within-individual increases in offending across time even after controlling for baseline levels of offending. Second, following the life-course notion of “cumulative disadvantage” and prior work documenting the cumulative effects of suspension on criminal justice outcomes (Mowen and Brent, 2016), we expect (Hypothesis 2) that between-individual increases in offending will become greater with each suspension. Specifically, we expect that youth who are suspended in one wave will report greater increases in offending compared to youth who are never suspended. Likewise, after controlling for prior offending, we expect that youth suspended in two waves will report even greater increases in offending, and youth suspended in three waves will report the greatest increase in offending.

Methods

Data

Data for this project come from the National Longitudinal Survey of Youth 1997 (NLSY97; Bureau of Labor Statistics, 2016). A national, household-based survey, the NLSY97 consists of a total of 8,984 youth between the ages of 12 and 18 at wave one. The overall goal and design of the NLSY97 was to “document the transition from school to work and into adulthood” (Bureau of Labor Statistics, 2016, n.p.). Toward this goal, the NLSY97 contains a variety of items capturing delinquency and offending, school experiences, family relationships, peer relationships, employment, education, and sexual activity, among other measures.

The NLSY97 is a longitudinal panel data set and, as such, contains data collected from the same respondents over time. Data collection occurred annually. The first wave of data were collected in 1997 (ages 12–18), wave two in 1998, wave three in 1999, and wave four data were collected in 2000 (ages 16–22). Although the NLSY97 was collected beyond wave four, we choose to focus on the first four waves of data since the vast majority of youth had aged out of school by wave five.

Dependent measure: Offending

The dependent measure in the current study encompasses offending. To capture offending, we draw data from six measures that asked the respondent about offending behaviors. These items asked respondents how many times within the last 12 months (since the last interview) they attacked or assaulted someone, carried a gun, sold illegal substances, destroyed property, stole an item worth less than $50, or stole an item worth more than $50. Respondents reported the total number of times they committed each act, meaning each item captures the total frequency the behavior was committed since the previous interview.

Since each act poses a different level of severity, we apply the severity weights developed by Wolfgang et al. (1985). These weights range from 2.88 (stealing an item worth less than $50) to 6.17 (attacking/assaulting someone). Each individual item was multiplied by the appropriate weight and then summed to create an offending index. The final measure, therefore, captures both the severity and frequency of offending. Due to the positive skew in the data, we use the natural logarithmic transformation in the forthcoming analysis.1 The mean of this logged measure of offending is .649 with an overall standard deviation of 1.467 and a range from 0 to 8.278. As this measure varies across time, the within-individual standard deviation of .964 indicates that individuals report changes in offending across waves. Descriptive statistics of all measures used in the analysis are available in Table 1.

Table 1.

National longitudinal survey of youth 1997 descriptive statistics (n = 6,876).

Variable Mean S.D. Range S.D. within S.D. between
Delinquency measures
 Delinquency index (logged) 0.649 1.467 0–8.278 0.964 1.164
 Lagged delinquency (wave 1) 0.793 1.534 0–8.278 – –
 Lagged delinquency (wave 2)a 0.779 1.613 0–8.278 – –
School suspension measures
 Suspension 0.123 0.328 0, 1 0.233 0.251
 Suspended in no wavesa 0.744 0.436 0, 1 – –
 Suspended in one wavea 0.161 0.232 0, 1 – –
 Suspended in two wavesa 0.074 0.144 0, 1 – –
 Suspended in three wavesa 0.021 0.144 0, 1 – –
School, family, and peer measures
 School drop out 0.044 0.206 0, 1 0.170 0.157
 School bonds 15.287 2.367 5–20 – –
 Gang membership 0.027 0.162 0, 1 0.089 0.222
 Delinquent peers index 8.042 2.124 5–15 – –
 Family bonds 9.558 5.788 0–28 4.469 3.922
 Income 47821 42908.8 0–246474.0 – –
Demographic controls
 Race/Ethnicity
  White 0.525 0.498 0, 1 – –
  Black 0.216 0.425 0, 1 – –
  Hispanic 0.202 0.405 0, 1 – –
  Other 0.057 0.217 0, 1 – –
 Age 14.048 1.427 12–18 – –
 Gender – –
  Female 0.514 0.499 0, 1 – –
 Male 0.486 0.499 0, 1 – –
Open in a new tab

Notes: n = sample size; S.D. = standard deviation.

a

Used in Table 3 Only.

Independent measure: School suspension

The focal independent measure in the current study is school suspension. At each of the four waves of data, respondents were asked if they had been suspended since the prior interview.2 Respondents could answer yes (coded as “1”) or no (coded as “0”). Overall, respondents reported being suspended 12.3 percent of the time during the first four waves of the NLSY97. As time variant measures, a within-individual standard deviation of .233 indicates that students who were suspended once were likely to report being suspended again.

Concurrent with our analytic strategy (outlined in the next sections), we also measure the between-person effect of school suspensions on offending. To create the between-person effects, we include a variable representing whether the respondent reported being suspended in one wave only (16.1 percent of respondents), two waves only (8.4 percent), or three waves only (2.1 percent) within the first three waves of the NLSY97 in contrast to youth who reported never being suspended (74.4 percent; contrast group).3 These categories are mutually exclusive.

Control measures

We control for a variety of factors included in the NLSY97 that have been shown to influence offending. First, we include a binary measure indicating whether the respondent dropped out of school (e.g. Na, 2017). Overall, about 4.4 percent of youth report dropping out of school at some point within the study timeframe. Since youth could report being enrolled in one wave and dropping out at another, this measure is time variant.

