School Discipline as a Turning Point

Abstract

Objectives:

To examine how school discipline may serve as a negative turning point for youth and contribute to increased odds of arrest over time and to assess whether suspensions received across multiple years may present a “cumulative” increase in odds of arrest.

Methods:

Using four waves of data from the National Longitudinal Survey of Youth, we use a longitudinal hierarchical generalized linear model (HGLM) to explore how school suspensions contribute to odds of arrest across time while controlling for a number of theoretically important dimensions such as race, age, delinquency, and gender among others.

Results:

Results show that youth who are suspended are at an increased risk of experiencing an arrest across time relative to youth who are not suspended and that this effect increases across time. Further, with each subsequent year the youth is suspended, there is a significant increase in odds of arrest.

Conclusion:

Supporting prior work, we find that youth who receive a suspension are at an increased odds of contact with the criminal justice system, and increases in the number of suspensions received contribute to significant increases in odds of arrest. Findings demonstrate that suspensions present a form of cumulative effect over time.

Keywords

life-course theory developmental theories criminological theory punishment

Introduction

There is a long criminological tradition exploring how life “turning points” and “transitions” impact pathways into and through adulthood (Elder 1985; Laub and Sampson 2003). Within the life-course literature, criminal justice involvement often stands as a negative turning point that obstructs prosocial bonds, reduces institutional attachments, and limits access to opportunity structures (Sampson and Laub 1997). Consequently, criminal sanctions can have a direct and indirect effect on the probability of further criminal involvement and additional formal punishment (Uggen and Manza 2002; Warr 1998). This insight becomes more salient, given the rise of punitive practices and penal sanctions over the last 30 years (Garland 2001). Perhaps more importantly, scholars suggest the cumulative effect of these practices “knife off” opportunities that can severely alter future outcomes (Hagan and Dinovitzer 1999; Sampson and Laub 1993; Western, Kling, and Weiman 2001).

The shift toward punitive governance, however, is not limited to the criminal justice system. Often termed the “criminalization of school discipline” (Hirschfield 2008; Hirschfield and Celinscka 2011), schools have increasingly relied on practices associated with the criminal justice system in order to identify and sanction student misconduct (Casella 2006; Kupchik 2010). Despite their growing presence, these practices correspond with harmful outcomes, specifically, racial disparities in punishment (Reyes 2006; Townsend 2000), decreased student engagement (Mowen and Manierre 2015), lower school performance (Gottfredson 2001; Gottfredson et al. 2005), and higher dropout rates (Nolan and Anyon 2004). Additionally, the punitive turn in school discipline has been charged with creating a “school-to-prison pipeline”—a national trend in which youth, especially racial/ethnic minorities, become enmeshed in the criminal justice system (Department of Education 2014b).

As punitive measures become a common part of the educational environment, the distinction between school discipline and criminal justice becomes increasingly blurred. Consequently, similar trends in criminal justice that create undesirable transition points for adults now appear to be in schools. While the life-course literature documents the importance of education and the effect of educational “snares” (Bersani and Chapple 2007; Moffit and Caspi 2001), it has overlooked whether the punitive sanctions associated with contemporary school disciple serve as a negative turning point for adolescents that may provoke future sanctions. Further, the school-to-prison pipeline shows that youth who are removed from school—often due to punitive punishment—are at increased chances of finding themselves within the correctional system (Department of Education 2014b). While research continues to document these effects (Kupchik 2010), what remains unclear is whether disciplinary involvement within schools has an effect on future sanctions, such as arrest.

By combining the life-course perspective with recent work on school discipline and the school-to-prison pipeline, we extend each of these literatures by assessing (1) whether school punishment acts as a negative turning point for youth by increasing their criminal justice contact in the form of arrest and (2) whether there is a cumulative effect of school discipline that corresponds to additional criminal justice involvement. To accomplish this, we use four waves of data from the National Longitudinal Survey of Youth (NLSY 1997) to explore how school suspensions relate to odds of arrest between youth who are suspended relative to those who are not, while jointly examining the effect of school suspension on arrest for youth over time. Further, we examine how school discipline may present a cumulative effect on arrest between youth due to variations in the number of waves youth reported receiving a suspension.

Literature Review

The Life-course Perspective

The life-course perspective has largely focused on how transitions create a catalyst of change in the form of turning points that can alter future outcomes (Elder 1985; Laub and Sampson 2003; Sampson and Laub 1993). To borrow from Sampson and Laub (1993:304), turning points represent “radical ‘turnarounds’ or changes in life history that separate the past from the present.” In earlier variations, the life-course perspective primarily focused on positive turning points that encourage desistance from criminal conduct. Commonly cited events include marriage, employment, educational success, mobility, parenthood, and military service (Elder, Gimbel, and Ivie 1991; Laub and Sampson 2003; Sampson and Laub 1993, 1996).

Despite a concentration on desistance, turning points can also be negative. School failure, unemployment, formal sanctions, and family instability have been found to promote criminal behavior or a persistence in crime (Bersani and Doherty 2013; Mowen and Visher 2015; Sampson and Laub 1997). Further, research finds that the adverse effects of these turning points can build on one another and increase incrementally. Often termed “cumulative continuity” or “cumulative disadvantage,” negative events can set in motion a sequence of reinforcing conditions that impact one’s future (Sampson and Laub 1997). If serious enough, these events can “knife off” the past and limit access to future opportunities (Moffitt 1993). For purposes of this effort, it is important to consider that turning points are not limited to those above; turning points can include involvement with criminal sanctions that have steadily increased.

Criminal Justice, the Life Course, and Labeling

Scholars frequently argue that the criminal justice system has experienced a punitive turn over the last 30 years (Garland 2001). This trend has resulted in a near sixfold increase throughout criminal justice processing with U.S. corrections now standing as the world’s leader in incarceration (Carson 2014). Given this expansion, many have taken stock of how criminal justice contact impacts the life course of individuals (Uggen and Manza 2002; Warr 1998). Consequently, scholars have noted that criminal justice involvement—for adults and youth—can serve as a negative turning point (Sampson and Laub 1993).

Research consistently finds that an arrest, conviction, and/or term of imprisonment creates structural impediments that block ties to conventional society and limits access to opportunity structures (Sampson and Laub 1997). Evidence suggests this cumulative disadvantage, arising from state sanctions, creates destabilizing conditions throughout the life course (Farrington 1977). Through decreasing prosocial bonds and conventional opportunities, formal punishment indirectly increases the chances of prolonged criminal conduct. Still further, the literature finds that criminal justice contact and formal punishment are closely associated with additional forms of punishment and high recidivism rates (Pettit and Western 2004; Sampson and Laub 1993). Overall, research suggests that the cumulative effect of formal sanctions can lead to adverse future consequences (Moffitt 1993).

