Macro-Correlates of Inmate Violence: The Importance of Programming in Prison Order

Abstract

Identifying the facility-level correlates of inmate misconduct is necessary for improving safety for both inmates and staff. Assessing these factors is also critical to ensure the effectiveness of administrative controls and facility programming. Using a repeat measures analysis, this study examines a panel of 487 state correctional facilities to assess the dynamic effects that factors of administrative control and treatment availability have on inmate–inmate and inmate–staff assaults. Findings show that several types of facility programs were negatively associated with inmate–inmate assaults but not inmate–staff assaults. Also, several dimensions of administrative control were associated with inmate–staff assaults. Implications for research and correctional practice are discussed.

Keywords

misconduct inmate prison treatment programming correctional staff

Introduction

Empirical research shows that mass incarceration in the United States has contributed to a myriad of correctional management issues. Overcrowding, increased costs of incarceration, constant need for scrutiny and order maintenance, correctional staff stress and turnover, and high rates of misconduct and violence have placed vast strain on corrections administration and management (Cullen, Link, Wolfe, & Frank, 1985; Stohr, Lovrich, Menke, & Zupan, 1994; Welsh, McGrain, Salamatin, & Zajac, 2007; Wooldredge, Griffin, & Pratt, 2001). Among all concerns for management, safety for inmates and correctional staff universally remains as the top priority (Cullen, Latessa, Burton, & Lombardo, 1993; Dilulio, 1987; Gendreau, Goggin, & Law, 1997; Steiner & Meade, 2014; Toch, Adams, & Grant, 1989). Inmate violence has a variety of adverse effects for inmates, correctional staff, and administrators alike. Each of these stakeholders in the prison environment possesses a need to feel safe and maintain some level of stability and order (Irwin & Cressey, 1962; Toch, 1997).

For inmates, acts of misconduct and violence by some increases the risk of injury to others. Such disorder impairs physical and emotional health, can lead to a loss of privileges, and, potentially, even increase the length of incarceration for those with serious disciplinary records (Bottoms, 1999; Gaes & McGuire, 1985; Toch, 1997). For correctional officers, inmate misconduct increases the threat of injury and stress, which, in turn, leads to higher rates of staff absenteeism, job dissatisfaction, and turnover (Cullen et al., 1993). From an administrative perspective, correctional disorder increases a variety of administrative costs, including security, detention of violators in disciplinary housing, placement of at-risk inmates in protective custody, medical treatment, and additional staff recruitment and training of staff (Goetting & Howsen, 1986; Langan & Pellsier, 2002; Welsh et al., 2007). Such social and fiscal costs may be mitigated if the sources of inmate misconduct were more adequately connected with or explained by administrative and institutional measures.

Typically, explanations of inmate misconduct focus on the individual as the primary unit of analysis (Blevins, Listwan, Cullen, & Jonson, 2010; Sorensen, Vigen, Woods, & Williams, 2015; Steinke, 1991. To date, the influence of facility-level factors on misconduct and violence is much less understood (Camp, Gaes, Langan, & Saylor, 2003; Gaes & McGuire, 1985; Lahm, 2008, 2009; McCorkle, Miethe, & Drass, 1995; Steiner, 2009; Steiner & Wooldredge, 2009b). Moreover, there is a growing number of studies evaluating the impact that programs, including drug treatment, educational, and vocational programs, have on the disciplinary records of individual inmates (Dietz, O’Connell, & Scarpitti, 2003; Welsh et al., 2007). However, fewer studies have examined the impact that the facility-level provision of programs have on aggregate rates of misconduct and violence within prisons. With inmate violence and serious misconduct creating a perpetual concern among correctional administrators (Catalano, 2004; Wolff, Shi, & Siegel, 2009; Wooldredge, 1998), improving our understanding of its correlates can guide the advancement of correctional policies. This knowledge can aid at improving prison administration in a variety of areas, such as inmate supervision, personnel management, resource allocation, and overall security.

Furthermore, most studies evaluating the correlates of inmate misconduct have employed cross-sectional research designs, thus focusing exclusively on the static influences of prison order (Camp et al., 2003; Gaes & McGuire, 1985; Lahm, 2008, 2009; McCorkle et al., 1995). Far fewer studies have used panel data analysis to test the factors that are predictive of changes in inmate violence over time (Steiner, 2009). To address these gaps in the literature and contribute our understanding of the facility-level predictors of inmate misconduct, and, more important, violence, this study uses a subsample of data drawn from the U.S. Bureau of Justice Statistics’ (BJS) 1995, 2000, and 2005 Census of State and Federal Adult Correctional Facilities.¹

Review of Literature

Deprivation Factors and Inmate Violence

Explanations of inmate misconduct typically encompass one of four theories, each having a slightly different unit of analysis: the importation model, the coping model, the situational model, and the deprivation model. Both the importation and situational models posit a micro-level account to misconduct that emphasizes the individual. Importation explanations hold that individual characteristics and inmate social histories before confinement are key predictors of inmate maladjustment and prison misconduct (Irwin & Cressey, 1962). The coping model suggests that the individual incarceree’s maladjustment once incarcerated is a product of the person’s inability to cope with the stressors of the new environment (Blevins et al., 2010; Toch, 1997).