We also include a measure capturing the respondent’s bonds to school (see Kirk & Sampson, 2013). To create this measure, we draw data from 5 items collected at wave one that asked the respondents if they felt safe at school, whether teachers were interested in students, if school discipline was fair, if they liked school, and whether students were graded fairly. Respondents were asked to respond along a four-point scale (1 = strongly disagree, 2 = disagree, 3 = agree, 4 = strongly agree). To create a measure capturing student bonds to the school, we performed a principal components analysis (Dunteman, 1989). Results of this analysis demonstrated a unidimensional factor (eigenvalue above 1, factor loadings all exceeding .6, alpha = .730). To create this scale, the items were summed. Student bonds to the school has an overall mean of 15.287 with a standard deviation of 2.367 and ranges from 5 (low bonds) to 20 (high bonds). We note that this measure is time invariant as data on student bonds to school were only collected at wave one.

To account for family bonds (Hirschi, 1969), we use the index of family routines, compiled by NLSY97 researchers (see Appendix 9, p. 43 of the NLSY97 codebook for a full description of this measure).4 This measure, modified from the Family Routines Inventory (Jensen et al. 1983), captures the sense of belonging the respondent has to his/her family. Higher scores indicate stronger family bonds. This measure has a mean of 9.558, has a standard deviation of 5.788, and ranges from 0 (very low bonds to family) to 28 (very high bonds to family). This measure is time variant (within-individual standard deviation = 4.469). In addition to bonds to the family, we also include a variable that captures the respondent’s total family income at wave one. This measure has a mean of 47,821.44 US dollars, a standard deviation of 42,908.80, and ranges from 0 to nearly 250,000 dollars.

As prior literature has demonstrated the important role peers play in offending (Haynie, 2002), we include two measures of peer relationships within the analysis. The first measure asked respondents if they were members of a gang. Overall, 2.7 percent of youth reported being in a gang, though this measure changes across time (within-individual standard deviation = .092). In addition, youth were asked how many of their peers: get drunk more than once a month, smoke, use illegal drugs, belong to a gang, and cut class at school.5 Respondents could respond almost none (coded 1), about half (coded 2), and almost all (coded 3). To create a measure of delinquent peers, each item was summed to create a scale so that higher values indicate a greater proportion of delinquent peers. This measure is time invariant as the questions capturing peer delinquency were collected at wave one only. The scale has an overall mean of 8.042, a standard deviation of 2.124, and ranges from 5 (almost no delinquent peers) to 15 (almost all delinquent peers). The reliability of this measure is .801 (Cronbach, 1951).

To account for the influence of race/ethnicity, we include measures indicating that the respondent was White (52.5 percent of the sample), Black (21.6 percent), Hispanic (20.2 percent) or Other race/ethnicity (5.7 percent). Gender is also included as a binary measure with females (51.4 percent of the sample) compared to males (48.6 percent of the sample). To account for age, we include age measured at wave 1. The mean age at wave one is 14.047 years (standard deviation = 1.427; the range is 12–18 years).

Missing data

As with all large-scale panel datasets, there are missing data present within the NLSY97. In our models, we rely on a sample size of 6,876, or about 77 percent of the original sample due to pairwise deletion (missing data on some measures across each wave). To examine the influence of missing data on our analysis, we performed a sensitivity analysis by comparing patterns of responses of missing and non-missing data through a series of t-tests (Brame & Paternoster, 2003). The results of the t-tests yielded non-significant results, suggesting that no covariate used in our analysis is significantly predictive of attrition. We also used maximum-likelihood estimation (Allison, 2012) to impute missing data in our models; results were substantively similar to the models we present in the next pages. In short, we note that missing data is present in our analysis but does not appear to be significantly impacting our results.

Analytic strategy

To overcome the three limitations of prior research which we previously discussed, we employ a cross-lagged dynamic fixed-effects panel. As Allison (2015, n.p.) highlights, “one of the best predictors of what happens at time t is what happened at time t-1.” However, it is not possible to include a lagged measure (t-1) of the dependent variable (t) in a traditional fixed- or mixed-effects model due to correlated error terms and a lack of independence (Allison, 2015). A cross-lagged dynamic panel model overcomes this limitation by using a series of chained equations. This characteristic of the modeling strategy is particularly important for the current effort as we would expect that delinquent youth could potentially be suspended because they committed crime. The use of the cross-lagged fixed-effects dynamic panel model accounts for this by directly modeling delinquency at t-1 (referred to as a “baseline” measure; see Boman & Mowen, 2018) on each progressive time point t while simultaneously meeting the assumption of independence. Put differently, this modeling approach directly models progressive pathways from wave 1 offending onto waves 2, 3, and 4 offending; wave 2 offending onto waves 3 and 4; and wave 3 offending onto wave 4 offending. Prior behavior is, therefore, modeled as a predictor of future behavior, thus creating a baseline measure of offending.

Another advantage of the cross-lagged dynamic panel model is the ability to model both between-individual differences as well as within-individual changes. While mixed-effects models can model both between-individual differences as well as within-individual changes, they are often susceptible to concerns over severe endogeneity (Rabe-Hesketh & Skrondal, 2012). A fixed-effects model, while not as prone to endogeneity, can traditionally only model within-individual changes. While fixed-effects models are powerful, they fail to directly model between-person differences that researchers know matter (such as gender or race). The cross-lagged dynamic panel model can estimate time invariant between-person covariates (i.e. race) while simultaneously modeling time variant within-person measures (i.e. suspension). As a result, we can produce models that (1) estimate the time variant within-person effect of suspension on offending, (2) model the time invariant between-person effects of suspension on offending, and (3) directly control for lagged, baseline levels of offending.

To address our first research question, we begin by examining the effect of suspension on within-individual changes in offending across time while accounting for baseline levels of offending. Then, to address our second research question, we examine whether suspensions create a cumulative effect on offending while again accounting for baseline levels of offending.

Results

The results of the analysis examining the effect of time variant school suspensions on self-reported delinquency are shown in Table 2. The model fit indices, shown at the bottom of the table, indicate acceptable fit to the data. The root mean square error of approximation (RMSEA) of .02 is below the accepted threshold of .06, and the comparative fit index (CFI) of .958 falls above the cutoff of .95 (Acock, 2013). Although the chi-square statistic is significant, it is not a preferable means of evaluating model fit in cross-lagged dynamic panel data models since it is particularly influenced by the size of the sample (Williams et al. 2016). Thus, based on the preferred criteria of the RMSEA and the CFI, we conclude that the model fits the data well.

