Journal of Criminal Justice 40 (2012) 194–201
Contents lists available at SciVerse ScienceDirect
Journal of Criminal Justice
Does prison strain lead to prison misbehavior? An application of general strain theory to inmate misconduct Robert G. Morris ⁎, Michael L. Carriaga, Brie Diamond, Nicole Leeper Piquero, Alex R. Piquero a r t i c l e
i n f o
Available online 16 January 2012
a b s t r a c t Purpose: This paper applies macro-micro General Strain Theory (GST) to predict prisoner misconduct. It is hypothesized that prison-based strain will adversely affect inmates and increase their misconduct. Methods: Data from a large southern state are used to examine how environmental strain measured at the prison level inﬂuence inmates’ violent misconduct. Analyses will include a group-based trajectory model on monthly counts of violent misconduct for the ﬁrst three years of incarceration and assess whether the strain of the environment (using a latent measure of prison deprivation as a proxy for environmental strain) distinguishes between trajectories. The analysis will employ ﬁnite multilevel mixture modeling with environmental strain as both a within- and between-class predictor, but at the prison level. Results: Findings suggest that prison strain is positively associated with prison misconduct; however, the magnitude of the effect varies across distinct inmate trajectories. Conclusions: Theoretical and practical implications are highlighted as are directions for future research. © 2011 Elsevier Ltd. All rights reserved.
Introduction Criminologists have long-focused on the extent to which institutionalization and the prison experience exert a negative effect on inmate attitudes and subsequent behavior upon release. Much of the early literature focused on the process of assimilation or “prisonization” where offenders learned the norms and mores of the prison community (Clemmer, 1958) and how the pains of imprisonment forced inmates to create a new social system in response to the deprivations of imprisonment (Sykes, 1958). Meanwhile, most of the contemporary literature focuses on importation, deprivation, and situational models of inmate adjustment to correctional environments (Berg & DeLisi, 2006; Gover, MacKenzie, & Armstrong, 2000; Tasca, Grifﬁn, & Rodriguez, 2010). In short, much of the research base has shown that offenders bring into prisons some of their own attitudes and behaviors that shape their prison experience (and subsequent outcomes), while the prison environment also exerts some inﬂuence on inmate attitudes, adjustment to prison, misconduct while institutionalized, and subsequent behavior upon release. Unfortunately, there have been few theoretical developments aimed at understanding the prison experience and the adaptations and behaviors that ensue—especially with regard to prison misconduct. Even fewer attempts have been made to sort out the extent to which adaptation to prison is a ﬁxed constant or whether it changes over time. ⁎ Corresponding author at: The University of Texas at Dallas, School of Economic, Political & Policy Sciences 800 W. Campbell Rd., GR 31 Richardson, TX 75080-3021, United States. Tel.: + 1 972 883 6728; fax: + 1 972 883 6572. E-mail address: [email protected]
(R.G. Morris). 0047-2352/$ – see front matter © 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.jcrimjus.2011.12.001
One particular criminological theory that may offer a unique lens through which to view adaptation to prison and subsequent prison misconduct is General Strain Theory (GST). Agnew (1992) developed GST as a micro-level version of strain that attempted to overcome previous problems and limitations associated with classic macrolevel conceptions of strain. Speciﬁcally, Agnew focused on both negative relations with others and negative experiences, expanded on the types of strain that individuals encounter (presentation of negative stimuli, removal of positive stimuli, failure to achieve positive goals), ascribed importance to the emotional reactions that individuals experience as a result of the various strains (e.g., anger, fear, rage, depression), and speciﬁed potential coping mechanisms that individuals may use as a way to deal with the emotional responses that arise from strainful events. It is not difﬁcult to see how GST is applicable to prison misconduct. Offenders enter prison (itself a strainful event), are presented with a series of strainful events and negative experiences (e.g., loss of freedom, few resources to purchase goods), which in turn likely generate an array of emotions that may ultimately spark misconduct while institutionalized. As Blevins, Listwan, Cullen, and Jonson (2010) observe, inmates experience a number of strains including the loss of autonomy, privacy, material goods, services, and are forced to live in overcrowded and oftentimes dilapidated conditions. Yet, to the best of our knowledge only one study has empirically applied GST to prison misconduct. Listwan, Sullivan, Agnew, Cullen, and Colvin (2011) used data from over one thousand recently released inmates in Ohio to examine whether exposure to strains associated with imprisonment (e.g., direct victimization, perception of a threatening prison environment, and hostile relationships with correctional ofﬁcers)
R.G. Morris et al. / Journal of Criminal Justice 40 (2012) 194–201
affected recidivism and found that one of four types of strain (negative prison environment—where inmates perceived the prison environment to be fearful, threatening, and violent) increased the likelihood of recidivism. 1 Although their work provides an important ﬁrst-step, several limiting features of their data and analysis prohibit more ﬁrm conclusions, including the exclusive focus on post-release recidivism, the limited range of strain-related variables, and lack of examining potential variation within their sample with respect to the effect of strains on antisocial behavior. 2 Our study builds upon and extends their research in important ways. We use longitudinal data from a sample of inmates from a large, southern prison system to study the extent to which GST can help understand prison misbehavior. Speciﬁcally, we acknowledge ﬁndings from prior research showing that the experience of imprisonment varies among inmates, with some inmates experiencing the pains more than others (Johnson, 2001; Toch, Adams, & Greene, 1987), and as noted by Listwan et al. (2011, p. 5) such inmates “ﬁnd the prison experience to be more coercive and thus feel greater levels of strain.” Our paper employs trajectory methodology, which permits an objective investigation into the extent to which there may be different groups of inmates who experience these pains differently and thus have distinct patterns of prison misconduct that unfold over time. After isolating these unique trajectories, we then examine how they may differ according to key covariates. Our paper thus contributes to theoretical, methodological, and policy-relevant literature and may help not only to assess the generality of GST but also whether there are unique misconduct trajectories and whether distinct covariates distinguish between them. Indication of such ﬁndings may be relevant for correctional programming as well as for correctional administrators who must deal with organizational complexities of the larger prison. Before we present the results of our investigation, we provide brief overviews of the literatures we seek to combine— GST and prison misconduct.