Certainly, the life-course literature and notion of cumulative disadvantage are firmly rooted in the labeling—or societal reaction—perspective (Becker 1963; Lemert 1951). Labeling theory generally focuses on the application of stigmatizing, or criminal, labels that increase the likelihood of future offending (Becker 1963; Lemert 1972). Here, formal reactions to primary deviance are positioned as independent variables that foster, rather than curb, secondary deviance. In order to understand how sanctions and attached labels can provoke further criminal conduct, existing work largely offers two mechanisms: one that proposes an internal process (Lemert 1951) and another suggesting external factors (Paternoster and Iovanni 1989).

The first mechanism argues that delinquent labels can shift identities toward a deviant self-concept through an internalization process. Seeing that identities take on other definitions (Becker 1963), individuals given a stigmatized label may adopt a criminal-self and become more deviant (Lemert 1951). The second mechanism states that being labeled through “primary sanctions” can provoke “secondary sanctions” via increased monitoring, exclusionary policies, and reduced opportunities (Liberman, Kirk, and Kim 2014; see also Doherty et al. 2015; Farrington 1977; Paternoster and Iovanni 1989). While formal punishment can increase misconduct, research finds that negative labels can increase secondary sanctions even when controlling for criminal conduct (Liberman et al. 2014; see also Huizinga et al. 2003; Kirk and Sampson 2013; Klein 1986). Overall, once a deviant label is attached, an internal and external detachment process occurs that promotes further criminal justice contact.

The Criminalization of School Discipline and School-to-prison Pipeline

Punitive practices, however, are not confined to criminal justice. Overwhelming evidence finds that they have become a natural part of the school environment (Casella 2006). For example, according to the 2014 Indicators of School Crime and Safety published by the U.S. Department of Education and Department of Justice, schools across the nation have gradually escalated their use of security measures. Between the 1999 and 2012, data show greater use of controlled access to school grounds, check-in areas, closed campus protocols, ID badges, and strict dress codes. During the same time frame, schools increasingly adopted criminal justice–based mechanisms including the use of surveillance systems, metal detectors, drug-sniffing dogs, random sweeps for contraband, drug tests, and school resource officers (SROs; Robers et al. 2015).

In addition to security features, a large body of research finds that school officials have adopted exclusionary disciplinary policies when responding to student misconduct. In a study in Texas, Fabelo and colleagues (2010) found that roughly 60 percent of students were expelled or suspended at least once while in middle and high school. Yet, Fabelo et al. (2010) found that only 3 percent of punished misconduct required removal. In a similar vein, Losen and Martinez (2013) estimate that roughly 2 million students—or one in nine—were suspended during the 2009 academic year. Suspension rates in 2000 represent more than an 80 percent increase to those of 1974 (Sykes et al. 2015).

The national rise of criminal justice–based practices in schools has been referred to as the “criminalization of school discipline”—the process through which school environments mimic the formal justice system, given their use of restrictions, security measures, and punishment practices that redefine, recognize, and sanction student misconduct (Hirschfield 2008). Given this trend, research finds that students who are suspended or expelled are more likely to be transferred to the juvenile and criminal justice system (Fabelo et al. 2010; Na and Gottfredson 2013). This has led scholars to document and examine the rise of a “school-to-prison pipeline,” a national trend in which youth—especially racial/ethnic minorities—are enmeshed in the criminal justice system as a result of being punished under punitive disciplinary policies (Skiba et al. 2014; Wald and Losen 2003).

Similar to disparities in criminal justice processing, contemporary school discipline is also disproportionately felt by racial/ethnic minorities (Nicholson-Crotty, Birchmeier, and Valentine 2009; Peguero and Shekarkhar 2011; Skiba et al. 2011). Studies consistently find that Black and Hispanic students are more likely to be referred for discipline, punished, and arrested in school (Skiba et al. 2002). As a result, reports often conclude that minority students are two to three times more likely than their White counterparts to receive exclusionary punishment (Department of Education 2014b). These outcomes have led federal reports to conclude that students of color are systematically being denied access to education due to disproportionate discipline (Department of Education 2014a).

Impact of School Discipline on Criminal Justice Involvement

Just as research has observed the fallout of intensified criminal justice efforts, scholars have examined the damaging effects associated with punitive school discipline. Aside from potentially increasing misconduct (Gottfredson 2001; Gottfredson et al. 2005), punitive discipline has been found to lower academic performance (Gottfredson 2001; Gottfredson et al. 2005) and increase dropout rates (Losen and Martinez 2013; Nolan and Anyon 2004). Research also reveals that the use of criminal justice measures to curb student misconduct corresponds with decreased attendance, overall engagement, and educational success (Department of Education 2014b; Gregory, Skiba, and Noguera 2010; Mowen and Manierre 2015). Similarly, research indicates that harsh punishment practices have the ability to create an undesirable school climate (Ayers, Dohrn, and Ayers 2001; Lyons and Drew 2006) that hinders the quality of education (Lawrence 2006; Webber 2003).

These consequences become more salient, given the influential role education plays in shaping youths movement into and through adulthood, thus suggesting that the intensification of punitive school discipline may act as turning points for youth. As Sampson and Laub (1993:304) highlight, “turning points can modify life trajectories–they can ‘redirect paths’” and that “pathways to both crime and conformity are modified by key intuitions of social control.” Thus, scholars have examined whether education can reorder the life course by opening or closing off conventional opportunity structures (Moffitt 1993; Sampson and Laub 1997; Wright et al. 2001). Following this line of inquiry, research finds that academic disruptions—similar to those outlined above—serve as turning points that can lead to adverse outcomes through adulthood (Bersani and Chapple 2007; Elder 1998; Jimerson 1999; Thornberry, Moore, and Christenson 1985)

Current Study

Overall, the life-course perspective suggests that criminal justice involvement stands as a negative turning point, and research on the school-to-prison pipeline shows that youth who experience punitive discipline are likely to be removed from school and experience undesirable life outcomes such as incarceration (Polakow-Suransky 2001). Yet, existing research stops at documenting the connection between discipline and criminal justice contact across time and for students who remain in school. Consequently, we do not know if school-based punishment relates to future arrest. Evidence suggests that the institutional snares of school discipline can create a snowball effect of disadvantage (Moffitt 1993; Sampson and Laub 1997). However, it is unclear whether the cumulative risk of adverse future outcomes, such as arrest, increases as involvement in school discipline increases.

In order to address these gaps, we assess the relationship between school suspension and arrest between and within students longitudinally. We hypothesize the following: (1) youth who report receiving a suspension in school will be placed at a greater odds of arrest over time, and they will be significantly more likely to experience an arrest than youth who do not report receiving a suspension even when accounting for delinquency and other theoretically important control variables; (2) each increase in the number of years in which the youth reports being suspended will lead to an increase in the odds of arrest for that youth, thus presenting a cumulative effect.