Table 2.

Cross-lagged dynamic panel model examining offending across time with time variant suspension (n = 6,876).

Variable Coefficient S.E.
School suspension measure
 Suspension 0.377 0.060***
Delinquency measure
 Lagged Offending (Wave 1) 0.142 0.014***
School, family, and peer Measures
 School drop out 0.111 0.099
 School bonds −0.039 0.006***
 Gang membership 1.740 0.329***
 Delinquent peers index 0.058 0.007***
 Family bonds −0.191 0.071***
 Household income 0.006 0.009
Demographic controls
 Race/Ethnicity
  Black −0.227 0.031***
  Hispanic −0.114 0.032***
  Other −0.124 0.122
 Age −0.077 0.011***
 Gender
  Female −0.402 0.003***
χ2 (Model vs. Saturated) 107.925***
Root mean square error of approximation (RMSEA) 0.020
Comparative fit index 0.958
Open in a new tab
*

p ≤ .05

**

p ≤ .01

***

p ≤ .001.

Notes: n = sample size; S.E. = standard error.

Turning to the results of the independent variables, school suspensions are related to a .377 unit increase in the logged delinquency index. As a within-individual effect only (a fixed effect), youth who experience a suspension report a significant within-person increase in offending across time. Not surprisingly, and highlighting the utility of this method, the lagged measure of delinquency—delinquency at wave one—is significantly associated with delinquency across time. As such, offending at wave one is tied to changes in offending across time. Results of the control measures reveal that family bonds relate to lower offending across time and individuals who are members of a gang report significant increases in offending. Several time invariant measures are also significantly related to offending. Specifically, individuals with a greater proportion of delinquent peers report significantly higher levels of offending than those with fewer delinquent peers. Black and Hispanic youth, relative to White youth, report lower levels of offending. Finally, results showed that higher levels of bonds to the school, being female, and being older are all associated with lower levels of offending.

Table 3 reports results relevant to the possibility that there may be a cumulative effect of suspension on offending. The fit indices in Table 3 demonstrate a strong fit to the data based on the RMSEA (.015) and CFI (.994). Results of the covariates demonstrate that individuals suspended in one wave report significantly higher levels of offending than those never suspended. Similarly, those suspended in two waves report significantly higher levels of offending than those never suspended, and respondents suspended in three waves report significantly higher levels of offending than those never suspended. In addition, the coefficients increase with each increase in waves suspended (.273 in wave one; .464 in two waves, .923 in all three waves). These results suggest there is a cumulative effect on offending. Importantly, this finding remains even after validly being able to control for prior delinquency. Results of the control measures are similar, though not the same, as the prior model. Specifically, family bonds are no longer significantly related to offending after accounting for lagged offending and the cumulative effect of suspension on offending. We posit some explanations for this difference in the following section.

Table 3.

Cross-lagged dynamic panel model examining change in offending between waves three and four with time invariant suspension (n = 6,876).

Variable Coefficient S.E.
School suspension measures
 Suspended in one wave 0.273 0.055***
 Suspended in two waves 0.464 0.084***
 Suspended in three waves 0.923 0.141***
Delinquency measure
 Lagged offending (wave 1) 0.156 0.034***
 Lagged offending (wave 2) 0.044 0.021*
School, family, and peer measures
 School drop out 0.189 0.156
 School bonds −0.200 0.007***
 Gang membership 1.858 0.472***
 Delinquent peers index 0.024 0.009***
 Family bonds −0.207 0.110
 Household income 0.017 0.011
Demographic controls
 Race/Ethnicity
  Black −0.227 0.041***
  Hispanic −0.089 0.038***
  Other −0.133 0.138
 Age −0.065 0.056***
 Gender
  Female −0.315 0.036***
χ2 (Model vs. Saturated) 30.500**
Root mean square error of approximation (RMSEA) 0.015
Comparative fit index 0.994
Open in a new tab
*

p ≤ .05

**

p ≤ .01

***

p ≤ .001.

Notes: n = sample size; S.E. = standard error.

To illustrate these findings, Figure 1 presents a line graph of the findings reported in Table 3. Although respondents who were never suspended report decreases in offending during this time period, youth who were suspended report significant increases in offending during this same time period even after prior offending is directly modeled. As a result, suspension presents a cumulative effect on offending for youth as they move through adolescence.

Figure 1.

Figure 1.

Open in a new tab

Predicted effects of suspension on offending, holding prior offending and covariates constant (n = 6,876).

Discussion and conclusion

In light of the increases in punitive discipline and suspensions in school hallways across the United States (Hirschfield, 2008; Kupchik, 2016), the goal of this study was to examine the extent to which school suspensions function as a turning point in offending pathways. The use of a series of cross-lagged dynamic panel models allowed us to examine the effect of suspension on within-person changes in offending as well as the cumulative effect of suspensions on offending between-individuals while simultaneously accounting for prior levels of offending. Our first hypothesis premised that school suspensions would contribute to within-individual increases in offending across time. The support for this hypothesis echoes broader research demonstrating that punishment can increase offending (e.g. Mowen et al., 2017; Wiley & Esbensen, 2016). More specifically, this suggests that school discipline can serve as a negative and harmful turning point in adolescence that increases offending within-individuals over time.

Our second hypothesis premised that youth suspended in a greater number of waves would report significantly higher levels of offending than youth suspended in fewer or no waves. This presumption is supported as accumulated suspensions over time were related to increases in offending, even after controlling for baseline offending. Since suspensions do appear to accumulate and compound in their effect on offending, this study’s findings support the notion that suspensions may contribute to “cumulative disadvantage.” This result suggests that suspensions exacerbate and reify between-individual differences in offending, meaning this study contributes to existing literature that has demonstrated a cumulative effect of suspension on arrest by demonstrating a similar dynamic for the cumulative effect on offending (e.g. Wiley & Esbensen, 2016).