General strain theory Strain theories argue that individuals use delinquency as a coping mechanism or as a problem-solving activity in response to their inability to achieve certain goals. Over the years, this view has brought about various criticisms which have engendered several revisions to strain theory. Agnew's (1992) GST was developed in order to improve upon classic strain theories (Cloward & Ohlin, 1960; Cohen, 1955; Merton, 1938). GST expanded upon traditional strain theories by including two new categories of strain. In addition to the conventional focus on the blockage of positively valued goals, Agnew posits that strain can also manifest from exposure to noxious stimuli such as criminal victimization (Lauritsen, Sampson, & Laub, 1991), child abuse and neglect (Rivera & Widom, 1990) as well as the removal of positive stimuli such as the death of a friend or relative and parental divorce/separation (Williams & Uchiyama, 1989). According to Agnew, exposure to any of these three types of strain increases the likelihood that an individual will engage in delinquency. The relationship between strain and deviance is not always direct because exposure to strainful events can lead an individual to experience a variety of negative emotions. For instance, Agnew (2001) explains that strain is likely to bring about feelings of disappointment, depression, fear, anger, or frustration. These negative emotions can operate as a mediating link between strain and delinquency (Agnew, 2001; Jang & Johnson, 2003). Anger, in particular, seems to be the most inﬂuential emotion because it increases the individual's sensitivity to the strainful event(s), incites retaliation, provides enough vigor to act upon frustrations and lowers inhibitions (Agnew, 2001; Mazerolle & Piquero, 1998). However, not everyone who is exposed to strain engages in delinquency. Typically, strainful experiences are most likely to lead
to criminal behavior if the individual is unable to cope with the strain(s) through legitimate coping mechanisms (Agnew, 2001; Broidy, 2001). In particular, those who perceive their strains as unjust and high in magnitude, who have low social control or added incentive to become delinquent are the most likely to engage in crime (Agnew, 2001, 2006). Furthermore, individuals who are both high in negative emotionality and low in constraint are also at a greater risk of becoming delinquent after strainful experiences (Agnew, Brezina, Wright, & Cullen, 2002). Agnew (2002) also points out that the effect of strain is not limited to personal encounters with strain, as strains that are anticipated or experienced vicariously can also have a positive effect on delinquency. Since its inception, GST has been subjected to a substantial amount of empirical research. Studies have consistently found overall support for the main hypothesis that strain positively inﬂuences delinquency, both directly and indirectly (i.e., mediated by negative emotions) (Agnew, 2006; Agnew & White, 1992; Johnson & Morris, 2008; Mazerolle & Piquero, 1998; Paternoster & Mazerolle, 1994). GST has also served to explain the impact that certain demographic factors such as race (Eitle & Turner, 2002; Jang, 2007; Jang & Johnson, 2003; Perez, Jennings, & Gover, 2008; Piquero & Sealock, 2010) and gender (Broidy & Agnew, 1997; Eitle, 2002; Hay, 2003; Hoffmann & Su, 1997; Jennings, Piquero, Gover, & Perez, 2009; Mazerolle, 1998; Piquero, Fox, Piquero, Capowich, & Mazerolle, 2010; Piquero & Sealock, 2000, 2004) have on delinquency. Although some studies have yielded mixed results regarding the straindelinquency relationship, almost all have shown a positive relationship between at least one measure of strain and delinquency (Agnew et al., 2002; Brezina, 1996; Capowich, Mazerolle, & Piquero, 2001; Johnson & Morris, 2008; Mazerolle & Maahs, 2000; Mazerolle, Piquero, & Capowich, 2003). GST has primarily been used to test the relationship between strain and crime, especially among juveniles (Broidy, 2001; Listwan et al., 2011). Consequently, with the exceptions noted earlier (Blevins et al., 2010; Listwan et al., 2011), GST has not been fully applied to one of the most strained populations, inmates, in arguably one of the most strain-laden environments, prisons. Given the stressful conditions of prison life, it is surprising that research has not tested the applicability of GST to the explanation of prison misbehavior. On the heels of Blevins et al. (2010), we combine GST with traditional theories of prison misconduct in order to better inform our understanding of inmate behavior. Prison misconduct Three theoretical perspectives dominate the prison misconduct literature: the deprivation model, the importation model, and the situational model. While each perspective covers different features of the prisoner and prison experience, the three are often used in tandem to provide a holistic representation of the causes of prison misconduct (Wooldredge, 2003). First, the deprivation perspective posits that the indigent and brutal realities of prison life provide the catalyst to inmate misconduct (Sykes, 1958). Thus, it is the pains of imprisonment (i.e., custody level, sentence length, prison gang membership) that may drive inmates to rebel against prison authority and be cited for misconduct (Camp, Gaes, Langan, & Saylor, 2003; Cunningham & Sorensen, 2007; DeLisi, Berg, & Hochstetler, 2004; Morris, Longmire, Bufﬁngton-Vollum, & Vollum, 2010; Worrall & Morris, 2011). Alternatively, the importation model places the onus of blame on those factors that an individual possesses prior to incarceration and, therefore, focuses on what the inmate brings into the institution (Irwin & Cressey, 1962). Certain individual characteristics such as educational attainment, marital status, and age may inﬂuence an inmate's response to incarceration. Research suggests that less educated, unmarried, and younger offenders tend to engage in more prison misconduct than their counterparts (Cao, Zhao, & Van Dine,
R.G. Morris et al. / Journal of Criminal Justice 40 (2012) 194–201
1997; Ellis, Grasmick, & Gilman, 1974; Gendreau, Goggin, & Law, 1997; Jiang & Winfree, 2006; Wooldredge, 1991). Findings are less clear concerning factors such as violent offending history (DeLisi et al., 2004; Gendreau et al., 1997; Wooldredge, Grifﬁn, & Pratt, 2001) and prior incarcerations (Cao et al., 1997; Cunningham & Sorensen, 2006, 2007; Morris et al., 2010; Sorensen & Pilgrim, 2000). Lastly, the situational perspective views the prison environment as active, rather than a static assemblage of inﬂuences. Explanations for prison misconduct are found in the interplay between inmate and the prison milieu (Steinke, 1991). Factors such as prison architecture, staff characteristics and even temperature can serve as dynamic circumstances that inﬂuence the likelihood of misconduct at any given time (Camp et al., 2003; Jiang & Fisher-Giorlando, 2002; Morris & Worrall, 2010; Steinke, 1991). Blevins et al. (2010) recently attempted to explain prison misconduct by integrating GST with the deprivation and importation theories of misconduct (along with the work of Toch, 1977). They argue that the deprivation perspective encapsulates three sources of strain (thwarting desired goals, denial of positive stimuli and the introduction of aversive stimuli) and that the importation perspective covers those characteristics of the individual that either promote or inhibit the ability to cope with these strains. Therefore, inmates with limited access to programs, living in dismal conditions, and stripped of their autonomy, will develop anger and umbrage toward the institution. These deprivation factors essentially produce strains that are ﬁltered through importation factors. An inmate's response to these strains is conditioned by personal characteristics such as age, education level and propensity for violence, making them more or less prone to misbehavior. In sum, while Blevins et al. (2010) argue for the capability of GST in providing an explanation of prison misconduct, research has yet to test the applicability of this framework to the unique circumstances surrounding prison misconduct (although see Listwan et al., 2011). The present study uses data from a large sample of state prisoners along with group-based trajectory modeling to assess whether the strains of imprisonment differentially impact inmates’ involvement in prison misconduct over time.