Often in contrast with these two models are the situational and deprivation models that focus on the meso- (small group) and macro-level factors, respectively. Situational models highlight the importance of interactions between the inmates, staff, and the environment as the nature of the situation can dictate the inmate’s response (Endler & Magnusson, 1976; Goodstein, MacKenzie, & Shotland, 1984; Steinke, 1991). Last, supplying a more macroscopic view of misconduct, the deprivation model suggests that overall repressive and stressful environmental conditions within prisons are key predictors of inmate misconduct (Cao, Zhao, & Van Dine, 1997; McCorkle et al., 1995; Sykes, 1958). Although a number of studies have examined the influence that individual characteristics of inmates have on misconduct (e.g., Cao, Zhao, & Van Dine, 1997; Goetting & Howsen, 1986; Jiang & Fisher-Giorlando, 2002; Lahm, 2008, 2009), a limited number of studies have examined the deprivation model and the associated effects that facility-level factors (e.g., availability of various programming) have on inmate misconduct (Huebner, 2003; Kigerl & Hamilton, 2016; McCorkle et al., 1995; Steiner & Wooldredge, 2008; Useem & Reisig, 1999). The few studies that have examined macro-level factors of deprivation have produced inconclusive results. For instance, in their examination of 371 state prisons, McCorkle et al. (1995) found little to no support for a deprivation model of inmate violence, suggesting that problems with order are more an issue of managerial practice. Similarly, Gaes and McGuire (1985) also found only modest support for an aggregate deprivation theory in their study of 19 federal prisons.

There are a number of prison-level correlates of inmate misconduct that have been examined in previous studies, but few have consistently emerged as being significant (Steiner & Wooldredge, 2009a; Worrall & Morris, 2011). These correlates include prison crowding (Lahm, 2008; Steiner, 2009; Wooldredge et al., 2001), prison size (Huebner, 2003; Lahm, 2008; McCorkle et al., 1995), racial integration (Camp et al., 2003; Steiner, 2009), facility location (Huebner, 2003), security level (Huebner, 2003; Lahm, 2008; McCorkle et al., 1995; Steiner, 2009; Steiner & Wooldredge, 2008), staff characteristics (Camp et al., 2003), total proportion of inmates who participate in programs (Huebner, 2003; McCorkle et al., 1995), and facility age (Camp et al., 2003).

Effects of Programs on Inmate Misconduct and Violence

Among facility-level factors, there are some variables that are still in need of investigation. A number of publications have recommended that prison facilities should establish or improve existing treatment programs to reduce incidents of inmate misconduct (Gendreau & Keyes, 2001; Welsh et al., 2007); however, fewer studies have actually evaluated the existence of such an effect (Kupers et al., 2009; Welsh et al., 2007). The argument in favor of a positive effect of treatment on inmate misconduct, also known as the “treatment hypothesis,” posits that participation in programs may improve adjustment to adverse conditions, reducing incidents of misconduct (Hiller, Knight, Broome, & Simpson, 1998; Maguire, 1992; Siddall & Conway, 1988; Wright, 1991).

Only a few studies have been conducted that discuss facility-level treatment effects on misconduct. While not the aim of these studies, programming participation has been used as a control variable and was found to have some negative association with assault rates. For example, Huebner (2003) researched the impact of certain administrative controls on inmate assaults. Using hierarchical linear modeling on a sample of more than 4,000 male inmates across 185 state facilities, Huebner found that inmates who participated in work programs (particularly when the programs were used as incentives) were associated with significantly fewer assaults on staff.

In an effort to identify exactly how beneficial treatment programming can be for the reduction of inmate misconduct, French and Gendreau (2006) conducted a meta-analysis of studies that contained sufficient data to calculate an effect size coefficient that used experimental or quasi-experimental designs, and included in-facility misconduct outcomes. A total of 68 studies were identified as yielding 104 effect sizes, with study years ranging from 1952 to 2003. Findings from their analysis noted that behavioral treatments (40 studies) were associated with a 26% reduction in misconduct and nonbehavioral (e.g., nondirective counseling—31 studies) were associated with a 10% reduction, consistent with the results of two previous, similar meta-analyses (Gendreau & Keyes, 2001; Morgan & Flora, 2002). In addition, French and Gendreau (2006) make a clear call for future research to be geared toward primary studies of prison contexts rather than on meta-analyses (p. 210).

In another study, Steiner and Wooldredge (2008) investigated the differences in inmate and environmental effects on misconduct using data from the 1991 and 1997 surveys conducted by the U.S. Bureau of Census. The researchers applied bi-level modeling on a final-censored sample of 9,828 male inmates across 204 state facilities for 1991, and 10,022 male inmates across 203 facilities for 1997. Results showed that among the few facility-level factors found to be significant, programming was a significant predictor of assaults on either inmates or staff members in 1997 (measured as proportion of programming and work assignment participants). These mixed and inconclusive findings document the need for more research to investigate the effect that the availability and overall in-prison utility of programming has on inmate misconduct (Lahm, 2008; Steiner & Meade, 2014).

Facility-Level Correlates of Inmate-on-Staff Assaults

Addressing another gap, this study addresses the understudied topic of inmate–staff assaults. Most research on inmate–staff assault has focused on individual-level predictors, and fewer studies have examined environmental or facility-level factor effects on staff risk of being assaulted by inmates (Gaes & McGuire, 1985; Lahm, 2009; McCorkle et al., 1995; Sorensen et al., 2015). Gaes and McGuire’s (1985) study of 19 federal prisons used bi-level modeling to examine the effect that both individual-level and facility-level factors had on inmate–staff assaults. Few factors were significant, but the authors found that crowding was the strongest predictor of inmate–staff assaults. McCorkle, Miethe, and Drass (1996) later research used a sample of 371 facilities to test the effect that structural, managerial, and environmental determinants had on staff assaults. They found that the security level of the institution, program involvement, and the White–Black guard ratio were significant predictors of inmate–staff assaults. Lahm’s (2009) more recent study used a sample of 1,000 inmates from 30 prisons and employed a bi-level modeling strategy to investigate both individual-level and facility-level predictors of inmate-on-staff assaults. She found that prison facilities with a greater proportion of non-White inmates and larger staff-to-inmate ratios experienced higher rates of assaults on prison staff members.

The goals of this study are twofold. First, this study uses a panel data design to contribute to our understanding of the dynamic effects that factors of deprivation and treatment availability have on inmate violence. Second, this study seeks to add to our understanding on inmate–staff assault, which is an understudied topic in the inmate misconduct literature (Gaes & McGuire, 1985; Lahm, 2009; McCorkle et al., 1995). Implications for research and correctional management will be discussed.