Considering these findings together, school suspensions appear to increase offending within-individuals across time and present a cumulative effect on offending between-individuals. To unpack these findings, we turn back to prior literature. First, while prior literature has demonstrated that formal sanctions like arrest can function as a negative turning point that increases offending (Wiley & Esbensen, 2016), findings from this study suggest that school suspensions also function to increase offending. In addition to creating important between-person differences (Rosenbaum, 2018), suspensions contribute to within-person increases in offending. Placed within the framework of labeling theory, results support the notion that suspension, as a form of sanction, may contribute to secondary deviance via deviance amplification even after accounting for prior levels of offending. This strongly suggests that school suspensions contribute to secondary acts of offending beyond baseline levels of offending.

Findings also carry implications for research examining the effect of labeling on offending. Although Liberman et al. (2014) find that additional arrests contributed to small marginal increases in offending, our findings demonstrate that suspensions at a certain threshold contribute to significantly stronger increases in offending. Results suggest that the effect of a label may increase significantly with additional labeling events. Thus, we suggest that the effects of stigmatizing labels may result in dramatic increases in offending as opposed to “smaller marginal” increases (Liberman et al., 2014, p. 35). This represents a fundamentally different understanding of the effects of labeling on secondary deviance. As a result, future research should consider how other forms of punishment (e.g. arrest) may present a compounding and cumulative labeling effect on offending across time.

Placed within the life-course framework, our findings suggest that school suspensions serve to amplify offending pathways both within-individuals across time as well as between-individuals. While these findings strongly demonstrate that school suspensions are an important life event—a turning point—in the short term for youth, they do not speak to longer-term outcomes. From a life-course perspective, it is entirely possible that the effect of school suspensions may fundamentally alter trajectories (e.g. Moffitt, 1993) of offending across time. While it is outside the scope of the current study to examine how school discipline may fundamentally alter longer-term outcomes, findings from this study highlight the need for scientists to consider the possibility that school discipline and security may influence offending trajectories into, and through, adulthood.

Relatedly, findings from this study also raise two important considerations regarding race/ethnicity, punishment, and offending. First, our analysis demonstrated that Black and Hispanic youth reported lower levels of offending than White youth. This finding is contrary to existing studies that have found that White youth report lower levels of offending than non-White youth (see Piquero, 2015 for an overview). Although it is outside of the scope of this research to examine this pattern of findings in greater detail, a number of studies using the NLSY97 have documented higher levels of offending among Whites relative to non-Whites (e.g. Kakade et al. 2012; Mitchell & Caudy, 2017; Mowen & Schroeder, 2018). Future research should comprehensively explore factors that may explain differences in offending behaviors across racial/ethnic groups in the NLSY97. Second, and perhaps more centrally concerned with the main findings of this study, prior work has demonstrated strong racial and ethnic inequalities in the use of suspensions in schools across the United States (Skiba et al., 2011). Specifically, studies have shown that Black and Hispanic youth are far more likely to receive suspensions than their White counterparts (e.g. Shollenberger, 2015; Skiba et al., 2011). Because suspensions are not experienced equally, findings from this study suggest that the effects of punitive school discipline may exacerbate differences in offending across racial/ethnic groups over time. Future research should examine the extent to which school punishment may produce disparate levels of offending across key sociodemographic indicators.

In addition to the theoretical implications, this study also carries important policy implications. American schools have increasingly relied on exclusionary sanctions, zero-tolerance policies, and criminal justice appendages (Hirschfield, 2008; Musu-Gillette et al., 2017) to maintain control and safety. These “criminalized” strategies have become a natural part of school environments (Hirschfield and Celinska, 2011) without much evidence as to their effectiveness (Musu-Gillette et al., 2017). Findings from this study add to the growing body of literature outlining some of the negative—and seemingly unintended—consequences of the use of punitive discipline. Findings underscore conclusions from other studies that call on school officials to examine more “responsive responses” (Kupchik et al., 2015) that address the underlying causes of offending instead of relying solely on punishment.

While results from this study demonstrate that school suspensions can exacerbate offending, it is inappropriate to suggest that school officials should completely abandon their use of suspensions. In fact, prior research demonstrates that fair and equitable punishment can contribute to positive outcomes among youth (Arum & Velez, 2012). Despite this, about 3.5 million students experience a suspension each year (Losen, 2015). To place this in perspective, more students are suspended each year in schools in the United States than there are seniors enrolled in high school (Shollenberger, 2015), meaning that suspensions are a normal experience for many youth in the United States. Moreover, when highly publicized events like shootings occur on school grounds, the response from policy makers, school officials, and the general public alike tends to center around increasing discipline and security (for an overview, see Kupchik et al., 2015). Yet, findings from this study demonstrate that intensification of disciplinary strategies may counterproductively increase offending behaviors. The use of zero tolerance strategies and punitive discipline may place a significant number of youth at risk of experiencing increases in offending unless there is an accompanying strategy to address the underlying factors that contribute to misbehavior (Skiba and Noam, 2002).

Although findings from this study demonstrate that school suspensions amplify offending, we do not believe that most school officials use suspensions with the belief that students will experience negative transitions. As Skiba and Sprague (2008, p. 41) note, “most administrators turn to school exclusion as a disciplinary tool because they need to do something and don’t know what else to do.” In this observation, Skiba and Sprague point to a growing awareness of the need to use alternatives to suspension like Positive Behavior-Intervention Support (PBIS) programs. PBIS programs have received considerable attention and have been shown to reduce school suspensions as well as offending (for a comprehensive study, see Bradshaw et al., 2010, see also Howell, 2013). Unfortunately, we are unable to examine the effect of PBIS on offending in the current study due to limitations in these data. However, we note the importance of prior work calling for alternatives to the use of suspensions to improve student outcomes (Bradshaw et al., 2010; Kupchik, 2016).