Data and methods Data used in the present study were culled from archival records representing the population of male inmates from a large southern state's primary corrections agency. Information was provided on each inmate's personal demographic characteristics, criminal offense history, as well as complete disciplinary infraction histories chronicling reports of misconduct for each day of incarceration. Information speciﬁc to the prison unit to which an inmate was assigned on the date of any given act of misconduct was also gathered. In the end, the data represent the cohort of male inmates entering one of 47 traditional prison facilities (excluding state jails, private prisons and substance abuse treatment) at any time between August 2004 and June 2006 and who served a continuous three years of their sentence for their offense of record (i.e., most recent offense). 3 This approach allowed for inmate behavior, as deﬁned by ofﬁcial disciplinary records, to be observed continuously, or longitudinally, over the observation period from the start of each inmate's speciﬁc entry date. For the present study, our interests lie in the effect of environmental strain as a potentially exacerbating factor on the development of misconduct among inmates. Therefore, we further limited the data to inmates who had at least one reported instance of violent misconduct (deﬁned below) reported at any time during the observation period. Inmates sentenced to death or life without the possibility of parole were also excluded because of their unique conﬁnement circumstances may make them less likely to engage in many forms of misconduct (DeLisi et al., 2004; Morris et al., 2010). This procedure
resulted in 6,328 inmates eligible for analysis. Descriptive statistics are presented in Table 1. Measures Data were collected for both inmates and the prison facility for which they were assigned when a disciplinary infraction occurred. Both inmate-level and prison unit-level measures are described in detail below. Analyses were limited to those inmates reported to have had at least one violent incident during the observation period. This was important as our focus is on how strain may evoke antisocial behavior, not abstention from it. Removing abstainers from the data also served to help eliminate heterogeneity by excluding the most qualitatively distinct group of inmates, those who are never reported for violent misconduct. Inmate-level measures Outcome variables The primary outcome measures were repeated observations for each inmate reﬂecting the number of reported acts of misconduct occurring within a series of six-month intervals over a three-year observation period (totaling 6 time periods, i.e., the number of infractions from admittance through month 6, from month 7 through 12, and so on). The misconduct outcomes represent any and all forms of misconduct reported by the agency during any given time period, respectively, and ranged in seriousness from common minor violations (e.g., verbal disruptions) to the more sparsely occurring extreme acts of violent misconduct (e.g., assaulting an inmate or guard that resulted in a serious injury). All forms of misconduct were eligible for at least minimal penalty from the prison administration. Inmate-level covariates Following prior inmate misconduct literature, several inmatelevel covariates were incorporated into the analyses. These variables served as control variables to assess the effect of environmental strain, including the inmate's age at the time of entry (for the offense of record), race (1 = White, 0 = Non-White), incarceration history (i.e., number of incarcerations), educational attainment, marital status (1 = married, 0 = else), IQ score, conﬁrmed prison gang membership (1 = gang member, 0 = non-member), sentence length in months (logged due to skewness), and whether they were incarcerated for a violent crime (1 = violent offense, 0 = else). Educational attainment scores and IQ scores were gathered by the corrections agency during the intake period of each inmate's most Table 1 Sample Demographics Combined
8.99 0.42 0.81 3.86 0.44 13.18 0.37 1.57 0.48
27.23 8.45 0.23 0.42 0.41 0.84 6.43 3.87 0.73 0.44 88.37 13.07 0.17 0.38 8.20 1.56 0.63 0.48
23.92 6.04 0.17 0.38 0.27 0.74 6.75 3.77 0.85 0.35 87.35 12.81 0.17 0.38 7.98 0.93 0.63 0.48
29.80 31.48 30.49 1.04 -0.11 1.02 0.09 0.33 0.09
27.00 24.16 0.05 1.02 0.31 0.08
Inmate-level Variables Age 27.53 8.79 27.82 White 0.23 0.42 0.23 Prior Incarceration 0.42 0.81 0.42 Education 6.46 3.86 6.45 Married 0.73 0.44 0.73 IQ 88.12 13.14 88.07 Gang 0.17 0.37 0.16 Sentence Length 8.20 1.55 8.21 Violent Offense 0.63 0.48 0.62 Prison-level Variables Age of unit 30.70 29.80 30.61 Deprivation -0.13 1.04 -0.14 Percent with 0.33 0.09 0.33 priors N 6,328 4,465
R.G. Morris et al. / Journal of Criminal Justice 40 (2012) 194–201
recent prison sentence. Educational attainment was based on an exam that resulted in a grade-level equivalence score ranging from 1 to 13 where 13 reﬂected an educational level beyond high school. IQ score was the test result of the WAIS-R intelligence quotient exam. Prison unit-level measures Of particular importance to the present study were prison-level characteristics. The prison-level variables included a latent measure of environmental stressors (i.e., deprivation), the age of the prison facility 4 (in years), and an indicator of the proportion of inmates assigned to a prison unit who had more than one experience with being incarcerated. Each prison-level measure utilized has been shown to be predictive of misconduct in cross-sectional studies of prison misconduct. Our primary independent variable, deprivation, or environmental strain in the parlance of GST, was composed of ﬁve prison-unitspeciﬁc characteristics that were present during the time of data collection. Recall that for Agnew (2006, pp.29-48), strains are things that can include speciﬁc negative experiences and events as well as a more general set of negative or adverse issues. Being housed in a highsecurity facility for example, speaks to the criminogenic potential of other individuals near the inmate, while exposure to gang members also increases the likelihood of offending and victimization. Similarly, over-crowded conditions are likely to lead to a sense of the loss of personal space and freedom as well as increased feelings of hostility, while being surrounded by inmates who have demonstrated previous violence is also likely to create a strainful and criminogenic milieu. Toward this end, Agnew (1999; see also Brezina, Piquero, & Mazerolle, 1999) has paid close attention to the macro-level analog of GST, macro-level strain theory, to focus on how GST explains community differences in offending—and the prison is one such community that centrally locates strainful and deprived conditions. Additionally, the cumulative nature of these strains (Agnew, 2006, p.81) is also likely to increase the likelihood of negative behaviors, including offending, victimization, and prison misconduct. Following Morris and Worrall (2010), the environmental strain index was composed of: (1) the proportion of inmates in the unit who were conﬁrmed prison gang members, (2) the proportion of inmates classiﬁed to high-security custody, (3) the unit's maximum inmate capacity, (4) the proportion of inmates convicted of a violent offense, and (5) a prison gang composition index that reﬂected the balance of different prison gangs within each unit. Internal consistency for the environmental strain measure was strong (α = .87). Factor analysis suggested a highly singular construct, thus factor scores were used to represent environmental strain. Analytical procedure As discussed above, this study examines the effect of environmental strain, deﬁned through deprivation theory, on inmate misconduct both within and between developmental trajectories of misconduct. 5 The analyses are multifaceted and were carried out in steps, all of which explicitly account for the nested nature of the data. Here, the data are nested at three levels: (a) repeated observations nested within (b) inmates, who are then nested within (c) prison units. The ﬁrst step was to estimate a longitudinal latent class analysis (LLCA), which is one form of group-based trajectory modeling (e.g., ﬁnite-mixture modeling). The LLCA procedure systematically classiﬁes observed misconduct trajectories into one of a speciﬁed number of latent trajectory classes. Details of the analytical procedures are discussed elsewhere (Feldman, Masyn, & Conger, 2009), but it is important to note that LLCA assumes no particular functional form of development and essentially classiﬁes patterns of development. For this approach, a random-intercept was modeled to account for betweenprison variation in trajectories of misconduct (i.e., a three-level model). For this component, we discuss the protocols used to