Method

Description of Data

Data for this study were obtained from BJS’s 1995, 2000, and 2005 Census of State and Federal Adult Correctional Facilities. These data collections were administered by the U.S. Census Bureau on behalf of the Department of Justice’s BJS. The Census is conducted every 5 years and provides facility-level data on all state and federal correctional facilities in the United States. The Census provides a periodic snapshot of a variety of data points, such as each correctional facility’s scope of functions, custody levels, inmate capacity, inmate demographics, number of correctional officers and other staff, programming availability, and the incident rates of inmate misconduct and assaults.

The target population consists of state-level correctional facilities in the United States that had the primary function of housing a general population of male inmates. The 1995, 2000, and 2005 datasets were restricted to include only data pertaining to this target population. Additional steps were taken to merge these respective samples for panel data analysis. First, the 1995 sample of 1,500 facilities was restricted to include only state correctional facilities that had the primary function of housing a general population of male inmates. Both female and coed correctional facilities were excluded due to administrative, structural, and cultural differences between facilities that house women compared with men. The same procedures were followed for the 2000 and 2005 datasets. Next, the sample from the 1995 dataset was matched with samples from the 2000 and 2005 datasets using several linkage variables. The sample was restricted to facilities that reported in each of the 3 years of the Census. The final sample consisted of 521 facilities. An examination of the data prior to the data merger revealed a small number of facilities failed to report in all 3 Census years due to closure; these were omitted from the final sample. Important, in addition, a few facilities changed their purpose from housing a general population of men to either coed or female detention and, therefore, were also excluded, leaving a final subsample of 487 state prisons.

With the exception of the inmate–staff assault variable, 6.5% of total observations were missing from the subsample. The results of the missing data analysis indicated that these observations were missing at random (MAR) and should not bias parameter estimates (Acock, 2005). The inmate–staff assault variable contained a substantial number of missing observations (18.7%). As a means of compensating for this non-response error, and to establish a balanced panel, linear interpolation and linear extrapolation procedures were used (see Kovandzic, Sloan, & Vieraitis, 2002). Linear interpolation is a data imputation method that uses data from two or more different reference points to estimate missing data points between the references. Linear extrapolation is a method that uses data from preceding time points to estimate values for missing subsequent observations in a panel of data. The use of interpolation and extrapolation methods for the treatment of missing repeat measures data is well established in criminal justice and criminology (Kovandzic et al., 2002; Worrall & Kovandzic, 2007).

Measures

Dependent Variables

The outcome measure of interest, violent acts committed by inmates, was operationalized as two variables. The first measure assesses the total number of inmate-inflicted physical or sexual assaults on other inmates for each facility. The second measure includes the total number of inmate-inflicted physical or sexual assaults on correctional staff (i.e., includes all staff types, such as correctional officers, treatment and program staff, administrative staff, etc.). A common concern in measuring incidents of inmate misconduct is the possibility that the criteria used to classify such cases varies across states or administrative units, thereby making meaningful comparison problematic (Steiner, 2009). Self-reported rates of deviancy have also been criticized because they are susceptible to systematic errors, such as the underreporting or overreporting of a particular facility’s rates of misconduct (Hewitt, Poole, & Regoli, 1984). In response to these concerns, it is important to note that these reporting errors are foremost a problem with the measurement of rule infractions in general, whereas several studies have shown that incidents of violence, such as those examined in this study, are more likely to be reported accurately and are much less prone to manipulation and measurement error (Camp et al., 2003; Dilulio, 1987). Furthermore, Steiner (2009) argues that the potential for measurement error is not as much of an issue at the aggregate level because the underreporting of inmate misconduct by some staff should be balanced out by the overreporting of such incidents by others.

Independent Variables

Following the extant literature on inmate violence, a theoretically informed model was developed to evaluate the effects between facility-level factors and inmate violence. This framework included aggregate measures of officer and inmate social demographics, the criminal propensity of inmate populations, measures of formal control, and the availability of treatment programming within prison facilities. These factors and their sample-specific means are reported in Table 1.

Table 1.

Sample-Specific Descriptive Statistics for Dependent and Independent Measures (N = 487).

Measures	1995	2000	2005
Dependent variables
Inmate–inmate assaults^a	30.03 (59.27)	33.43 (62.24)	27.07 (64.58)
Inmate–staff assaults^a	15.50 (29.91)	18.57 (36.80)	14.14 (36.04)
Independent variables
Crowding	1.28 (0.37)	1.26 (0.43)	1.19 (0.40)
Proportion inmates African American	0.54 (0.29)	0.49 (0.19)	0.49 (0.32)
Proportion inmates Hispanic^a	0.12 (0.15)	0.11 (0.13)	0.15 (0.15)
Proportion non-White staff	0.25 (0.22)	0.29 (0.25)	0.26 (0.26)
Ratio correctional officers to inmates^a	0.26 (0.15)	0.24 (0.09)	0.30 (0.09)
Proportion maximum security inmates	0.21 (0.32)	0.19 (0.29)	0.22 (0.03)
Proportion minimum security inmates	0.42 (0.43)	0.40 (0.39)	0.44 (0.39)
Proportion inmates in disciplinary housing^a,b	0.03 (0.06)	0.03 (0.06)	0.08 (0.09)
Proportion inmates in protective custody	0.01 (0.03)	0.01 (0.03)	0.01 (0.02)
Proportion inmates with work assignment	0.71 (0.38)	0.65 (0.30)	0.69 (0.26)
Facility programs
Substance abuse	0.91 (0.29)	0.93 (0.15)	0.92 (0.17)
Mental health^b	0.70 (0.46)	0.79 (0.40)	0.84 (0.37)
Cognitive behavioral change^a,c	0.12 (0.33)	0.17 (0.38)	0.12 (0.33)
Sex offender^b,c	0.24 (0.19)	0.41 (0.49)	0.45 (0.50)
Family and life skills^b	0.41 (0.49)	0.38 (0.49)	0.35 (0.48)
Vocational	0.76 (0.43)	0.76 (0.43)	0.76 (0.43)
Educational^a,b	1.87 (0.88)	1.87 (0.85)	2.36 (0.73)
Employment^a,b	0.63 (0.48)	0.65 (0.48)	0.81 (0.40)
Types of work assignments
Prison industry	0.57 (0.49)	0.57 (0.50)	0.57 (0.50)
Farming and agriculture^b,c	0.31 (0.46)	0.39 (0.49)	0.40 (0.49)
Public works	0.64 (0.48)	0.67 (0.47)	0.59 (0.49)

Note. Standard deviations are listed in parentheses.