Interestingly, one’s level of bonds to his/her family no longer significantly related to offending after accounting for the cumulative effect of suspension. Although this finding was unexpected, it may suggest that while family bonds are generally a protective factor for offending (results in Table 2), they no longer significantly protect against offending once the between-person, cumulative effect of suspension is captured (results in Table 3). Thus, it seems possible that the between-person, cumulative effect of suspension on offending is so robust that the effect of some protective factors (such as family bonds) may be proven spurious. Relatedly, future research should examine the extent to which school discipline may interact with, or impede, family processes and relationships (see, e.g. Dunning-Lozano, 2018).

Despite the contributions of this study, it is not without important limitations. First, as a household-based survey, we are unable to examine specific characteristics of the school. Although prior work examining the link between adolescent experiences and offending later in life finds that individual-level characteristics are more “consistent and robust predictors of future negative outcomes” (Wolf & Kupchik, 2017, p. 424) than school characteristics, this remains an important limitation. In reality, it is possible that a variety of school characteristics like the presence of school resource officers (Theriot, 2009) may affect student behaviors.

Additionally, prior work has also established the importance of peer offending on adolescent delinquency. Although we control for the effects of peer delinquency at wave one, the NLSY97 did not include questions about peer delinquency in subsequent waves. It is likely that changes in peer delinquency relate to significant changes in respondent behaviors (Haynie, 2002). This measure is also perceptual, lending to questions about its validity (see Boman et al. 2012). Future work should examine the effect of suspension on offending across time while also accounting for the influence of peer behavior using different measurement strategies. Additionally, this study relies on the use of self-report data which may be limited by respondent’s knowledge, ability to recall past events, need to provide socially acceptable responses, and any other unacknowledged biases.

Another notable limitation concerns the changing landscape of school discipline. This issue presents at least two specific limitations. First, disciplinary strategies have expanded to include more punitive responses associated with zero-tolerance policies, use of police officers, and “criminalized” conduct manuals (see Hirschfield, 2008; Kupchik, 2010). However, reflecting a limitation within the larger literature, we are unable to examine the link between delinquency and other forms of discipline. Future research should examine how other forms of punishment such as resource officer intervention, expulsion, and placement in alternative schools influence offending pathways. Second, school discipline has changed a great deal since the time these data were collected. In fact, evidence suggests that school security and discipline has intensified since the first four waves of the NLSY97 were collected (see, generally, Kupchik, 2010). Since the use of suspension and security has increased in the last few decades, findings from this study may underestimate the effect of school discipline on offending behaviors. Future research should examine the influence of suspension and other security interventions on student offending behaviors using more contemporary data that reflect these changes.

Despite some limitations, findings from this study demonstrate that school discipline can serve as an important turning point during adolescence. Given the relatively recent attention to many of the deleterious outcomes associated with “criminalized” school discipline (Hirschfield, 2008), this finding is particularly important as it provides systematic evidence that exclusionary school discipline increases offending. School suspensions amplify levels of offending across time, establish salient and meaningful between-person differences in offending, and can present a cumulative effect that substantially amplifies deviance as youth move through adolescence and into emerging adulthood.

Funding

This research was supported in part by the Center for Family and Demographic Research, Bowling Green State University, which has core funding from the Eunice Kennedy Shriver National Institute of Child Health and Human Development (P2CHD050959).

Footnotes

1

We used both the transformed and untransformed measure in the analysis. The substantive results were identical, but the logged measure produced models that fit the data more closely. As a result, we report results using the logged measure.

2

At wave 1, respondents were asked if they had been suspended within the last year.

3

We do not include wave 4 suspensions in this category as our outcome variable in these models is change in offending between wave 3 and wave 4. Thus, including wave 4 suspensions would result in a violation of temporal ordering

4

NLSY97 researchers discuss the psychometric properties of this measure in the NLSY97 codebook and note that this measure has strong predictive validity.

5

Being in a gang and having friends in a gang could present issues of multicollinearity. We examined the relationship between both measures and the correlation was very low (r = .092) suggesting having friends in a gang, and being in a gang, are not highly correlated in these data.

Disclosure statement

No potential conflict of interest was reported by the author(s).