determine the number of latent developmental proﬁles (i.e., classes) to retain below. Upon specifying a base trajectory model, we estimated the effect of each of the above-noted covariates on predicting membership to one class or another. This approach allows us to examine whether exposure to environmental strain distinguishes between trajectory classiﬁcation, net of other inmate and prison-level inﬂuences. This between-class approach explicitly accounts for developmental group heterogeneity in prison behavior, which has not been previously carried out in studies of prisoner misconduct, but has become routine in developmental studies of criminal behavior (Piquero, 2008). In the end, this component involves a multilevel multinomial logistic regression model predicting class membership. Upon determining whether and how strain might impact classiﬁcation to a particular trajectory group, we estimated the withinclass effect of environmental strain on both the within-class intercept and slope of each speciﬁc trajectory using the same outcome measures and predictor variables. Thus, beyond assessing whether strain predicts class membership, we examine whether strain increases the expected rate of misconduct within each speciﬁc classiﬁcation of inmates and whether this effect changes (i.e., grows stronger or weaker) over the course of time. For this latter component, we estimated group-speciﬁc multilevel models for change, also referred to as growth curve models or hierarchical generalized linear models (Raudenbush & Bryk, 2002) and assessed the net effect of environmental strain on both the intercept and growth parameters of development. Results Trajectory analysis The goal of trajectory analysis or ﬁnite mixture modeling (Land, McCall, & Nagin, 1996; Nagin, 2005) is to assess and classify behavioral changes over a speciﬁed period of time and it is commonly used in developmental/life-course criminology (Piquero, 2008). While there are varying types of trajectory modeling techniques (e.g., hierarchical generalized linear modeling (HGLM), ﬁnite mixture modeling, growth mixture modeling, latent class growth curve modeling, longitudinal latent class analysis (LLCA)), each approximates a continuous distribution of developmental trajectories in the population by a ﬁnite number of groups or (latent) classes and each are based on the generalized linear model (Feldman et al., 2009). Speciﬁc differences across these methods involve assumptions about proportional odds, assumptions about the functional form of development over the observation period, and/or about how within-class variability (i.e., random effects) is treated. One form of trajectory analysis, LLCA, does not model growth along a single continuous scale per se (i.e., time), rather LLCA models patterns of states over a period of time, here the number of reported violent infractions for a six-month period (Feldman et al., 2009). An advantage of this approach is that it assumes no particular function of time, nor does it assume proportional odds, thus complex patterns of change (e.g., periods of intermittency) can be assessed without violating model assumptions. Upon classiﬁcation to a particular latent class of states (i.e., trajectories), covariates can be used to model the likelihood of belonging to one class over another. 6 The ﬁrst step in a trajectory analysis is to determine the appropriate number of classes, which is usually carried out using unconditional trajectory models (i.e., models without covariates; see Nylund & Masyn, 2008). Indices of comparative ﬁt are commonly used to assess model selection (i.e., the number of classes to retain and which form of modeling technique to capitalize upon). Here, multiple forms of trajectory analysis were carried out and it was determined that a multilevel LLCA assuming a Zero-Inﬂated Poisson (ZIP) distribution on the outcomes was appropriate. 7 We assessed Bayesian Information
R.G. Morris et al. / Journal of Criminal Justice 40 (2012) 194–201
Criterion (BIC), posterior probabilities, visualization plots, and group proportions to determine the number of latent classes to retain. Fit statistics are presented in Table 2 for a two, three, and four latent class solution. A preponderance of evidence from the LLCA models suggested the use of the 3-class solution, which was retained for subsequent analyses. The estimated trajectory proﬁles are graphically illustrated in Fig. 1 and show three distinct developmental proﬁles. We termed the proﬁles as the chronic class, the early-onset limited class, and the delayed-onset/lower-risk class.
Predicting class membership The second component of the analysis involved estimating the effect of environmental strain on distinguishing membership across developmental proﬁles. This involved using a nominal indicator of class membership as an outcome variable (1 = delayed-onset, 2 = earlyonset limited, and 3 = chronic) for a multinomial logistic regression model in which the chronic class was the reference category. As a reminder, we estimated multilevel models throughout, thus inmatelevel intercepts were allowed to vary across prison units. Results are presented in Table 3. This component was carried out in two steps (models 1 and 2). Model 1 estimates a base model where all inmate-level covariates are assessed. Model 2 extends the base model by incorporating prison-level covariates. As shown, incorporating prison-level effects in predicting class membership partially or fully confounds some, but not all, inmate-level covariates and draws out other effects. Our focus variable, environmental strain which was included as part of Model 2, is statistically signiﬁcant and negative for both stratiﬁcations of analysis. In other words, the chronic class may be characterized by acute susceptibility to such strain, net of other effects, relative to the early-onset limited and delayed-onset proﬁles, respectively. Alternative parameterizations of the reference category (not shown) suggested that the environmental strain effect is statistically indistinguishable between the early-onset and delayed-onset limited classiﬁcations. Overall, these ﬁndings suggest that the chronic class can be characterized by relatively stronger reactivity to increased exposure to environmental strain. Compared to delayed-onset inmates, chronics are also more likely to be younger, less educated, have lower IQ scores, and slightly shorter sentences. Compared to the early-onset limited group, chronics are younger, have a more extensive incarceration history, but have longer sentence lengths. While interesting, the results to this point do not yet provide the information needed to determine whether strain matters in explaining within-class development (i.e. change) of misconduct, which is the focus of the following section.