Method

Data

The data used for this project are panel data taken from wave 1 (collected in 1997) through wave 4 (collected in 2000) of the NLSY97. The NLSY97 is a household-based random sample of youth between the ages of 12 and 16 (1997) in the United States. NLSY97 researchers asked each youth, as well as their parent, to complete hour-long interviews and questionnaires assessing measures dealing with education, family, delinquency, crime and arrest, and future aspirations in life at each wave (Bureau of Labor Statistics 2013). With the use of sampling weights derived from Census data, the NLSY97 is a nationally representative sample comprised 8,984 youth at wave 1.

The NLSY97 is suitable for this project for a number of reasons. First, the data contain measures on school discipline, arrest, and theoretically important control variables including measures of race, gender, and delinquency. Second, the longitudinal nature of the NLSY97 allows us to take advantage of multiple waves of data by nesting time within the individual to examine the impact of school discipline on arrest between respondents (time invariant) and within respondents (time variant). The use of multiple waves also provides a means to gain a better understanding of the cumulative effect of school discipline on arrest by assessing how receiving suspensions in multiple years relates to odds of arrest between students. Finally, although some respondents drop out of, or graduate from, school over time, because the average age of youth at wave 1 is about 14 years, the majority of respondents are in school across the first four waves of data.

Dependent Variable

The dependent variable in the current study is arrest. At each wave, youth were asked if they had been arrested since the previous interview date (not including any arrest due to minor traffic violations). Youth could answer either yes or no, making this a binary response. As shown in Table 1, on average, about 5.4 percent of the sample reported receiving an arrest at each wave. Yet, when examining the prevalence of arrest between respondents, descriptives show that about 14.5 percent of all youth reported receiving at least one arrest during the first four waves of the NLSY97. Perhaps unsurprisingly, the variance within respondents shows that youth who are arrested once appear to be at a much greater risk of reporting an arrest again. Overall, these descriptive statistics suggest very important trends in arrest both between respondents and within respondents over time.

Table 1.

Time-variant Measures.

Binary Measures	Overall		Between	Within
Binary Measures	Mean	SD	Mean	Mean
Arrest (1 = yes)	0.054	0.225	0.145	0.382
Suspension (1 = yes)^a	0.108	0.302	0.261	0.460
Continuous measures	Mean	SD	SD	SD
Age	15.3	1.98	1.45	1.37
Delinquency (log)	1.10	1.14	0.971	0.647
Income (log)	10.4	1.20	1.07	0.591
Drop out	0.067	0.249	0.181	0.171

Note: N = 7,397.

^aUsed in first modeling strategy only.

Independent Variables

The independent variable under examination within this study is school discipline. At each wave, the NLSY97 asked youth to report if they had received a school suspension in the prior year. Youth could respond either yes or no for each wave making this a binary measure. We examine suspension in two different models. First, we explore suspension as a time-variant and time-invariant measure (discussed in greater detail below). As shown in Table 1, about 10.8 percent of the sample reported receiving a suspension within each wave, though this varies between individuals and within individuals over time.

In addition, we examine the cumulative effect of receiving suspensions in multiple years as a time-invariant (between) effect. As shown in the bottom of Table 2, about 73 percent of youth reported not receiving a suspension in any year, 7 percent reported receiving a suspension in two years, about 4 percent reported receiving a suspension in three years, and slightly less than 1 percent reported a suspension in all four years (suspension in one year contrast: about 14 percent).

Table 2.

Time-invariant Measures.

	Mean	SD	Range
Race (White contrast)
Black	.229	.421	0, 1
Hispanic	.211	.408	0, 1
Other race	.010	.098	0, 1
Gender (male contrast)
Female	.518	.499	0, 1
Cumulative suspension (one suspension contrast)^a
Suspended in zero waves	.739	.438	0, 1
Suspended in two waves	.071	.256	0, 1
Suspended in three waves	.035	.184	0, 1
Suspended in four waves	.007	.085	0, 1

Note: N = 7,397.

^aUsed in second modeling strategy only.

Control Variables

Analytically, the relationship between suspension and arrest suggests that individuals who are suspended are also likely to be arrested for similar underlying reasons such as social class, race/ethnicity, and delinquency. To this end, we include a number of theoretically important control variables in the analysis including time-variant measures shown in Table 1 and time-invariant measures shown in Table 2. First, we account for the influence of race/ethnicity on arrest by including variables representing Black, Hispanic, and Other Race youth (White contrast). We also include a variable representing female (male contrast). Both gender and race are time-invariant measures and shown in Table 2.

To account for class differences in arrest, we include a control variable to account for total family income. As this measure is highly skewed, we have transformed this variable using the natural logarithm. Because this measure is time variant, descriptives are shown in Table 1. As shown in Table 1, we control for whether or not the student dropped out of school prior to graduation. This is important for two reasons. First, prior work demonstrates that students who leave school prior to graduation are more likely to experience negative outcomes including arrest (Pettit and Western 2004). In addition, and particularly important to this analysis, students who were not enrolled in school could not report a suspension. To account for this, we coded individuals who reported that they had stopped going to school prior to receiving their diploma or General Education Development (GED) as a binary measure (still enrolled contrast). As shown in Table 1, this measure is time variant, as individuals could report dropping out of school at one wave and reenroll in subsequent waves.

Finally, from both a theoretical and an empirical standpoint, we know that delinquency is generally quite correlated with arrest and suspension. For example, youth who are more delinquent should be placed at a greater likelihood than youth who are less delinquent to receive both a suspension and an arrest. Therefore, to account for delinquency, we rely on a number of measures within the data set. The NLSY97 asked youth to report how often they have engaged in any of the following activities within the previous 12 months: selling drugs, carrying a gun, belonging to a gang, destroying property, stealing an item worth less than US$50, stealing an item worth more than US$50, committing any other property crime, and attacking or assaulting someone. To create a single dimension to capture delinquency, we summed each of these seven measures to generate a count of the total number of delinquent acts engaged in over the prior year. Because this measure is negatively skewed creating heteroscedasticity and potentially biasing regression estimates (Wooldridge 2005), we have taken the natural logarithm and included this transformed measure in the analysis.