References

  1. Acock AC (2013). Discovering structural equation modeling using Stata. College Station, TX: Stata Press. [Google Scholar]
  2. Allison PD (2012). Why Maximum Likelihood Is Better Than Multiple Imputation. Statistical Horizons Retrieved from http://statisticalhorizons.com/ml-better-than-mi [Google Scholar]
  3. Allison PD (2015). Don’t Put Lagged Dependent Variables in Mixed Models. Statistical Horizons Retrieved from http://statisticalhorizons.com/lagged-dependent-variables [Google Scholar]
  4. Arum R, & Velez M. (Eds.). (2012). Improving learning environments: School discipline and student achievement in comparative perspective. Stanford, CA: Stanford University Press. [Google Scholar]
  5. Becker H. (1963). Outsiders: Studies in the sociology of deviance. New York, NY: Free Press. [Google Scholar]
  6. Benda BB, Harm NJ, & Toombs NJ (2005). Survival analysis of recidivism of male and female boot camp graduates using life-course theory. Journal of Offender Rehabilitation, 40(3–4), 87–113. doi: 10.1300/J076v40n03_05 [DOI] [Google Scholar]
  7. Bernburg JH, & Krohn MD (2003). Labeling, life changes, and adult crime: The direct and indirect effects of official intervention in adolescence on crime in early adulthood. Criminology, 41(4), 1287–1318. doi: 10.1111/j.1745-9125.2003.tb01020.x [DOI] [Google Scholar]
  8. Bernburg JG, Krohn MD, & Rivera CJ (2006). Official labeling, criminal embeddedness, and subsequent delinquency: A longitudinal test of labeling theory. Journal of Research in Crime and Delinquency, 43(1), 67–88. doi: 10.1177/0022427805280068 [DOI] [Google Scholar]
  9. Blomberg TG, Bales WD, & Piquero AR (2012). Is educational achievement a turning point for incarcerated delinquency across race and sex? Journal of Youth and Adolescence, 41(2), 202–216. doi: 10.1007/s10964-011-9680-4 [DOI] [PubMed] [Google Scholar]
  10. Boman JH, & Mowen TJ (2018). The role of turning points in establishing baseline differences between people in developmental and life-course criminology. Criminology, 56(1), 191–224. doi: 10.1111/1745-9125.12167 [DOI] [PMC free article] [PubMed] [Google Scholar]
  11. Boman JH, Stogner JM, Miller BL, Griffin OH III., & Krohn MD (2012). On the operational validity of peer delinquency: Exploring projection and elements contained in perceptions. Journal of Research in Crime and Delinquency, 49(4), 601–621. doi: 10.1177/0022427811419367 [DOI] [Google Scholar]
  12. Bracy N. (2011). Student perceptions of high-security school environments. Youth & Society, 43, 365–395. doi: 10.1177/0044118X10365082 [DOI] [Google Scholar]
  13. Bradshaw CP, Mitchell MM, & Leaf PJ (2010). Examining the effects of schoolwide positive behavioral interventions and supports on student outcomes: Results form a randomized controlled effectiveness trial in elementary schools. Journal of Positive Behavior Interventions, 12(3), 133–148. doi: 10.1177/1098300709334798 [DOI] [Google Scholar]
  14. Brame R, & Paternoster R. (2003). Missing data problems in criminological research: Two case studies. Journal of Quantitative Criminology, 19(1), 55–78. doi: 10.1023/A:1022518712229 [DOI] [Google Scholar]
  15. Bureau of Labor Statistics (2016). National Longitudinal Surveys: The NLSY97. Retrieved from https://www.bls.gov/nls/nlsy97.htm
  16. Cronbach LJ (1951). Coefficient alpha and the internal structure of tests. Psychometrika, 16(3), 297–334. doi: 10.1007/BF02310555 [DOI] [Google Scholar]
  17. Dunning-Lozano JL (2018). Secondary discipline: The unintended consequences of zero tolerance school discipline for low-income Black and Latina mothers. Urban Education (Online First) 1–28. doi: 10.1177/0042085918817343 [DOI] [Google Scholar]
  18. Dunteman GH (1989). Principal components analysis. Newbury Park, CA: Sage University Paper Series. [Google Scholar]
  19. Fabelo T, Thompson M, Plotkin M, Carmichael D, Marchbanks M III., & Booth EA (2011). Breaking schools’ rules: A statewide study of how school discipline relates to students’ success and juvenile justice involvement. New York, NY: Council of State Governments Justice Center and Public Policy Research Institute. [Google Scholar]
  20. Farrington DP (1977). The effects of public labelling. The British Journal of Criminology, 17(2), 112. doi: 10.1093/oxfordjournals.bjc.a046802 [DOI] [Google Scholar]
  21. Gottfredson D. (2001). Schools and delinquency. New York, NY: Cambridge University Press. [Google Scholar]
  22. Gregory A, Cornell D, Fan X, & Sheras P. (2010). Authoritative school discipline: High school practices associated with lower bullying and victimization. Journal of Educational Psychology, 102, 482–496. [Google Scholar]
  23. Haynie DL (2002). Friendship networks and delinquency: the relative nature of peer delinquency. Journal of Quantitative Criminology, 18(2), 99–134. doi: 10.1023/A:1015227414929 [DOI] [Google Scholar]
  24. Hemphill SA, Toumbourou JW, Herrenkohl TI, Mcmorris BJ, & Catalano RF (2006). The effect of school suspensions and arrests on subsequent adolescent antisocial behavior in Australia and the United States. Journal of Adolescent Health, 39, 736–744. [DOI] [PubMed] [Google Scholar]
  25. Hirschfield PJ (2008). Preparing for prison? The criminalization of school discipline in the USA. Theoretical Criminology, 12(1), 79–101. doi: 10.1177/1362480607085795 [DOI] [Google Scholar]
  26. Hirschfield PJ (2018). Schools and Crime. Annual Review of Criminology, 1(1), 149–169. doi: 10.1146/annurev-criminol-032317-092358 [DOI] [Google Scholar]
  27. Hirschfield P, & Celinska K. (2011). Beyond fear: Sociological perspectives on the criminalization of school discipline. Sociology Compass, 5(1), 1–12. doi: 10.1111/j.1751-9020.2010.00342.x [DOI] [Google Scholar]
  28. Hirschi T. (1969). Causes of delinquency. Berkeley, CA: University of California. [Google Scholar]