Table 2 ZIP Longitudinal Latent Class Analysis Results
Posterior Probabilities and Class Proportions for 3 Group Model
Class 1 Class 2 Class 3
0.863 0.753 0.808
70.5 25.8 3.7
# 4465 1630 233
Posterior Probabilities and Class Proportions for 4 Group Model Class 1 Class 2 Class 3 Class 4
.695 .860 .774 .830
14.2 81.2 3.1 1.5
901 5140 194 93
Solid Line = Chronic Class Dotted Line = Early-onset Limited Class Dash-dot Line = Delayed-onset Class
Fig. 1. Violent Misconduct Trajectory Proﬁles.
Does environmental strain predict misconduct within various trajectories? The results from the within-class analysis are presented in Table 4. This component involved the estimation of class-speciﬁc multilevel models for change where repeated observations over time were nested within an inmate, and inmates were nested within prison units (i.e., a 3-level multilevel model for change, or growth curve model). For each model, both linear and quadratic functions of time (i.e., growth parameters) were speciﬁed and since the outcomes were over-dispersed counts, the negative binomial link function was
Table 3 Multilevel-Multinomial Logistic Regression Predicting Class Membership Model 1 DELAYED-ONSET versus CHRONIC INMATE-LEVEL COVARIATES Est. AGE .087 WHITE -.124 PRIOR INCARCERATION -.242 EDUCATION .159 MARRIED -.331 IQ .018 GANG MEMBER -.390 SENTENCE LENGTH .200 VIOLENT OFFENSE -.285 PRISON-LEVEL COVARIATES AGE OF UNIT DEPRIVATION PERCENT WITH PRIORS INTERCEPT -2.554
Model 2 SE .026 .217 .128 .023 .250 .007 .199 .094 .190
p .001 .569 .059 .000 .186 .012 .050 .033 .134
Est. .078 -.139 -.228 .154 -.321 .019 -.346 .210 -.299
-.013 -1.343 2.608 -2.994
p .002 .744 .024 .858 .424 .052 .341 .000 .078
Est. .116 -.161 -.666 .004 .663 .028 -.604 -7.379 -.757
SE .038 .449 .320 .104 .782 .016 .699 1.873 .421
p .002 .720 .037 .967 .397 .074 .388 .000 .072
-.007 -1.047 -.154 49.328
.012 .371 3.690 12.738
.538 .005 .967 .000
EARLY-ONSET LIMITED versus CHRONIC INMATE-LEVEL COVARIATES Est. SE AGE .113 .036 WHITE -.149 .456 PRIOR INCARCERATION -.692 .306 EDUCATION -.019 .105 MARRIED .686 .859 IQ .029 .015 GANG MEMBER -.593 .623 SENTENCE LENGTH -7.682 2.048 VIOLENT OFFENSE -.726 .412 PRISON-LEVEL COVARIATES AGE OF UNIT DEPRIVATION PERCENT WITH PRIORS INTERCEPT 51.347 13.919
SE .024 .216 .130 .023 .243 .007 .201 .096 .189
p .001 .521 .080 .000 .186 .011 .085 .028 .113
.009 .226 2.008 1.396
.125 .000 .194 .032
R.G. Morris et al. / Journal of Criminal Justice 40 (2012) 194–201
Table 4 Within-Class Multilevel Model for Change Results (3-Level Analysis) Delayed-Onset
INMATE-LEVEL COVARIATES AGE WHITE PRIOR INCARCERATION EDUCATION MARRIED IQ GANG MEMBER SENTENCE LENGTH VIOLENT OFFENSE
-.097 .025 .013 -.087 .035 -.041 .003 -.033 .021
.017 .014 .013 .013 .015 .012 .013 .013 .013
.000 .070 .317 .000 .022 .001 .804 .009 .092
-.080 .024 .019 -.079 .054 -.065 .051 -.047 .035
.030 .016 .023 .029 .023 .026 .019 .022 .027
.008 .120 .410 .006 .017 .014 .007 .033 .204
-.061 -.069 -.106 -.015 .131 -.022 -.036 -.200 .025
.125 .055 .066 .048 .064 .072 .040 .074 .046
.624 .214 .107 .749 .040 .760 .369 .007 .589
PRISON-LEVEL COVARIATES AGE OF UNIT DEPRIVATION PERCENT WITH PRIORS Constant
.001 .133 .009 -.438
.018 .019 .018 .073
.966 .000 .611 .000
.000 .120 .008 -.438
.030 .022 .026 .073
.996 .000 .754 .000
.049 .192 .046 -.907
.057 .045 .055 .164
.394 .000 .399 .000
utilized. These models estimate the effect of the predictor variables on class-speciﬁc trajectories or in our case, the impact of covariates on the development of misconduct for the delayed-onset, earlyonset limited, and chronic classes, respectively and not the probability of classiﬁcation to one group over another. As shown, the impact of environmental strain on misconduct is positive and statistically signiﬁcant for each respective class, net of other effects. Thus, regardless of which trajectory an inmate is assigned, if he is assigned to a unit that is characterized by higher levels of environmental strain, then we would expect his misconduct to be more pronounced at any given time during the observation period. On its face, the environmental strain effect is strongest for the chronic class (est. = .192), followed by the delayed-onset class (est. = .133), and is weakest for the early-onset limited class (est. = .120). However, post hoc tests of coefﬁcient equality (Paternoster, Brame, Mazerolle, & Piquero, 1998) suggested that differences between any two effects were not statistically signiﬁcant. In sum, while environmental strain plays a role in predicting which trajectory an inmate falls into, the group-speciﬁc effect of environmental strain is statistically indistinguishable from one group to the next. 8 Discussion The pains of imprisonment have been a core feature of theoretical and empirical research, yet theoretical development and applications of existing theory to the variation in inmate behavior has not kept pace. One theory that may have much to offer with respect to understanding prison misbehavior is Agnew's GST, which focuses on the strain that individuals are subject to and then, in turn, how they react to strain through an array of coping mechanisms—including offending (in our case, prison misconduct). Our study examined how a measure of environmental strain, assessed at the prison level, inﬂuenced inmates’ prison misconduct. A unique feature of our study was its application of a group-based trajectory model to examine heterogeneity in misconduct. Findings showed that prison strain was positively associated with violent prison misconduct and that that the magnitude of the effect varied somewhat across different inmate trajectories. To be sure, our study was limited in some respects. First, we did not have the full range of GST-related variables, thereby prohibiting a more complete assessment of the theory. At the same time, it is likely that no public data source contains the array of variables needed to fully test GST—especially a measure of negative emotionality, which is an important variable within the theory. Nevertheless, that we