Analytic Strategy

Multilevel Longitudinal Modeling

To understand the relationship between school discipline and arrest, we model the data in two ways. First, we examine the influence of suspension on arrest as both a time-variant covariate and a time-invariant covariate through the use of a hierarchical generalized linear model (HGLM; Rabe-Hesketh and Skrondal 2012; Raudenbush and Bryk 2002). The use of this multilevel modeling approach nests time within the individual, as longitudinal panel data violate the assumption of independence made in ordinary least squares regression due to a correlated error term. HGLM introduces a random intercept to help correct for a lack of independence and endogeneity (see Osgood 2009) to model differences between and within individuals over time. For example, Bersani and Doherty (2013) show how to delineate the effects of a key independent variable into two levels of effects: between and within. To explore the within-individual effect, the independent variable is group mean centered and captures within individual change over time. To explore between-individual effects, the independent variable is aggregated across each time point for each individual (see also Horney, Osgood, and Marshall 1995). We apply this procedure to the key independent variable in this study: school suspension. The level 1 effect represents the “within” effect of suspension on arrest over time, thus capturing the time-variant effect. The level 2 effect represents the between effect of suspension, which captures the time-invariant effect. Both levels of effects are dichotomous. This procedure allows us to explore the general trends in the effect of suspension on arrest both between and within individuals in the NLSY97 sample.

Second, we also examine the cumulative effect of suspension on arrest, as an individual who is suspended in three or four waves, as opposed to just one or two, may be placed at a significantly greater odds of arrest. Because youth could report receiving a suspension at each wave (yes/no), we calculated the total number of waves each respondent reported having received a suspension. As a result, youth could report receiving a suspension in one of the five mutually exclusive categories: never, once, twice, three, or four times. Because respondents could fall into only one of these categories, these variables represent a between (level 2) effect. This allows us to examine how receiving a suspension in none, one, or multiple waves may relate to different odds of arrest. Finally, we note that this second approach would allow for an individual to report being arrested prior to receiving a suspension, therefore creating an issue of temporal ordering. To address this, we forced temporal ordering by dropping all cases where the youth reported receiving an arrest prior to receiving a suspension. Overall, this resulted in the loss of 79 respondents because it was rare that a youth reported receiving an arrest prior to a suspension.

Missing Data and Attrition

There are two important notes concerning missing data and sample attrition. First, while we note the majority of individuals remain within the average age of school attendance across all four waves of data, some individuals leave school. In addition to controlling for dropouts as we note above, we allow individuals who have graduated to drop out of the analysis. This strategy highlights one of the important advantages of HGLM in that it is robust to missing data across time (Rabe-Hesketh and Skrondal 2012). Unlike other forms of regression analysis that use pairwise deletion to exclude cases if they are missing any data across a given set of variables or waves, HGLM allows for individuals with incomplete data to contribute information across the sample time frame through the use of maximum likelihood estimates (for a description, see Rabe-Hesketh and Skrondal 2012:243). As a result, although many individuals leave school at some point following wave 1, they still contribute valuable information to the longitudinal analysis. So long as the youth were enrolled in school at wave 1, and therefore could experience a suspension, they were included in the model. For each model, we provide the minimum, average, and maximum number of waves in which youth were included in the analysis.

Second, although there is missing data across most variables in the data set, one key control variable (income) contained significantly more missing data than others. For example, in waves 3 and 4, the measure of income was missing significantly more responses than waves 1 and 2, and although HGLM is robust to missing data as we note above, proceeding with the analysis would have resulted in about a 35 percent reduction in sample size. Therefore, in order to maintain power and reduce the likelihood of introducing significant bias in the results due to sample attrition, multiple imputation in Stata 14 was used. Multiple imputation by chained equations (MICE) using all complete variables in the data set was used to impute missing values for each variable used in the analysis. In order to impute data, MICE matches variables with missing data to variables without missing data. Then, using the variance among these measures, MICE generates imputations by performing a series of univariate regressions (see Royston and White 2011). Using the results of these chained equations, missing data are imputed on a case-by-case basis using sampling weights.

Overall, by allowing youth who graduate to drop out of the sample and imputing missing data, the final sample size at wave 4 is comprised of a total of 7,397 youth, or about 82 percent of the original sample in 1997. As school discipline is often a robust—and important—contributing factor to leaving school, we performed a logistic attrition analysis on our sample to assess whether or not youth who were dropped from the sample scored significantly different on any of the measures used in our analysis (Brame and Paternoster 2003). Sensitivity analyses were performed by dichotomizing this variable and running t-tests as well as estimating a logit model to examine patterns of missing data (Brame and Paternoster 2003). Overall, these results demonstrate no significant predictors of attrition within our sample and the covariates used in the analysis, suggesting that these data are missing at random.

Results

Model 1: Between and Within Effects of Suspensions Over Time

First, we present the results of the HGLM exploring both within (level 1) and between (level 2) effects of suspension on arrest. To explore the relationship between the theoretically important control variables, suspension, and arrest, we present stepwise results by first introducing demographic variables in step 1. In step 2, we introduce delinquency, income, and dropout as additional predictors. In step 3, we introduce both the level 1 and level 2 measures of school suspensions.

As shown by step 1 in Table 3, Black youth, relative to White youth, are 48 percent more likely to report being arrested. Similarly, older respondents are more likely to report receiving an arrest; with each one year increase in age, youth become 19 percent more likely to report having received an arrest. Finally, females, relative to males, are about 67 percent less likely to report being arrested. These findings tend to support prior literature on the relationship between race, age, and gender on arrest.

Table 3.

Results of HGLM Predicting Arrest Over Four Waves of Data.

Variable	Step 1		Step 2		Step 3
Variable	Odds Ratio	SE	Odds Ratio	SE	Odds Ratio	SE
Time invariant
Black	1.48	.101***	1.37	.096***	1.08	.098
Hispanic	1.17	.108	1.09	.101	1.04	.103
Other	1.27	.419	1.22	.383	1.01	.369
Female	0.326	.089***	0.442	.081***	0.509	.082***
Time variant
Age	1.19	.021***	1.18	.021***	1.23	.021***
Delinquency (logged)	—	—	2.43	.035***	2.13	.036***
Income (logged)	—	—	0.956	.021*	0.968	.022*
Drop out	—	—	3.39	.114***	3.06	.117***
Level 1 suspension (within)	—	—	—	—	2.57	.154***
Level 2 suspension (between)	—	—	—	—	5.17	.116***
Random intercept	.064	.088***	0.072	.090***	0.065	.090***
Variance of random intercept	1.73	.076	1.31	.070	1.31	.072
Intraclass correlation	.478	.022	0.343	.024	0.342	.025
F-test	47.41***		105.05***		93.43***
Total number of time points	24,612		24,612		24,612
Total number of individuals	7,397		7,397		7,397
Minimum/average/max	1/3.3/4		1/3.3/4		1/3.3/4

Note: HGLM = hierarchical generalized linear model.

*p < .05.