  29. Howell JC (2013). GREAT results: Implications for PBIS in Schools. Criminology & Public Policy, 12, 413–420. doi: 10.1111/1745-9133.12051 [DOI] [Google Scholar]
  30. Jaggers JW, Robison SB, Rhodes JLF, Guan X, & Church WT (2016). Predicting adult criminality among Louisiana’s urban youth: Poverty, academic risk, and delinquency. Journal of the Society for Social Work and Research, 7(1), 89–116. doi: 10.1086/685089 [DOI] [Google Scholar]
  31. Jenkins PH (1997). School delinquency and the school social bond. Journal of Research in Crime and Delinquency, 34(3), 337–367. doi: 10.1177/0022427897034003003 [DOI] [Google Scholar]
  32. Jensen EW, James SA, Boyce WT, & Hartnett SA (1983). The family routines inventory: Development and validation. Social Sciences Medicine, 17(4), 201–211. doi: 10.1016/0277-9536(83)90117-X [DOI] [PubMed] [Google Scholar]
  33. Kakade M, Duarte CS, Liu X, Fuller CJ, Drucker E, Hoven CW, … Wu P. (2012). Adolescent substance use and other illegal behaviors and racial disparities in criminal justice system involvement: Findings from a US National Survey. American Journal of Public Health, 102(7), 1307–1310. doi: 10.2105/AJPH.2012.300699 [DOI] [PMC free article] [PubMed] [Google Scholar]
  34. Kirk DS (2012). Residential change as a turning point in the life course of crime: Desistance or temporary cessation? Criminology, 50(2), 329–358. doi: 10.1111/j.1745-9125.2011.00262.x [DOI] [Google Scholar]
  35. Kirk DS, & Sampson RJ (2013). Juvenile arrest and collateral educational damage in the transition to adulthood. Sociology of Education, 88, 36–62. doi: 10.1177/0038040712448862 [DOI] [PMC free article] [PubMed] [Google Scholar]
  36. Kupchik A. (2010). Homeroom security: School discipline in the age of fear. New York, NY: NYU Press. [Google Scholar]
  37. Kupchik A, Brent JJ, & Mowen TJ (2015). The aftermath of Newtown: More of the same. British Journal of Criminology, 55(6), 1115–1130. doi: 10.1093/bjc/azv049 [DOI] [Google Scholar]
  38. Kupchik A. (2016). The real school safety problem: The long-term consequences of harsh school punishment. Oakland, CA: University of California Press. [Google Scholar]
  39. Kupchik A, & Catlaw TJ (2014). The dynamics of school discipline in a neoliberal era. In Muschert GW, Henry S, Bracy N, & Peguero A. (Eds.), Responding to school violence: Confronting the Columbine effect (pp. 53–68). Boulder, CO: Lynne Rienner Publishers. [Google Scholar]
  40. Laub JH, & Sampson RJ (1993). Turning points in the life course: Why change matters to the study of crime. Criminology, 31(3), 301–325. doi: 10.1111/j.1745-9125.1993.tb01132.x [DOI] [Google Scholar]
  41. Lemert EM (1951). Social pathology: A systematic approach to the theory of sociopathic behavior. New York, NY: McGraw-Hill. [Google Scholar]
  42. Liberman AM, Kirk DS, & Kim K. (2014). Labeling effects of first juvenile arrests: Secondary deviance and secondary sanctioning. Criminology, 52(3), 345–370. doi: 10.1111/1745-9125.12039 [DOI] [Google Scholar]
  43. Link B. (1982). Mental patient status, work, and income: An examination of the effects of a psychiatric label. American Sociological Review, 47(2), 202–215. [PubMed] [Google Scholar]
  44. Loeber R, & Le Blanc M. (1990). Toward a developmental criminology. Crime and Justice, 12, 375–473. doi: 10.1086/449169 [DOI] [Google Scholar]
  45. Lopes G, Krohn MD, Lizotte AJ, Schmidt NM, Vasquez BE, & Bernburg JG (2012). Labeling and cumulative disadvantage: The impact of formal police intervention on life chances and crime during emerging adulthood. Crime & Delinquency, 58, 456–488. [Google Scholar]
  46. Losen D. (2015). Closing the school discipline gap: Equitable remedies for excessive exclusion. New York, NY: Teachers College Press. [Google Scholar]
  47. Losen D, & Martinez T. (2013). Out of school and off track: The overuse of suspensions in American middle and high schools. Los Angeles, CA: The UCLA Center for Civil Rights Remedies at the Civil Rights Project. [Google Scholar]
  48. Lyons W, & Drew J. (2006). Punishing schools: Fear and citizenship in American public education. Ann Arbor, MI: The University of Michigan Press. [Google Scholar]
  49. Marx GT (1981). Ironies of social control: Authorities as contributors to deviance through escalation, nonenforcement and covert facilitation. Social Problems, 28(3), 221–246. [Google Scholar]
  50. Meighan R, & Harber C. (2007). A sociology of educating. London: Continuum International Publishing Group. [Google Scholar]
  51. Melde C, & Esbensen F. (2011). Gang membership as a turning point in the life course. Criminology, 49(2), 513–552. [Google Scholar]
  52. Meyer JW (1977). The effects of education as an institution. American Journal of Sociology, 83(1), 55–77. doi: 10.1086/226506 [DOI] [Google Scholar]
  53. Mitchell O, & Caudy MS (2017). Race differences in drug offending and drug distribution arrests. Crime & Delinquency, 63, 91–112. doi: 10.1177/0011128714568427 [DOI] [Google Scholar]
  54. Moffitt T. (1993). Adolescence-limited and life-course-persistent antisocial behavior: A developmental taxonomy. Psychological Review, 100(4), 674–701. doi: 10.1037/0033-295X.100.4.674 [DOI] [PubMed] [Google Scholar]
  55. Mowen TJ, & Brent JJ (2016). School discipline as a turning point: The cumulative effect of suspension on arrest. Journal of Research in Crime and Delinquency, 53(5), 628–653. doi: 10.1177/0022427816643135 [DOI] [Google Scholar]
  56. Mowen TJ, Brent JJ, & Bares KJ (2018). How arrest impacts delinquency over time between and within individuals. Youth Violence and Juvenile Justice, 16(4), 358–377. doi: 10.1177/1541204017712560 [DOI] [Google Scholar]
  57. Mowen TJ, & Manierre MJ (2017). School security measures and extracurricular participation: an exploratory multi-level analysis. British Journal of Sociology of Education, 38(3), 344–363. doi: 10.1080/01425692.2015.1081091 [DOI] [Google Scholar]