established prison environment strain to be an important correlate of inmate misconduct provides an important ﬁnding consistent with GST as well as a useful baseline to which future results can be compared. Second, our analysis was unable to examine various individual-level characteristics that inmates may bring into the facility with them, and that, in turn, could inﬂuence how they adapt (or fail to adapt) to the prison environment. Future research should carefully consider a larger range of such measures. Third, it would be useful for subsequent research to conduct moderation analyses. One project would examine the extent to which an inmate's prior offending history conditions how they adjust to the prison environment. For example, do individuals with a minor criminal history become ‘damaged’ by the prison environment and thus engage in (and are cited for) prison misconduct? This question invokes the notion of state dependence. Within the context of GST, it may be good to examine how inmates who have more extensive prosocial coping skills deal with prison strain. Perhaps among these inmates, their stock and use of coping functions as a buffer and they somehow manage to avoid conﬂict-laden situations (see Hochstetler, DeLisi, & Pratt, 2010). GST also offers much to consider with respect to correctional policy. Most importantly, speciﬁc features of prison environments appear to do more harm than good. Thus, identiﬁcation of adverse features of correctional institutions highlight important points of intervention for correctional administrators as they seek to build and retain facilities that are designed to reform and rehabilitate. Finally, the data were limited to ofﬁcial records, which cannot account for misconduct that were unknown to or not recorded by the corrections agency. Our analytic framework and study ﬁndings also relate to future GST-related research that could examine speciﬁc policy changes. For example, our ﬁndings showed that environmental strain was related to prison misconduct—and its effect was most pronounced among the chronic trajectory. Correctional initiatives or policy implementations aimed at limiting these sorts of strain could reduce inmate misbehavior, which in turn, could also lessen the likelihood of recidivism upon release (as research shows that prison misconduct is a strong correlate of post-release recidivism). Of course, reducing prison strain is a worthwhile endeavor more generally and its adverse effects are not necessarily limited to a speciﬁc group of inmates. Nevertheless, as some policies may affect some types of inmates more so than other types of inmates, researchers and policy ofﬁcials should carefully consider exploring the variability in what inmates bring into prison, the variability in how the prison inﬂuences inmates, and then in turn, how inmates deal with and adjust to the strains of imprisonment. In this regard, GST offers promise as a useful theoretical
R.G. Morris et al. / Journal of Criminal Justice 40 (2012) 194–201
framework that is easily grasped and assessed by researchers and corrections ofﬁcials. Notes 1. They also found that not only was a negative prison environment related to a higher likelihood of re-incarceration, but so too was negative relations with other inmates. 2. We recognize that their primary interest was in predicting recidivism, in a customary recidivism-oriented framework. 3. The decision to explore the ﬁrst three years of incarceration was made due to prior research indicating that this is the period for which inmates are most likely to engage in misconduct (see Adams, 1992; Bottoms, 1999; Flanagan, 1980; Grifﬁn & Hepburn, 2006; Zamble, 1992). 4. The age of the prison unit was included to reﬂect potential variation in any prison-guard culture that may be speciﬁc to the geographic location of the prison unit (Morris & Worrall, 2010). 5. Group-based trajectory (GBT) modeling has been used to understand life course transitions in offending. Nagin and Land (1993) developed the technique as a way to draw out underlying heterogeneity and extract homogenous groups with similar developmental trajectories (see review in Piquero, 2008). Despite the popularity of group-based models for explaining the progression of offending (Bushway, Piquero, Broidy, Cauffman, & Mazerolle, 2001; Cohen, Piquero, & Jennings, 2010; Hay & Forrest, 2006; Laub, Nagin, & Sampson, 1998; Nagin, Farrington, & Mofﬁtt, 1995; Sampson & Laub, 2003), it has yet to be applied to misconduct within prison settings. GBT models are well suited to tease out unobserved heterogeneity in the progression of prison misconduct, thereby providing a more nuanced understanding of how inmates respond and/or adapt to prison environments. 6. At the prison-unit level, a “common factor” was utilized to improve model efﬁciency and to reduce demands on computation. Following Henry and Muthén (2010), a common factor was used to model the random means and covariances at the prison unit level for the LLCA model. We found that the common factor approach produced better model ﬁt and greatly reduced computation time in generating model estimates. The approach has been recommended for multilevel mixture models, such as that applied here (see Asparouhov & Muthén, 2008; Vermunt, 2008). 7. Models were also run using the negative binomial link function. BIC statistics were not improved by doing so and posterior probabilities were reduced and group proportions were questionable. The functional form of latent classes for the 3-class LLCA model between the negative binomial model, Poisson, and ZIP models were similar. 8. How the effect of environmental strain operates within a particular class's development is an interesting question. To assess whether this strain effect remains stable or changes over the course of time, we also assessed an interaction effect between strain and the linear function of time (results not shown). This effect failed to reach statistical signiﬁcance for any particular latent class.
References Adams, K. (1992). Adjusting to prison life. Crime and Justice: A Review of Research, 16, 275–359. Agnew, R. (1992). Foundation for a general strain theory of crime and delinquency. Criminology, 30, 47–87. Agnew, R. (1999). A general strain theory of community differences in crime rates. Journal of Research in Crime and Delinquency, 36, 123–155. Agnew, R. (2001). Building on the foundation of general strain theory: Specifying the types of strain most likely to lead to crime and delinquency. Journal of Research in Crime and Delinquency, 38, 319–361. Agnew, R. (2002). Experienced, vicarious and anticipated strain: An exploratory study on physical victimization and delinquency. Justice Quarterly, 19, 603–632. Agnew, R. (2006). Pressured into crime: An overview of general strain theory. Los Angeles, CA: Roxbury. Agnew, R., Brezina, T., Wright, J. P., & Cullen, F. T. (2002). Strain, personality traits, and delinquency: Extending general strain theory. Criminology, 40, 43–71. Agnew, R., & White, H. R. (1992). An empirical test of general strain theory. Criminology, 30, 475–499. Asparouhov, T., & Muthén, B. (2008). Multilevel mixture models. In G. R. Hancock & M. Samuelsen (Eds.), Advances in latent variable mixture models (pp. 27–51). Charlotte, NC: Information Age. Berg, M. T., & DeLisi, M. (2006). The correctional melting pot: Race, ethnicity, citizenship, and prison violence. Journal of Criminal Justice, 34, 631–642. Blevins, K. R., Listwan, S. J., Cullen, F. T., & Jonson, C. L. (2010). A general strain theory of prison violence and misconduct: An integrated model of inmate behavior. Journal of Contemporary Criminal Justice, 26, 148–166. Bottoms, A. E. (1999). Interpersonal violence and social order in prisons. Crime and Justice, 26, 205–281. Brezina, T. (1996). Adapting to strain: An examination of delinquent coping responses. Criminology, 34, 39–60. Brezina, T., Piquero, A. R., & Mazerolle, P. (1999). Student anger and aggressive behavior in school: An initial test of Agnew's macro-level strain theory. Journal of Research in Crime and Delinquency, 38, 362–386.