Statistically significant difference between 2000 and 2005.

Statistically significant difference between 1995 and 2005.

Statistically significant difference between 1995 and 2000.

p < .05.

Problems of prison overcrowding have become an increasingly salient issue in prison administration. Presumably, prison facilities that contain conditions of overcrowding are also likely to experience greater levels of inmate violence. Although several studies have evaluated the relationship between prison crowding and rates of inmate violence, empirical research in this area has yielded mixed results (Lahm, 2008; Steiner, 2012; Wooldredge et al., 2001). A measure of prison crowding was included in the current study, operationalized as the daily average population of each facility divided by its design capacity.

Prior research has shown that prisons that house a higher proportion of racial minority inmates experience higher rates of violence (Steiner, 2009). Also, correctional facilities that have a higher proportion of non-White staff have been hypothesized to have lower rates of violence (McCorkle et al., 1995). In their research, McCorkle and his colleagues suggested that a racially diverse staff should establish a more normalized prison environment for a diverse inmate population and thereby improve prison order. Given these observations, several variables were included to examine the relationship between racial composition and prison order. The racial diversity of inmate populations was assessed using two variables. One measures the proportion of African American inmates and the other evaluates the proportion of Hispanic inmates. A variable that assesses the overall proportion of non-White correctional officers was used to test the hypothesis that greater racial diversity in correctional officer staff is associated with lower rates of inmate violence.

Variables measuring the proportion of the inmate population in maximum security and the proportion of inmates in minimum security were used to control for the risk level or criminal propensity of inmate populations within each facility. Prison facilities with greater proportion of inmates in minimum security are hypothesized to experience lower rates of inmate violence, whereas facilities housing a greater proportion of maximum security inmates are expected to have higher rates of violence.

Several variables were included to assess the degree of formal control within prison facilities—officer-to-inmate ratios, disciplinary segregation, and work assignment. A ratio of correctional officers to inmates was used to test the hypothesis that a greater proportion of correctional officers to inmates are associated with lower levels of violence (Lahm, 2009; Steiner, 2009).

The proportion of inmates in disciplinary housing in each facility was assessed with the assumption that the use of disciplinary segregation will be associated with lower levels of violence. However, a variable measuring the proportion of inmates in protective custody was used with the expectation that increased use of protective segregation will be associated with greater levels of inmate violence. Prior research examining the effect that work assignments have on inmate misconduct and violence has shown that the remunerative benefits of being placed on work assignment lead to greater compliance and lower rates of overall violence (Colvin, 1992; Huebner, 2003; Steiner, 2009). To test this hypothesis, a variable measuring the proportion of inmates on work assignments was included.

The Census surveys address the availability of facility programming in three target areas: work assignments, counseling or special programming, and educational programs. These response items force the respondent to classify their available programs into a finite number of subgroupings that were provided on the surveys. As a result, many respondents felt that the programs offered by their facility were not encompassed by the items provided in the survey. For instance, while most potential work assignment and educational program options were covered by their respective subgroupings, a number of respondents, instead, provided responses into the “Other—Specify” category. This was particularly the case with the “counseling or special programming” category where common forms of treatment (e.g., cognitive behavioral therapy or reentry programs) did not have an item listed on the survey. In doing so, the data from this survey pertaining to programming were laden with many qualitative, typed responses that required content analysis to capture a more accurate range of programs offered. Thus, to achieve a more accurate measure of programming, we coded the written responses as they pertain to each of the targeted categories.

Upon coding this category’s responses in the first stage, we created new subgroups to better capture the scope of programs that were commonly specified: drug/alcohol focused, stress/anger management, generic cognitive behavioral therapy (CBT), brand name CBT (e.g., Moral Reconation Therapy), violent/assault and domestic violence focused, family focused, financial planning, employment focused, sex offender focused, and mental health-related programs. Each of the reported programs falling into these categories was given a score of 1. The next stage of the coding process involved the consolidation of data collected through content analysis with the programming options that were provided in the Census survey. This produced the following variables for this analysis: (a) substance abuse (drug/alcohol focused); (b) mental health (psychological and psychiatric services); (c) cognitive behavioral focus (includes both generic and brand name CBT, violent/assault, domestic violence, and stress/anger management programs); (d) sex offender counseling; (e) family and life skills (includes financial planning); (f) educational; (g) employment; (h) vocational; (i) prison industry; (j) public works; (k) farming and agricultural programs (see Table 1 for descriptive statistics).

Each of the treatment variables, except for educational programs, was dichotomously coded. Prisons that reported having at least one or more programs under each of the respective categories were coded as 1, and facilities that reported having no programs in each of the categories were coded as 0. Educational programs were combined as an ordinal scale ranging from 0 to 4, including the following subgroupings of educational items: common education (Literacy, General Educational Development [GED], English as Second Language [ESL]); collegiate education; special education; and study-release education. Facility self-reports of offering transitional/reentry programs were omitted from the final model because only 3% of prisons reported having such programs.