**p < .01.

***p < .001.

Next, we introduce measures of delinquency, income, and dropout status in step 2 of Table 3. Overall, the effect of the time-invariant measures on arrest remains the same. With the addition of the time-variant measures, we see that as individuals increase one unit on the logged delinquency scale, the odds of receiving an arrest increase 143 percent. Likewise, an individual who drops out of school is placed at a 239 percent greater chance of reporting an arrest. On the other hand, as individuals move one unit up in logged income, their odds of arrest decrease by about 4 percent. Finally, we introduce the between and within effects of school suspension in step 3.

As shown by step 3 in Table 3, both the between (level 2) and within (level 1) effects of suspension on odds of arrest are significant. The within effect shows that an individual is 157 percent more likely to report an arrest each year they are suspended relative to a year in which they are not suspended. The between effect shows that an individual who is suspended relative to an individual who is not suspended is 417 percent more likely to report having been arrested. Like the prior steps, both dropout status and delinquency relate to increased odds of arrest, while females and higher-income students report lower odds of arrest. Although prior literature has clearly demonstrated the importance of race on arrest and suspension (see Nicholson-Crotty et al. 2009; Skiba et al. 2002; Skiba et al. 2011), once we account for the influence of between and within effects of suspension on arrest, race drops from significance.

Overall, this model demonstrates that receiving a suspension in school places an individual at increased odds of arrest over time, and individuals who are suspended relative to those who are not are significantly more likely to be arrested, even while accounting for theoretically important constructs such as self-reported delinquency, gender, race, and family income. Next, to further disentangle the impact of suspension on arrest, we run a similar model but replace the existing measures of suspension with time-invariant dummy variables representing the total number of years the youth reported receiving a suspension.

Model 2: Cumulative Effects of Suspension on Arrest

The results of this second modeling strategy are similar to the first analysis. The stepwise results (introducing demographic variables, followed by time-variant controls) are identical to the previous model; therefore, we show only the full model with the time-invariant measures of suspension in Table 4. Turning to the control variables, like the prior analysis, we do not observe any statistically significant differences across racial groups on odds of arrest once the measures of suspension are included in the analysis. Similar to the previous model, both females and individuals from higher-income families are significantly less likely than their counterparts to report receiving an arrest. Likewise, youth who reported greater levels of delinquency, those who dropped out of school, and older youth are more likely to report receiving an arrest.

Table 4.

HGLM Predicting Arrest Over Four Waves of Data Using Time-invariant Suspension.

Variable	Odds Ratio	SE
Time invariant
Black	1.04	.102
Hispanic	1.06	.104
Other	1.03	.394
Female	0.555	.073***
Time variant
Delinquency (logged)	2.08	.037***
Income (logged)	0.979	.020*
Age	1.21	.022***
Drop out	2.38	.122***
School discipline
Suspended in zero waves	0.453	.140***
Suspended in two waves	2.36	.206***
Suspended in three waves	3.52	.253***
Suspended in four waves	4.81	.399***
Random intercept	0.002	.460***
Variance of random intercept	1.31	.075
Intraclass correlation	0.343	.026
F-test	64.93***
Total number of time points	24,612
Total number of individuals	7,397
Minimum/average/max	1/3.3/4.0

Note: HGLM = hierarchical generalized linear model.

*p < .05.

**p < .01.

***p < .001.

Turning our attention to the key independent variables, findings indicate that the total number of waves in which a respondent receives a suspension is significantly related to odds of arrest. Individuals who reported no suspension in any wave reported 54.7 percent lower odds of reporting an arrest relative to an individual who was suspended within one wave. Likewise, an individual who reported receiving a suspension within two waves reported a 136 percent increase in the odds of arrest relative to a student who was suspended within only one wave. Similarly, an individual who was suspended within three waves reported a 252 percent increase in the odds arrest, and a youth who reported being suspended within all four waves reported a 381 percent increase in the odds of arrest, compared to a youth who was suspended within just one wave.

Together, results show that youth who report never being suspended in any wave are placed at a significantly lower risk of arrest than youth who are suspended once. Each increase in the number of waves a youth reports receiving a suspension significantly relates to increased odds of arrest.

Overall, the results of the two modeling strategies present similar findings. The results of the first HGLM showed that while race was a significant predictor of arrest over time, once the effect of school suspension was accounted, race was no longer a significant predictor of arrest. Supporting prior literature, the “between” effect showed that individuals who are suspended compared to those who are not are significantly more likely to experience an arrest. Exploring the “within”-individual effect showed that individuals who reported receiving a suspension reported increased odds of arrest across time. In the second modeling strategy, to understand the cumulative effect of suspension on odds of arrest, we introduced time-invariant variables capturing the total number of waves in which a youth reported receiving suspensions. Results showed that youth who reported never being suspended were placed at a lower risk of experiencing an arrest, and odds of arrest increased with each increase in the number of waves a youth reported being suspended.

Discussion and Conclusion

Overall, the goal of this article was to examine the impact of school discipline on arrest with particular attention given to exploring the cumulative effect of suspension on odds of arrest using four waves of data from the NLSY (1997). Results from two models revealed that suspension is significantly related to arrest over time within the individual and between individuals—even when delinquency is controlled. Further, we found that there is a cumulative effect of suspension on odds of arrest.

Our first hypothesis, that youth who report receiving a suspension in school would be placed at a greater odds of arrest over time and youth who are suspended would be more likely to report receiving an arrest than youth who are not suspended, is supported. Results of an HGLM reveal that youth who are suspended within one wave are likely to be arrested (level 1, time-variant effect) and that youth who are suspended are more likely to be arrested than youth who are not suspended (level 2, time-invariant effect). Further, these relationships exist even in the presence of theoretically important control variables. At present, we offer a number of theoretical explanations.

Prior work shows that formal sanctioning within the criminal justice system often contributes to future contact with the criminal justice system (Laub and Sampson 2003; Sampson and Laub 1997). As we note, scholars have shown that school discipline has become more punitive (Casella 2006; Hirschfield and Celinscka 2011), and many argue that school punishment now mimics formal sanctioning of the criminal justice system (see Kupchik 2010). Our findings suggest that youth who are punished within school are significantly more likely to be punished within the formal justice system, even when we control for self-reported levels of delinquency, race, gender, age, income, and time.

Within the life-course perspective, these findings suggest that school discipline functions as a negative turning point for some youth. Prior work, for example, has shown that contact with the criminal justice system often begets increased contact with the criminal justice system in the future (see Sampson and Laub 1993). Yet, outside of the “school-to-prison pipeline” that shows that youth who are removed from school are more likely to end up in prison than youth who are not removed from school (Department of Education 2014b), understanding how school discipline—like suspensions—relates to additional forms of criminal sanctioning such as arrest has not yet been explored. Our findings extend prior literature—including work on the school-to-prison pipeline—as our findings show that youth who are punished in school are placed at much higher risk of arrest relative to youth who are not punished and that this effect increases across time. Additionally, one of the main features of the school-to-prison pipeline is that youth who drop out of school—often due to punitive discipline—are likely to have criminal justice contact (Department of Education 2014a; Polakow-Suransky 2001). Results from our analysis show that even when youth stay in school but receive formal punishment, they are likely to experience formal contact with the criminal justice system through arrest.