Broidy, L. M. (2001). A test of general strain theory. Criminology, 39, 9–35. Broidy, L. M., & Agnew, R. (1997). Gender and crime: A general strain theory perspective. Journal of Research in Crime and Delinquency, 34, 275–306. Bushway, S. D., Piquero, A. R., Broidy, L. M., Cauffman, E., & Mazerolle, P. (2001). An empirical framework for studying desistance as a process. Criminology, 39, 491–516. Camp, S. D., Gaes, G. G., Langan, N. P., & Saylor, W. G. (2003). The inﬂuence of prisons on inmate misconduct: A multilevel investigation. Justice Quarterly, 20, 501–533. Cao, L., Zhao, J., & Van Dine, S. (1997). Prison disciplinary tickets: A test of the deprivation and importation models. Journal of Criminal Justice, 25, 103–113. Capowich, G. E., Mazerolle, P., & Piquero, A. (2001). General strain theory, situational anger, and social networks: An assessment of conditioning inﬂuences. Journal of Criminal Justice, 29, 445–461. Clemmer, D. (1958). The prison community. Holt, Rinehart, & Winston, Inc. Cloward, R. A., & Ohlin, L. E. (1960). Delinquency and opportunity. New York: Free Press. Cohen, A. K. (1955). Delinquent boys. New York: Free Press. Cohen, M. A., Piquero, A. R., & Jennings, W. G. (2010). Studying the costs of crime across offender trajectories. Criminology and Public Policy, 9, 279–305. Cunningham, M. D., & Sorensen, J. R. (2006). Actuarial models for assessment of prison violence risk: Revisions and extensions of the risk assessment scale for prison (RASP). Assessment, 13, 253–265. Cunningham, M. D., & Sorensen, J. R. (2007). Predictive factors for violent misconduct in close custody. The Prison Journal, 87, 241–253. DeLisi, M., Berg, M. T., & Hochstetler, A. (2004). Gang members, career criminals and prison violence: Further speciﬁcation of the importation model of inmate behavior. Criminal Justice Studies, 17, 369–383. Eitle, D. J. (2002). Exploring a source of deviance-producing strain for females: Perceived discrimination and general strain theory. Journal of Criminal Justice, 30, 429–442. Eitle, D. J., & Turner, R. J. (2002). Exposure to community violence and young adult crime: The effects of witnessing violence, traumatic victimization, and other stressful live events. Journal of Research in Crime and Delinquency, 39, 214–237. Ellis, D., Grasmick, H., & Gilman, B. (1974). Violence in prisons: A sociological analysis. American Journal of Sociology, 80, 16–43. Feldman, B. J., Masyn, K. E., & Conger, R. D. (2009). New approaches to studying problem behaviors: A comparison of methods for modeling longitudinal, categorical adolescent drinking data. Developmental Psychology, 45, 652–676. Flanagan, T. J. (1980). Time served and institutional misconduct: Patterns of involvement in disciplinary infractions among long-term and short-term inmates. Journal of Criminal Justice, 8, 357–367. Gendreau, P., Goggin, C. E., & Law, M. A. (1997). Predicting prison misconducts. Criminal Justice and Behavior, 24, 414–431. Gover, A. R., MacKenzie, D. L., & Armstrong, G. S. (2000). Importation and deprivation explanations of juveniles’ adjustment to correctional facilities. International Journal of Offender Therapy and Comparative Criminology, 44, 450–467. Grifﬁn, M. L., & Hepburn, J. R. (2006). The effect of gang afﬁliation on violent misconduct among inmates during the early years of conﬁnement. Criminal Justice and Behavior, 33, 419–448. Hay, C. (2003). Family strain, gender, and delinquency. Sociological Perspectives, 46, 107–135. Hay, C., & Forrest, W. (2006). The development of self-control: examining self-control theory's stability thesis. Criminology, 44, 739–774. Henry, K., & Muthén, B. (2010). Multilevel latent class analysis: An application of adolescent smoking typologies with individual and contextual predictors. Structural Equation Modeling, 17, 193–215. Hochstetler, A., DeLisi, M., & Pratt, T. C. (2010). Social support and feelings of hostility among released inmates. Crime and Delinquency, 56, 588–607. Hoffmann, J., & Su, S. (1997). The conditional effects of stress on delinquency and drug use: A strain theory assessment of sex differences. Journal of Research in Crime and Delinquency, 34, 46–78. Irwin, J. K., & Cressey, D. R. (1962). Thieves, convicts and the inmate culture. Social Problems, 10, 142–155. Jang, S. J. (2007). Gender differences in strain, negative emotions, and coping behaviors: A general strain theory approach. Justice Quarterly, 24, 523–553. Jang, S. J., & Johnson, B. R. (2003). Strain, negative emotions and deviant coping among African Americans: A test of general strain theory. Journal of Quantitative Criminology, 19, 79–105. Jennings, W. G., Piquero, N. L., Gover, A. R., & Perez, D. M. (2009). Gender and general strain theory: A replication and extension of Broidy and Agnew's gender/strain hypothesis among a sample of southwestern Mexican American adolescents. Journal of Criminal Justice, 37, 404–417. Jiang, S., & Fisher-Giorlando, M. (2002). Inmate misconduct: A test of the deprivation, importation and situational models. The Prison Journal, 82, 335–358. Jiang, S., & Winfree, L. T. (2006). Social support, gender, and inmate adjustment to prison life: insights from a national sample. The Prison Journal, 86, 32–55. Johnson, R. (2001). Hard time: Understanding and reforming the prison. Belmont, CA: Wadsworth. Johnson, M. C., & Morris, R. G. (2008). The moderating effects of religiosity on the relationship between stressful life events and delinquent behavior. Journal of Criminal Justice, 36, 486–493. Land, K. C., McCall, P., & Nagin, D. S. (1996). A comparison of poisson, negative binomial, and semiparametric mixed poisson regression models with empirical applications to criminal careers data. Sociological Methods and Research, 24, 387–440. Laub, J. H., Nagin, D. S., & Sampson, R. J. (1998). Trajectories of change in criminal offending: Good marriages and the desistance process. American Sociological Review, 63, 225–238.