Statistical Modeling

Negative binomial regression analysis was used to examine the effect that facility-level factors had on inmate–inmate and inmate–staff assaults. This technique was a preferable alternative to Poisson regression due to the presence of overdispersion in the distribution of the outcomes variables. Negative binomial is an alternative that generalizes the Poisson regression model by adding a dispersion parameter (see Raudenbush & Bryk, 2002). Diagnostic statistics indicated overdispersion in the distributions for both inmate–inmate and inmate–staff assault rates. Overdispersion in the distributions for the inmate–inmate assault variable was suggested by variances that are greater than their respective means (1995: x- = 30.03, σ² = 3512.9; 2000: x- = 33.43, σ² = 3873.8; 2005: x- = 27.07, σ² = 4170.6). This was also the case with the distributions for the inmate–staff assault variable (1995: x- = 15.50, σ² = 894.6; 2000: x- = 18.57, σ² = 1354.24; 2005: x- = 14.14, σ² = 1298.9).²

Pooled cross-sectional time-series (or panel data) analysis was used to examine the repeated effects that facility-level factors had on inmate–inmate and inmate–staff assaults. For this analysis, we specified a two-way random effects negative binomial (i.e., overdispersion) model.³ This specification yields the following:

\Pr {(Y}_{i t} = y_{i t} | x_{i t}, δ_{i}) = \frac{Γ (λ_{i t} + y_{i t})}{Γ (λ_{i t}) Γ (y_{i t} + 1)} {(\frac{1}{1 + δ_{i}})}^{λ_{i t}} {(\frac{δ_{i}}{1 + δ_{i}})}^{λ_{i t}}

where $y_{i t}$ is the dependent variable for facility i at year t. Moreover, $x_{i t}$ represents a set of explanatory variables, and $δ_{i}$ is the dispersion parameter. For this two-way random effects model, $δ_{i}$ is allowed to vary across groups (see Hausman, Hall, & Griliches, 1984, p. 927). Poisson and negative binomial (aka, generalized Poisson) regression have been widely used in the social sciences and criminal justice for the analysis of count data, and these techniques are widely accepted (Camp et al., 2003; Gaes & McGuire, 1985; Lahm, 2008, 2009; McCorkle et al., 1995; Steiner, 2009).

Findings and Discussion

Table 2 reports the results from the random effects negative binomial models (panel data) that estimated the effect between facility-level factors and both inmate–inmate and inmate–staff assaults.⁴ These models yield a number of noteworthy findings.

Table 2.

Panel Data Analysis: Facility-Level Predictors of Inmate–Inmate and Inmate–Staff Assaults (N = 1,461).

Crowding	Inmate–inmate assaults	Inmate–staff assaults
	−0.00 (0.00)	0.00 (0.00)
Proportion inmates African American	0.40 (0.16)*	0.09 (0.16)
Proportion inmates Hispanic	0.79 (0.16)	0.37 (0.20)
Proportion non-White staff	0.16 (0.13)	−0.08 (0.13)
Ratio correctional officers to inmates	0.11 (0.10)	0.30 (0.11)**
Proportion maximum security inmates	0.02 (0.05)	0.03 (0.05)
Proportion minimum security inmates	−0.01 (0.03)	−0.01 (0.03)
Proportion inmates in disciplinary housing	−0.13 (0.27)	0.96 (0.27)***
Proportion inmates in protective custody	2.76 (0.80)***	3.01 (0.77)***
Proportion inmates with work assignment	−0.01 (0.03)	−0.13 (0.04)***
Facility programs
Substance abuse	−0.30 (0.10)**	−0.10 (0.09)
Mental health	0.22 (0.08)	0.39 (0.08)***
Cognitive behavioral change	0.13 (0.08)	0.01 (0.08)
Sex offender	−0.13 (0.06)*	−0.05 (0.06)
Family and life skills	−0.14 (0.06)*	−0.12 (0.06)
Vocational	0.23 (0.09)*	−0.00 (0.08)
Educational	−0.04 (0.04)	0.12 (0.04)**
Employment	0.16 (0.07)*	0.22 (0.07)***
Types of work assignments
Prison industry	0.33 (0.07)***	0.21 (0.07)***
Farming and agriculture	0.09 (0.06)	−0.12 (0.06)*
Public works	−0.24 (0.06)***	−0.10 (0.06)
Intercept	−0.59	−1.28
Log likelihood	−5,777.9	−5,138.7
Wald χ²	156.05 (21), p < .001	155.82 (21), p < .001

Note. A total of 487 groups were estimated in these random effects negative binomial models. Models report unstandardized coefficients and standard errors in parentheses. STATA statistical software does not report a pseudo R² coefficient for these generalized models.

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

Racial Integration, Formal Control, and Inmate Assaults

First, higher proportions of African American inmates predicted higher rates of inmate assaults over the panel years 1995 to 2005. Prior research had also identified this dynamic relationship between the proportion of African American inmates and rates of inmate–inmate violence. For instance, Steiner (2009) used multilevel models and a longitudinal design to predict the effect of facility-level factors on inmate violence between 1995 and 2000, and found a similar result. This finding was expected, given that racial integration within a prison’s general population is often accompanied by social segregation, racial tension, conflict (Steiner, 2009), and gang rivalry (Drury & DeLisi, 2011). These conditions undoubtedly cause a great deal of stress for many inmates and, thus, exacerbate the risk for assaults. The association between racial heterogeneity and deviancy has been fairly well established in other social contexts (e.g., Sampson & Groves, 1989; Sampson, Raudenbush, & Earls, 1997).

Next, findings from Table 2 show that dimensions of formal control were significantly associated with inmate violence. Interestingly, higher ratios of correctional officers to inmates were a significant predictor of higher rates of inmate–staff assaults. Lahm’s (2009) study similarly found that prison facilities with larger staff-to-inmate ratios experience higher rates of assault on prison staff. There are several possible explanations for this finding. First, administrators of correctional facilities may have chosen to hire and staff more correctional officers in direct response to security concerns. Second, perhaps this outcome lends itself to a routine activities type of explanation. It is possible that having a higher ratio of correctional officers to inmates had increased the number of interactions between correctional officers and inmates, thus creating more opportunities for inmate–staff assaults to occur.