Our second hypothesis that there would be a cumulative effect of school suspensions on arrest is supported. As we note above, while prior work has assessed the school-to-prison pipeline (Polakow-Suransky 2001; Wald and Losen 2003), our study is the first to demonstrate the cumulative effect of school suspension on formal sanctions within the criminal justice system. Our results indicate clear increases in the odds of arrest across time that increase with each year a youth is suspended, even when they remain in school. Drawing from prior work, our results suggest that the cumulative continuity arising from formal sanctions that may significantly increase contact with the criminal justice system (Farrington 1977) occurs within the educational system through the use of suspensions. In this case, school suspensions serve as a mechanism that increases criminal justice contact for youth. Youth who reported being suspended in multiple years reported greater odds of arrest; thus, there is a cumulative effect of school discipline on formal contact with the criminal justice system.

Our results taken together suggest that punishment within the school has the potential to significantly increase future contact with the criminal justice system, even when students remain in school. As noted above, the cumulative effect of discipline on arrest holds when accounting for delinquency. Through the labeling perspective, this suggests that primary sanctions administered in school can provoke formal secondary sanctions (arrest). That is, once a negative label is applied, the mechanisms associated with punitive school discipline—increased surveillance, exclusionary policies, and reduced opportunities—can increase youth criminal justice contact. Certainly, it is still possible that stigmatizing labels can alter self-concepts toward deviance and increase criminal justice contact through misconduct. Echoing the work of others (Doherty et al. 2015; Liberman et al. 2014), future research should further untangle and compare the effects of punitive discipline on student conduct and criminal justice contact.

In addition, though we note the importance of prior work assessing racial inequalities in both school discipline (see Skiba et al. 2002) and criminal justice contact (Fabelo et al. 2010), our analysis reveals that once school discipline is accounted for, significant differences in youth reports of arrest cut across racial and ethnic boundaries. This unexpected finding suggests that while Black students are significantly more likely than White youth to be arrested, once we account for the effect of school discipline, this significant difference is minimized. That is, while Black youth are disproportionately represented in the school-to-prison pipeline, our findings expand on previous research by showing that the more involvement a youth has with school discipline, the more likely they are to experience an arrest, and this relationship appears to cut across race/ethnicity.

Despite the contributions of this project, there are also limitations. First, we note that youth who are suspended are likely more delinquent than youth who are not suspended. As a result, it is possible that youth who are suspended are predisposed to experiencing an arrest, although we control for self-reported delinquency. Second, the life-course literature suggests that turning points impact other prosocial influences such as family, marriage, and employment (Laub and Sampson 2003; Sampson and Laub 1993, 1996). Future work should assess how school discipline may impact forms of prosocial attachment, which in turn may impact future experiences with the criminal justice system. That is, formal sanctions have been found to reduce employment prospects, create job instability, decrease financial mobility, and interrupt family structure (Hagan and Dinovitzer 1999; Sampson and Laub 1993; Uggen and Manza 2002; Western et al. 2000), which may contribute to future contact with the criminal justice system. Within our study, it may be the case that youth are suspended, then they experience a change in prosocial attachments, which then contributes to contact with the criminal justice system. Yet, it may also be the case that youth experience a change in prosocial attachment which then contributes to school discipline and arrest. Unfortunately, our analytic strategy does not allow us to examine specific pathways of these additional mechanisms.

Third, schools in the United States have experienced a significant increase in the use of SROs. It is possible that youth who are suspended in school were also arrested in school, yet we are unable to examine this, as the NLSY97 does not include data on SROs or location of arrest. Finally, we know that there are important variations in the operation of school discipline both between schools and within schools. As the NLSY97 is a household-based sample, we are unable to examine differences at the school level. It is possible that some schools use suspensions fairly and consistently (e.g., see Arum and Velez 2012), and in these cases, there may not be a link between school discipline and arrest.

Overall, our findings support prior work demonstrating that youth who experience exclusionary school discipline are likely to experience more formal sanctions within the criminal justice system (Skiba et al. 2014; see also Department of Education 2014a). While prior work on the school-to-prison pipeline demonstrates that youth who experience punitive punishment are likely to find themselves in prison due, in large part, to dropping out of school (Department of Education 2014b; Polakow-Suransky 2001; Wald and Losen 2003), our findings suggest there is also a significant and positive relationship between school suspension and arrest within students over time as well as between students. In addition, the cumulative effect of school discipline appears to further compound the likelihood that youth will experience an arrest. While prior work has shown that punitive school discipline can lead to a whole host of destructive outcomes within school, such as diminished academic performance (Gottfredson 2001; Gottfredson et al. 2005), grade retention (Losen and Martinez 2013; Nolan and Anyon 2004), and decreased extracurricular participation (Mowen and Manierre 2015), our findings show that school discipline may serve as an important negative turning point setting youth up for additional formal sanctions even when they remain in school.

Footnotes

Acknowledgements

The authors would like to thank Dr. Bianca Bersani for helpful feedback and insight on the methodological technique used in this project. The authors would also like to thank Dr. John Boman for helpful feedback on a prior draft of this manuscript.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

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

References

Arum

Richard

Velez

Melissa

. 2012. Improving Learning Environments in Schools: Lessons from Abroad. Palo Alto, CA: Stanford University Press.

Ayers

William

Dohrn

Bernardine

Ayers

Rick

. 2001. Zero Tolerance: Resisting the Drive for Punishment in Schools. New York: Free Press.

Becker

Howard S.

1963. Outsiders: Studies in the Sociology of Deviance. London, UK: Free Press Glencoe.

Bersani

Bianca E.

Chapple

Constance L.

2007. “School Failure as an Adolescent Turning Point.” Sociological Focus 40:370–391.

Bersani

Bianca

Doherty

Eggleston

. 2013. “When the Ties that Bind Unwind: Examining the Enduring and Situational Processes of Change Behind the Marriage Effect.” Criminology 51:399–433.

Brame

Robert

Paternoster

Raymond

. 2003. “Missing Data Problems in Criminological Research: Two Case Studies.” Journal of Quantitative Criminology 19:55–78.

Bureau of Labor Statistics. 2013. National Longitudinal Survey of Youth, 1997-2000. Retrieved September 15, 2015 (http://www.bls.gov/nls/nlsy97.htm).