R.G. Morris et al. / Journal of Criminal Justice 40 (2012) 194–201 Lauritsen, J. L., Sampson, R. J., & Laub, J. H. (1991). The link between offending and victimization among adolescents. Criminology, 7, 91–108. Listwan, S. J., Sullivan, C. J., Agnew, R., Cullen, F. T., & Colvin, M. (2011). The pains of imprisonment revisited: The impact of strain on inmate recidivism. Justice Quarterly OnlineFirst published on July 21, 2011 as doi:10.1080/07418825.2011.597772 Mazerolle, P. (1998). Gender, general strain, and delinquency: An empirical examination. Justice Quarterly, 15, 65–92. Mazerolle, P., & Maahs, J. (2000). General strain and delinquency: An alternative examination of conditioning inﬂuences. Justice Quarterly, 17, 753–778. Mazerolle, P., & Piquero, A. R. (1998). Linking exposure to strain with anger: An investigation of deviant adaptations. Journal of Criminal Justice, 26, 195–211. Mazerolle, P., Piquero, A. R., & Capowich, G. (2003). Examining the links between strain, situational and dispositional anger, and crime: Further specifying and testing general strain theory. Youth and Society, 35, 131–157. Merton, R. K. (1938). Social structure and anomie. American Sociological Review, 3, 672–682. Morris, R. G., Longmire, D. R., Bufﬁngton-Vollum, J., & Vollum, S. (2010). Differential parole eligibility and institutional misconduct among capital inmates. Criminal Justice and Behavior, 37, 417–438. Morris, R. G., & Worrall, J. L. (2010). Prison architecture and institutional misconduct. Crime and Delinquency OnlineFirst, published on November 7, 2010 as doi:10. 1177/0011128710386204 Nagin, D. S. (2005). Group-based modeling of development. Cambridge, MA: Harvard University Press. Nagin, D. S., Farrington, D. P., & Mofﬁtt, T. E. (1995). Life-course trajectories of different types of offenders. Criminology, 33, 111–139. Nagin, D. S., & Land, K. C. (1993). Age, criminal careers, and population heterogeneity: Speciﬁcation and estimation of a nonparametric, mixed poisson model. Criminology, 31, 327–362. Nylund, K. L., & Masyn, K. E. (2008). Covariates and latent class analysis: Results of a simulation study. Paper presented at the Society for Prevention Research Annual Meeting. Paternoster, R., Brame, R., Mazerolle, P., & Piquero, A. R. (1998). Using the correct statistical test for the equality of regression coefﬁcients. Criminology, 36, 859–866. Paternoster, R., & Mazerolle, P. (1994). General strain theory and delinquency: A replication and extension. Journal of Research in Crime and Delinquency, 31, 235–263. Perez, D. M., Jennings, W. G., & Gover, A. R. (2008). Specifying general strain theory: An ethnically relevant approach. Deviant Behavior, 29, 544–578. Piquero, A. R. (2008). Taking stock of developmental trajectories of criminal activity over the life course. In A. Liberman (Ed.), The long view of crime: A synthesis of longitudinal research (pp. 23–78). New York, NY: Springer. Piquero, N. L., Fox, K., Piquero, A. R., Capowich, G., & Mazerolle, P. (2010). Gender, general strain theory, negative emotions and disordered eating. Journal of Youth and Adolescence, 39, 380–392.
Piquero, N. L., & Sealock, M. D. (2000). Generalizing general strain theory: An examination of an offending population. Justice Quarterly, 17, 449–484. Piquero, N. L., & Sealock, M. D. (2004). Exploring general strain: Gender and coping skills in an offender population. Justice Quarterly, 21, 125–158. Piquero, N. L., & Sealock, M. D. (2010). Race, crime and general strain theory. Youth Violence and Juvenile Justice, 8, 170–186. Raudenbush, S. W., & Bryk, A. S. (2002). Hierarchical linear models. Thousand Oaks, CA: Sage. Rivera, B., & Widom, C. S. (1990). Childhood victimization and violent offending. Violence and Victims, 5, 19–35. Sampson, R. J., & Laub, J. H. (2003). Life-course desisters? Trajectories of crime among delinquent boys followed to age 70. Criminology, 41, 555–592. Sorensen, J., & Pilgrim, R. L. (2000). An actuarial risk assessment of violence posed by capital murder defendants. Journal of Criminal Law and Criminology, 90, 1251–1270. Steinke, P. (1991). Using situational factors to predict types of prison violence. Journal of Offender Rehabilitation, 17, 119–132. Sykes, G. M. (1958). The society of captives: A study of a maximum security prison. Princeton, NJ: Princeton University Press. Tasca, M., Grifﬁn, M. L., & Rodriguez, N. (2010). The effect of importation and deprivation factors on violent misconduct: An examination of black and latino youth in prison. Youth Violence and Juvenile Justice, 8, 234–249. Toch, H. (1977). Living in prison: The ecology of survival. New York: Free Press. Toch, H., Adams, K., & Greene, R. (1987). Ethnicity, disruptiveness, and emotional disorder among prison inmates. Criminal Justice and Behavior, 14, 93–109. Vermunt, J. K. (2008). Latent class and ﬁnite mixture models for multilevel data sets. Statistical Methods in Medical Research, 17, 33–51. Williams, C. L., & Uchiyama, C. (1989). Assessment of life events during adolescence: The use of self-report inventories. Adolescence, 2, 95–118. Wooldredge, J. D. (1991). Correlates of deviant behavior among inmates of U.S. correctional facilities. Journal of Crime and Justice, 14, 1–25. Wooldredge, J. D. (2003). Keeping pace with evolving prison populations for effective management. Criminology and Public Policy, 2, 253–258. Wooldredge, J., Grifﬁn, T., & Pratt, T. (2001). Considering hierarchical models for research on inmate behavior: Predicting misconduct with multilevel data. Justice Quarterly, 18, 203–231. Worrall, J. L., & Morris, R. G. (2011). Taking inmate misconduct to the next level: The effect of custody classiﬁcation on prison rule violations. The Prison Journal, 91, 131–157. Zamble, E. (1992). Behavior and adaptation in long-term prison inmates: Descriptive longitudinal results. Criminal Justice and Behavior, 19, 409–425.