As expected, other factors of formal control were also associated with inmate violence. First, the proportion of inmates in disciplinary housing was a significant predictor of inmate–staff assaults. From the perspective of prison administrators who follow control-oriented management styles, it has long been assumed that the use of disciplinary segregation has the aggregate effect of reducing problem behavior in the general population and, thus, improves prison order by incapacitating the most violent offenders and deterring other inmates from misconduct. This relationship suggests that the practice of disciplinary segregation has the contrary effect and may even exacerbate the violent propensity of offenders. Similarly, a growing body of research shows that the practice of administrative segregation is frequently used as a substitute for treating mental illnesses and behavioral disorders. Studies illustrate that in these situations, the use of segregation is, in fact, associated with increased violence (Briggs, Sundt, & Castellano, 2003; Kupers et al., 2009; O’Keefe, 2008; Toch & Kupers, 2007).

Next, the proportion of inmates in protective custody was also a significant predictor of both increased inmate–inmate and inmate–staff assaults. This finding is contrary to the assumption that placing inmates who are at higher risk of victimization in protective custody may improve prison order. A possible explanation for this finding is that the underlying conditions of these facilities, such as insufficient controls or a relatively more criminogenic inmate population, were actually leading to this increase in inmate assault rates. Specifically, the use of protective custody for a few inmates (average of 1%) simply failed to have the effect of decreasing aggregate rates of assault victimization. However, a greater proportion of protective custody use may be used as a response to increased violence among the inmate general population. Such a case may be an indication that prison managers perceived a heightened level of violence and aimed to neutralize the threat to certain inmates.

Programming Availability and Inmate–Inmate Assaults

Results from Table 2 indicate that the general availability as well as type of programming potentially played a role in reducing inmate violence toward other inmates, and, thus, lending some support for the treatment hypothesis. As shown in Table 2, 50% of program types modeled indicated an association with decreased inmate–inmate violence: substance abuse, sex offender, family and life skills, and educational programming (although education was not significant). Similarly, the availability of public works programming was also associated with a significant decrease among inmate–inmate assaults. Interestingly, other programs, however, demonstrated an association with significant increases of inmate–inmate assaults: vocational, employment, and prison industry programs.

There are two important contextual caveats in which these results should be interpreted. First, given the changes of correctional management and movement toward more evidence-based rehabilitation, it is only logical to expect that programming would become more effective at reducing institutional misconduct as it does reducing recidivism. The more honed programming becomes and the more willing institutions are to adopt such honed programming, the more we can expect to see a decrease in violence when programs (particularly specialized treatment programs) are made available. Second is the fact that virtually every correctional program, with exception given to institutional work assignments, is designed to decrease the offenders’ likelihood of recidivism upon release. They are not designed to necessarily decrease misconduct while still incarcerated. This means that, typically, only those close to release will be receiving the needed programming that helps with reentry. It also means that any reduction in misconduct is an added bonus, as misconduct is an intermediate outcome for institutional programming. In the same token, any program’s associated increases with assaults must not be viewed as “ineffective.” Instead, such programs are primed for future research to include misconducts as an intermediate outcome measure for determining program effectiveness.

Programming Availability and Inmate–Staff Assaults

In contrast with the inmate–inmate assault model, findings from the model that estimates inmate–staff assaults yield a rather different picture. As shown in Table 2, the panel analysis results show that only one program was a significant predictor of decreased inmate–staff assaults (farming and agriculture), whereas four were associated with increases (mental health, educational, employment, and prison industry). In light of these findings regarding inmate–staff assaults, it appears that there is little support for the treatment hypothesis. The availability of programs does not appear to reduce inmate assaults on staff. Steiner and Wooldredge (2008) similarly found that increased levels of inmate participation in programs were associated with inmate misconduct. The authors concluded that,

the measure of program participation did not reflect reduced opportunities for misconduct as much as it reflected an inmate’s need for counseling, and those who need counseling and education might be at higher risk for violating facility rules, at least in the early phases of their incarceration. (p. 452)

The current study lends additional support for this interpretation.

Regardless of the intended purpose of these programs, it is rather peculiar that the availability of some programs was significantly associated with increases in assaults for both inmate–inmate and inmate–staff. One would expect that if programming is designed to reduce problematic or criminal behavior in the society, it should also have similar effects with those still in the prison. However, such an expectation assumes that just because a program is available, it is (a) being used by the majority of the inmates, and (b) is achieving expected fidelity with what has been shown to be evidence-based, effective programming (see Lipsey and Cullen, 2007).

Perhaps even more peculiar is that there are distinct differences in which programs predict assault occurrences, depending on who is the victim of the assault. While inmate–staff assaults do occur far less often than inmate–inmate assaults, these findings appear to beg further questions about the nature of the programs that are associated with increased inmate–staff assaults. In this sense, however, such findings do appear to lend support to a routine activities type of explanation. Each of those programs that predicts increased assaults on staff tend to be characterized by more interactions with either a higher risk population (mental health programs) or with a greater number of specialized staff (educational, employment, prison industry programs) who may be less equipped to deal with inmates than correctional officers.

Conclusion

Limitations

Before discussing the implications that these findings have for research and correctional policy, it is important to acknowledge the limitations of this study. First, the BJS’s 1995, 2000, and 2005 Census of State and Federal Adult Correctional Facilities were primarily designed for administrative purposes, and, therefore, these data sources are not optimal for addressing the theoretical questions that are addressed here. This issue is common with the analysis of administrative data. This study uses a number of dichotomous variables as measures of facility-level treatment that may result in some degree of measurement error. The Census’ categorical variables lack the measurement precision that interval-based measures could alternatively provide. For instance, the Census’ section for “counseling or special programming” uses dichotomous variables to measure treatment availability, but it does not ask facilities to report the numbers of offenders who participated in these programs, and it does not provide any indication of the quality of treatment that was provided.