Carson

Ann

. 2014. Prisoners in 2013. Washington DC: U.S. Department of Justice, Office of Justice Programs, Bureau of Justice Statistics.

Casella

Ronnie

. 2006. Selling Us the Fortress: The Promotion of Techno-security Equipment for Schools. New York: Routledge.

10.

Department of Education. 2014a. Dear Colleague Letter. Washington DC: U.S. Department of Education.

11.

Department of Education. 2014b. Guiding Principles: A Resource Guide for Improving School Climate and Discipline. Washington DC: U.S. Department of Education.

12.

Doherty

Elaine E.

Cwick

Jaclyn M.

Green

Kerry M.

Ensminger

Margaret E.

. 2015. “Examining the Consequences of the “Prevalent Life Events” of Arrest and Incarceration among an Urban African-American Cohort.” Justice Quarterly. doi:10.1080/07418825.2015.1016089.

13.

Elder

Glen

. 1985. “Perspectives on the Life Course.” Pp. 23–49 in Life Course Dynamics: Trajectories and Transitions, 1968-1980, edited by Elder

Glen

. Ithaca, NY: Cornell University Press.

14.

Elder

Glen H.

Jr . 1998. “The Life Course as Developmental Theory.” Child Development 69:1–12.

15.

Elder

Glen

Gimbel

Cynthia

Ivie

Rachel

. 1991. “Turning Points in Life: The Case of Military Service and War.” Military Psychology 3:215–31.

16.

Fabelo

Tony

Thompson

Michael

Plotkin

Martha

Carmichael

Dottie

Marchbanks

Miner

III Booth

Eric A.

. 2010. Breaking Schools’ Rules: A Statewide Study of How School Discipline Relates to Students’ Success and Juvenile Justice Involvement. New York: Council of State Governments Justice Center and Public Policy Research Institute.

17.

Farrington

David

. 1977. “The Effects of Public Labelling.” British Journal of Criminology 17:112–25.

18.

Garland

Dave

. 2001. The Culture of Control: Crime and Social Order in Contemporary Society. Chicago, IL: University of Chicago Press.

19.

Gottfredson

Denise

. 2001. Schools and Delinquency. New York: Cambridge University Press.

20.

Gottfredson

Gary

Gottfredson

Denise

Payne

Allison

Gottfredson

Nisha

. 2005. “School Climate Predictors of School Disorder: Results from a National Study of Delinquency Prevention in Schools.” Journal of Research in Crime and Delinquency 42:412–44.

21.

Gregory

Anne

Skiba

Russell J.

Noguera

Pedro A.

. 2010. “The Achievement Gap and the Discipline Gap: Two Sides of the Same Coin?” Educational Researcher 39:59–68.

22.

Hagan

John

Dinovitzer

Ronit

. 1999. “Collateral Consequences of Imprisonment for Children, Communities, and Prisoners.” Crime and Justice 36:121–62.

23.

Hirschfield

Paul J.

2008. “Preparing for Prison? The Criminalization of School Discipline in the USA.” Theoretical Criminology 12:79–101.

24.

Hirschfield

Paul J.

Celinska

Katarzyna

. 2011. “Beyond Fear: Sociological Perspectives on the Criminalization of School Discipline.” Sociology Compass 51:1–12.

25.

Horney

Julie

Osgood

Wayne

Marshall

Ineke

. 1995. “Criminal Careers in the Short-term: Intra-individual Variability in Crime and Its Relation to Local Life Circumstances.” American Sociological Review 60:655–73.

26.

Huizinga

David

Schumann

Karl

Ehret

Beate

Elliott

Amanda

. 2003. The Effect of Juvenile Justice System Processing on Subsequent Delinquent and Criminal Behavior: A Cross-national Study. Washington, DC: National Institute of Justice.

27.

Jimerson

Shane R.

1999. “On the Failure of Failure: Examining the Association between Early Grade Retention and Education and Employment Outcomes During Late Adolescence.” Journal of School Psychology 37:243–72.

28.

Kirk

David S.

Sampson

Robert J.

. 2013. “Juvenile Arrest and Collateral Educational Damage in the Transition to Adulthood.” Sociology of Education 86:36–62.

29.

Klein

Malcolm W.

1986. “Labeling Theory and Delinquency Policy: An Experimental Test.” Criminal Justice and Behavior 13:47–79.

30.

Kupchik

Aaron C.

2010. Homeroom Security: School Discipline in the Age of Fear. New York: New York University Press.

31.

Laub

John H.

Sampson

Robert J.

. 2003. Shared Beginnings, Divergent Lives: Delinquent Boys to Age 70. Cambridge, MA: Harvard University Press.

32.

Lawrence

Richard

. 2006. School Crime and Juvenile Justice. New York: Oxford Press.

33.

Lemert

Edwin M.

1951. Social Pathology: A Systematic Approach to the Theory of Sociopathic Behavior. New York: McGraw-Hill.

34.

Lemert

Edwin M.

1972. Human Deviance, Social Problems, and Social Control. Englewood Cliffs, NJ: Prentice Hall.

35.

Liberman

Akiva M.

Kirk

David S.

Kim

Kideuk

. 2014. “Labeling Effects of First Juvenile Arrests: Secondary Deviance and Secondary Sanctioning.” Criminology 52:345–70.

36.

Losen

Daniel

Martinez

Tina

. 2013. Out of School and Off Track: The Overuse of Suspensions in American Middle and High Schools. Los Angeles: The UCLA Center for Civil Rights Remedies at the Civil Rights Project.

37.

Lyons

William

Drew

Julie

. 2006. Punishing Schools: Fear and Citizenship in American Public Education. Ann Arbor: University of Michigan Press.

38.

Moffitt

Terrie

. 1993. “Adolescent-limited and Life-course-persistent Antisocial Behavior: A Developmental Taxonomy.” Psychological Review 100:674–701.

39.

Moffit

Terrie

Caspi

Avshalom

. 2001. “Childhood Predictors Differentiate Life-course Persistent and Adolescence-limited Antisocial Pathways among Males and Females.” Development and Psychopathology 13:355–75.

40.

Mowen

Thomas J.

Manierre

Matthew J.

. 2015. “School Security Measures and Extracurricular Participation: An Exploratory Multi-level Analysis.” British Journal of Sociology of Education. doi:10.1080/01425692.2015.1081091.

41.

Mowen

Thomas J.

Visher

Christy A.

. 2015. “Drug Use and Crime After Incarceration: The Role of Family Support and Family Conflict.” Justice Quarterly 32:337–59.

42.

Chongmin

Gottfredson

Denise

. 2013. “Police Officers in Schools: Effects on School Crime and the Processing of Offending Behaviors.” Justice Quarterly 304:619–50.

43.