  58. Mowen TJ, & Schroeder RD (2018). Maternal parenting style by race and the moderating effect of structural disadvantage. Youth & Society, 50, 139–159. doi: 10.1177/0044118X15598028 [DOI] [Google Scholar]
  59. Musu-Gillette L, Zhang A, Wang K, Zhang J, & Oudekerk BA (2017). Indicators of school crime and safety: 2016. Washington, DC: National Center for Education Statistics, U.S. Department of Education, and Bureau of Justice Statistics, Office of Justice Programs, U.S. Department of Justice. [Google Scholar]
  60. Na C. (2017). The consequences of school dropout among serious adolescent offenders: More offending? More arrest? Both? Journal of Research in Crime and Delinquency, 54(1), 78–110. doi: 10.1177/00224278166641187 [DOI] [Google Scholar]
  61. Nolan K. (2011). Police in the hallways: Discipline in an urban high school. Minneapolis, MN: University of Minnesota Press. [Google Scholar]
  62. Piquero A. (2015). Understanding race/ethnicity differences in offending across the life course: gaps and opportunities. Journal of Developmental and Life-Course Criminology, 1(1), 21–32. doi: 10.1007/s40865-015-0004-3 [DOI] [Google Scholar]
  63. Rabe-Hesketh S, & Skrondal A. (2012). Multilevel and longitudinal modeling using stata (3rd ed.) Stata Press, TX: College Station. [Google Scholar]
  64. Rausch MK, & Skiba RJ (2005). The academic cost of discipline: The relationship between suspension/expulsion and school achievement. Retrieved from content/uploads/2014/03/Academic-Cost-of-School-Discipline.pdf [Google Scholar]
  65. Rosenbaum J. (2018). Educational and criminal justice outcomes 12 years after school suspension. Youth & Society, (Online First) 1–23. doi: 10.1177/0044118X17752208 [DOI] [PMC free article] [PubMed] [Google Scholar]
  66. Paternoster R, & Iovanni L. (1989). The labeling perspective and delinquency: An elaboration of the theory and an assessment of the evidence. Justice Quarterly, 6(3), 359–394. [Google Scholar]
  67. Payne AA, & Welch K. (2016). The centrality of schools in the life course: The case for focusing on school-related influences in developmental theory and research. Deviant Behavior, 37(7), 748–760. doi: 10.1080/01639625.2015.1071134 [DOI] [Google Scholar]
  68. Sampson RJ, & Laub JH (1993). Crime in the making: Pathways and turning points throughout life. Cambridge, MA: Harvard University Press. [Google Scholar]
  69. Sampson RJ, & Laub JH (1997). A life-course theory of cumulative disadvantage and the stability of delinquency. In Thornberry T. (Ed.), Developmental theories of crime and delinquency (pp. 133–161). New Brunswick, NJ: Transaction Publishers. [Google Scholar]
  70. Shollenberger TL (2015). Racial disparities in school suspension and subsequent outcomes. In Losen D. (Ed.), Closing the school discipline gap: Equitable remedies for excessive exclusion (pp. 31–44). New York, NY: Teachers College Press. [Google Scholar]
  71. Skardhamar T, & Savolainen J. (2014). Changes in criminal offending around the time of job entry: A study of employment and desistance. Criminology, 52(2), 263–291. [Google Scholar]
  72. Skiba R, & Sprague J. (2008). Safety without suspensions. Educational Leadership. Retrieved from http://www.pbis.org/common/cms/files/Coach_Trainer/Articles/Safety%20Without%20Suspensions.pdf [Google Scholar]
  73. Skiba RJ, & Noam GG (2002). Zero tolerance: Can suspension and expulsion keep schools safe? new directions for youth development. San Francisco, CA: Jossey-Bass Publishers. [Google Scholar]
  74. Skiba RJ, Horner RH, Chung C, Rausch MK, May SL, & Tobin T. (2011). Race is not neutral: A national investigation of African American and Latino disproportionately in school discipline. School Psychology Review, 40, 85–107. [Google Scholar]
  75. Simon J. (2007). Governing through crime: How the war on crime transformed American democracy and created a culture of fear. New York, NY: Oxford University Press. [Google Scholar]
  76. Tannenbaum F. (1938). Crime and the Community. New York, NY: Columbia University Press. [Google Scholar]
  77. Theriot MT (2009). School resource officers and the criminalization of student behavior. Journal of Criminal Justice, 37(3), 280–287. doi: 10.1016/j.jcrimjus.2009.04.008 [DOI] [Google Scholar]
  78. Warr M. (1998). Life-course transitions and desistance from crime. Criminology, 36(2), 183–216. doi: 10.1111/j.1745-9125.1998.tb01246.x [DOI] [Google Scholar]
  79. Welsh WN, & Harding C. (2015). School effects on delinquency and school-based prevention. In Krohn MD & Lane J. (Eds.), The Handbook of Juvenile Delinquency and Juvenile Justice (pp. 181–198). Malden, MA: Wiley. [Google Scholar]
  80. Wilkins L. (1965). Social Deviance. Englewood Cliffs, N.J.: Prentice-Hall. [Google Scholar]
  81. Wiley SA, Slocum LA, & Esbensen FA (2013). The unintended consequences of being stopped or arrested: An exploration of the labeling mechanisms through which police contact leads to subsequent delinquency. Criminology, 51(4), 927–966. [Google Scholar]
  82. Wiley SA, & Esbensen F. (2016). The effect of police contact: Does official intervention result. In Deviance Amplification? Crime & Delinquency, 62, 283–307. doi: 10.1177/0011128713492496 [DOI] [Google Scholar]
  83. Williams R, Allison PD, & Moral-Benito E. (2016). xtdpdml: linear dynamic panel-data estimation using maximum likelihood and structural equation modeling. Retrieved from https://www3.nd.edu/∼rwilliam/dynamic/SJPaper.pdf [Google Scholar]
  84. Wolf KC, & Kupchik A. (2017). School suspensions and adverse experiences in adulthood. Justice Quarterly, 34(3), 407–430. doi: 10.1080/07418825.2016.1168475 [DOI] [Google Scholar]
  85. Wolfgang ME, Figlio RM, Tracy PE, & Singer SI (1985). The National Survey of Crime Severity. U.S. Department of Justice, Bureau of Justice Statistics. NCJ-96017. [Google Scholar]

RESOURCES