Second, it is important to note that, on average, a very small proportion of offenders were placed in disciplinary segregation or protective custody, in which case, the significant effects for these variables should be interpreted with caution. With low base rates, it is questionable whether these measures of formal control have any substantial relationship with inmate violence (Steiner, 2009). Third, it is important to acknowledge that many of the violent violations may have occurred in response to issues that were not accounted for in these data, such as shifts in personnel and management styles, internal policies that dictate inmate movement within the facility, and informal personal tensions that play out between inmates and staff. One particularly important theoretical variable not accounted for in these data is the effect of inmate transfers. Transfer theory suggests that there may be a bit of both importation and deprivation elements that construct the events of inmate violence (Kigerl & Hamilton, 2016). Specifically, as inmates are transferred between facilities, so, too, are violent tendencies and cultures of a separate facility. In such instances, the deprivation of one facility may instill a higher likelihood of violence within the inmates leaving that facility. Such factors are also not accounted for in this dataset.

Despite these limitations, the BJS’s Census of State and Federal Correctional Facilities is a nationally representative data source that offers several advantages, including high response rates, breadth of pertinent facility-level data points, and repeat measures. It is with these advantages in mind, coupled with our content analysis of programming availability and robust statistical approach, that we contend the findings are meaningful to the field and practice of prison management.

Implications

This study provides some support for deprivation theory. Deprivation theory posits that inmate misconduct and violence is a product of stressful and repressive conditions within a correctional facility (Gaes & McGuire, 1985; McCorkle et al., 1995 Sykes). Several factors related to deprivation were shown to be significant, including the use of disciplinary housing and racial integration. Building on previous study evidence (Cao et al., 1997; Goetting & Howsen, 1986; Jiang & Fisher-Giorlando, 2002; Lahm, 2008, 2009), we suggest that correctional administrators can potentially reduce inmate violence with policies and practices that mitigate the effects that deprivation has on inmates. Although, at the facility level, administrators have limited control over the rate of inmate admissions and the respective levels of risk that they import, it is feasible for them to take remedial steps to mitigate the effect that stressful conditions, such as crowding or segregation, have on inmates (e.g., use of alternative sanctions for low-level infractions, group therapy, work assignments, and programs aimed at building social bonds with family and community). Consistent with previous work (Bottoms, 1999; Camp et al., 2003; Steiner, 2009), our results confirm that the degree of racial integration within correctional facilities is a key factor in explaining variations in prison order. This evidence suggests that there is potential for corrections officials to reduce violence with efforts to improve interracial relations within facilities. Systematic assessments of interracial dynamics and conflicts within inmate population(s) is an important first step toward improving prison order. Research examining methods for reducing racial conflict and improving interracial relations within prisons is direly needed.

This research poses challenges for advocates of administrative control theory (Dilulio, 1987). Several dimensions of formal control, including higher rates of correctional officers to inmates, greater proportions of inmates in disciplinary segregation, and protective custody, did not lead to lower rates of assaults. Rather, they were predictive of higher rates of inmate violence. This implies that improving prison order is not as simple as increasing formal controls within prisons. Craig (2004) similarly observed a number of problems with control-oriented models for prison administration. The effects of formal control should be examined more closely by correctional administrators as well as researchers. Evidence showing that the hiring of correctional officers was associated with higher rates of violence suggests that facility administrators are simply hiring and staffing more correctional officers in direct response to prison disorder and violence (see Lahm, 2009). If this interpretation is correct, correctional managers should take a closer look at the shortcomings of this reactive strategy to maintaining prison order. This illuminates the need for a more proactive and problem-oriented approach to maintaining prison order, one that involves the systematic evaluation of the fundamental causes of prison violence.

Next, consistent with the work of Colvin (1992), Huebner (2003), and Steiner (2009), the use of work assignments was predictive of lower rates of inmate–staff assaults over the panel years. This indicates that practices associated with increasing the provision of work assignments, including providing inmates with a structured routine, greater supervision, and coercive controls, coupled with the remunerative benefits that inmates receive from being assigned a work detail, lead to lower rates of inmate–staff assaults. It is important to highlight that some consider work assignments as a form of treatment, thus having a therapeutic or rehabilitative effect. In their efforts to improve prison order, corrections administrators should take note that increasing work assignments may decrease violence.

In addition, important findings from our modeling of inmate–inmate assaults provides some support for the treatment hypothesis, whereas, in contrast, our model evaluating inmate–staff assaults illustrates that, at the aggregate level, the increased availability of programming was actually associated with higher levels of inmate–staff assaults. While inmate–staff assaults do occur far less often than inmate–inmate assaults, this raises important questions concerning the nature of these programs that are associated with increased inmate–staff assaults, the quality in implementation and administration of these programs, safety protocols, and training for correctional staff. Certainly, more research is needed that evaluates both the effect of programming on inmate violence and the predictors of inmate–staff assaults. Overall, these findings lend support for a routine activities type of explanation for incidents of inmate–staff assaults. Each of the programs that were associated with increased assaults on staff were characterized with either increased interactions between staff and an unstable population (mental health programs) or more interactions between specialized staff (i.e., educational, employment, prison industry programs) who, presumably, were less equipped to deal with inmates than correctional officers.

Footnotes

Declaration of Conflicting Interests

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

Funding

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

Notes

Author Biographies

Blake M. Randol, PhD, is an assistant professor of criminal justice, California State University, Stanislaus. His research interests focus on the application of organizational theory to the study of police and correctional agencies.

Christopher M. Campbell, PhD, is an assistant professor of criminology and criminal justice, Portland State University. His research interests focus on penology, offender rehabilitation and reentry, juvenile justice, and criminological integration.

References

Acock

(2005). Working with missing values. Journal of Marriage and Family, 67, 1012-1028.