Nicholson-Crotty

Sean

Birchmeier

Zachary

Valentine

David

. 2009. “Exploring the Impact of School Discipline on Racial Disproportion in the Juvenile Justice System.” Social Science Quarterly 90:1003–18.

44.

Nolan

Kathleen

Anyon

Jean

. 2004. “Learning to Do Time: Willis’ Cultural Reproduction Model in an Era of Deindustrialization, Globalization, and the Mass Incarceration of People of Color.” Pp. 133–49 in Learning to Labor in New Times, edited by Dolby

Nadine

Dimitriadis

Greg

Willis

Paul

. New York: Routledge.

45.

Osgood

D. Wayne

. 2009. “Statistical Models for Life Events and Criminal Behavior.” Pp. 375–396 in Handbook of Quantitative Criminology, edited by Piquero

Alex

Wesiburd

David

. New York: Springer.

46.

Paternoster

Raymond

Iovanni

Leeann

. 1989. “The Labeling Perspective and Delinquency: An Elaboration of the Theory and Assessment of the Evidence.” Justice Quarterly 6:359–94.

47.

Peguero

Anthony A.

Shekarkhar

Zahra

. 2011. “Latino/a Student Misbehavior and School Punishment.” Hispanic Journal of Behavioral Sciences 33:54–70.

48.

Pettit

Becky

Western

Bruce

. 2004. “Mass Imprisonment and the Life Course: Race and Class Inequality in US Incarceration.” American Sociological Review 69:151–69.

49.

Polakow-Suransky

Sasha

. 2001. “America’s Least Wanted: Zero Tolerance Policies and the Fate of Expelled Students.” Pp. 101–39 in The Public Assault on America’s Children: Poverty, Violence and Juvenile Justice, edited by Polakow-Suransky

Valarie

. New York: Teacher’s College Columbia University.

50.

Rabe-Hesketh

Sophia

Skrondal

Anders

. 2012. Multilevel and Longitudinal Modeling Using Stata. 3rd ed. Stata Press, TX: College Station.

51.

Raudenbush

Stephen

Bryk

Anthony

. 2002. Hierarchical Linear Models. 2nd ed. Thousand Oaks, CA: Sage.

52.

Reyes

Augustina H.

2006. Discipline, Achievement, and Race: Is Zero Tolerance the Answer? Lanham , MD: Rowman and Littlefield Education.

53.

Robers

Simone

Kemp

Jana

Truman

Jennifer

Snyder

Thomas D.

. 2015. Indicators of School Crime and Safety: 2013. Washington, DC: U.S. Department of Justice, Bureau of Justice Statistics and U.S. Department of Education, National Center for Education Statistics.

54.

Royston

Patrick

White

Ian

. 2011. “Multiple Imputation by Chained Equations (MICE): Implementation in STATA.” Journal of Statistical Society 45:1–20.

55.

Sampson

Robert J.

Laub

John H.

. 1993. Crime in the Making: Pathways and Turning Points throughout Life. Cambridge, MA: Harvard University Press.

56.

Sampson

Robert J.

Laub

John H.

. 1996. “Socioeconomic Achievement in the Life Course of Disadvantaged Men: Military Service as a Turning Point, Circa 1940-1965.” American Sociological Review 61:347–67.

57.

Sampson

Robert J.

Laub

John H.

. 1997. “A Life-course Theory of Cumulative Disadvantage and the Stability of Delinquency.” Pp. 133–61 in Developmental Theories of Crime and Delinquency, edited by Thornberry

Terence

. New Brunswick, NJ: Transaction.

58.

Skiba

Russell

Chung

Choong-Guen

Trachok

Megan

Baker

Timberly

Sheya

Adam

Hughes

Robin

. 2014. “Where Should We Intervene? How Infractions, students, and schools all contribute to out-of-school suspensions.” In Closing the School Discipline Gap: Equitable Remedies for Excessive Exclusion, edited by Losen

Daniel

. New York: Teachers College Press. pp. 132–146.

59.

Skiba

Rusell

Horner

Robert

Chung

Choong-Guen

Karega Rausch

May

Seth

Tobin

Tary

. 2011. “Race Is Not Neutral: A National Investigation of African American and Latino Disproportionality in School Discipline.” School Psychology Review 40:85–107.

60.

Skiba

Russell

Michael

Robert

Nardo

Abra

Peterson

Reece

. 2002. “The Color of Discipline: Sources of Racial and Gender Disproportionality in School Punishment.” Urban Review 34:317–42.

61.

Sykes

Bryan

Piquero

Alex R.

Gioviano

Jason

Pittman

Nicols

. 2015. “The School-to-prison Pipeline in America, 1972-2012.” Oxford Handbooks. Retrieved September 17, 2015 (http://www.oxfordhandbooks.com/view/10.1093/oxfordhb/9780199935383.001.0001/oxfordhb-9780199935383-e-110).

62.

Thornberry

Terence P.

Moore

Melanie

Christenson

R. L.

. 1985. “The Effect of Dropping out of High School on Subsequent Criminal Behavior.” Criminology 23:3–18.

63.

Townsend

Barbara

. 2000. “The Disproportionate Discipline of African American Learners: Reducing School Suspensions and Expulsions.” Exceptional Children 66:381–91.

64.

Uggen

Christopher

Manza

Jeff

. 2002. “Democratic Contraction? Political Consequences of Felon Disenfranchisement in the United States.” American Sociological Review 67:777–803.

65.

Wald

Johanna

Losen

Daniel

. 2003. “Defining and Redirecting a School-to-prison Pipeline.” Framing Paper for the School-to-Prison Pipeline Research Conference, The Civil Rights Project, Harvard University, Cambridge, MA.

66.

Warr

Mark

. 1998. “Life-course Transitions and Desistance from Crime.” Criminology 36:183–216.

67.

Webber

Julie A.

2003. Failure to Hold: The Politics of School Violence. New York: Rowman and Littlefield.

68.

Western

Bruce

Kling

Jeffrey R.

Weiman

David F.

. 2001. “The Labor Market Consequences of Incarceration.” Crime and Delinquency 47:410–27.

69.

Western

Bruce

Pettit

Becky

. 2000. “Incarceration and Racial Inequality in Men’s Employment.” Industrial & Labor Relations Review 54:3–16.

70.

Wooldridge

Jeffery

. 2005. “Simple Solutions to the Initial Conditions Problem in Dynamic, Nonlinear Panel Data Models with Unobserved Heterogeneity.” Journal of Applied Econometrics 20:39–54.

71.

Wright

Bradley R.

Caspi

Entner A.

Moffitt

Terrie E.

Silva

Phil A.

. 2001. “The Effects of Social Ties on Crime Vary by Criminal Propensity: A Life-course Model of Interdependence.” Criminology 39:321–48.