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

(1999). Interpersonal violence and social order in prisons. In Tonry

Petersilia

(Eds.), Crime and justice: A review of research (Vol. 26, pp. 205-281). Chicago, IL: University of Chicago Press.

Briggs

C. S.

Sundt

J. L.

Castellano

T. C.

(2003). The effect of supermaximum security prisons on aggregate levels of institutional violence. Criminology, 41, 1341-1376.

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

Zhao

Van Dine

(1997). Prison disciplinary tickets: A test of the deprivation and importation models. Journal of Criminal Justice, 25, 103-113.

Catalano

S. M.

(2004). Criminal victimization, 2003 (Bureau of Justice Statistics National Crime Victimization Survey, NCJ 210674). Washington, DC: U.S. Department of Justice, Office of Justice Programs.

Colvin

(1992). The penitentiary in crisis. Albany: State University of New York Press.

Craig

S. C.

(2004). Rehabilitation versus control: An organizational theory of prison management. The Prison Journal, 84, 92-114.

10.

Cullen

F. T.

Latessa

E. J.

Burton

V. S.

Lombardo

L. X.

(1993). Correctional orientation of prison wardens: Is the rehabilitation ideal supported. Criminology, 31, 69-92.

11.

Cullen

F. T.

Link

B. G.

Wolfe

N. T.

Frank

(1985). The social dimensions of correctional officer stress. Justice Quarterly, 2, 505-533.

12.

Dietz

E. F.

O’Connell

D. J.

Scarpitti

F. R.

(2003). Therapeutic communities and prison management: An examination of the effects of operating an in-prison therapeutic community on levels of institutional disorder. International Journal of Offender Therapy and Comparative Criminology, 47, 210-223.

13.

Dilulio

J. J.

(1987). Governing prisons. New York, NY: Free Press.

14.

Drury

A. J.

DeLisi

(2011). Gangkill: An exploratory empirical assessment of gang membership, homicide offending, and prison misconduct. Crime & Delinquency, 57, 130-146.

15.

Endler

N. S.

Magnusson

(1976). Toward an interactional psychology of personality. Psychological Bulletin, 83, 956-974.

16.

Fisher

J. E.

Mason

R. L.

(1981). The analysis of multicollinear data in criminology. In Fox

J. A.

(Ed.), Criminology, 99-125.

17.

French

S. A.

Gendreau

(2006). Reducing prison misconducts What works! Criminal Justice and Behavior, 33, 185-218.

18.

Gaes

G. G.

McGuire

W. J.

(1985). Prison violence: The contribution of crowding versus other determinants of prison assault rates. Journal of Research in Crime & Delinquency, 22, 41-65.

19.

Gendreau

Goggin

C. E.

Law

M. A.

(1997). Predicting prison misconducts. Criminal Justice and Behavior, 24, 414-431.

20.

Gendreau

Keyes

(2001). Making prisons safer and more humane environments. Canadian Journal of Criminology, 43, 123-130.

21.

Goetting

Howsen

R. M.

(1986). Correlates of prisoner misconduct. Journal of Quantitative Criminology, 2, 49-67.

22.

Goodstein

MacKenzie

D. L.

Shotland

R. L.

(1984). Personal control and inmate adjustment to prison. Criminology, 22, 343-369.

23.

Hausman

J. A.

Hall

B. H.

Griliches

(1984). Econometric models for count data with an application to the Patents-R&D relationship Econometrica, 52, 909-938.

24.

Hewitt

J. D.

Poole

E. D.

Regoli

R. M.

(1984). Self-reported and observed rule-breaking in prison: A look at disciplinary response. Justice Quarterly, 1, 437-447.

25.

Hiller

M. L.

Knight

Broome

K. M.

Simpson

D. D.

(1998). Legal pressure and treatment retention in a national sample of long-term residential programs. Criminal Justice and Behavior, 25, 463-481.

26.

Huebner

(2003). Administrative determinants of inmate violence: A multi-level analysis. Journal of Criminal Justice, 31, 107-117.

27.

Irwin

J. K.

Cressey

(1962). Thieves, convicts, and the inmate culture. Social Problems, 10, 142-155.

28.

Jiang

Fisher-Giorlando

(2002). Inmate misconduct: A test of deprivation, importation, and situational models. The Prison Journal, 83, 335-358.

29.

Kigerl

Hamilton

Z. K.

(2016). The impact of transfers between prisons on inmate misconduct: Testing importation, deprivation, and transfer theory models. The Prison Journal, 96, 232-257.

30.

Kovandzic

Sloan

Vieraitis

(2002). Unintended consequences of politically popular sentencing policy: The homicide promoting effects of “three strikes” in U.S. cities (1980-1990). Criminology & Public Policy, 1, 399-424.

31.

Kupers

T. A.

Dronet

Winter

Austin

Kelly

Cartier

. . . McBride

(2009). Beyond Supermax administrative segregation: Mississippi’s experience rethinking prison classification and creating alternative mental health programs. Criminal Justice and Behavior, 36, 1037-1050.

32.

Lahm

K. F.

(2008). Inmate-inmate assault: A multi-level examination of prison violence. Criminal Justice and Behavior, 35, 120-137.

33.

Lahm

K. F.

(2009). Inmate assaults on prison staff: A multilevel examination of an overlooked form of prison violence. The Prison Journal, 89, 131-150.

34.

Langan

N. P.

Pellsier

B. M.

(2002). The effect of drug treatment of inmate misconduct in federal prisons. Washington, DC: Federal Bureau of Prisons.

35.

Lipsey

M. W.

Cullen

F. T.

(2007). The effectiveness of correctional rehabilitation: A review of systematic reviews. Annual Review of Law and Social Science, 3, 297-320.

36.

Maguire

(1992). Prison industry: The effect of participation on inmate disciplinary adjustment. Dissertation Abstracts International, 53, 312.

37.

McCorkle

R. C.

Miethe

T. D.

Drass

K. A.

(1995). The roots of prison violence: A test of the deprivation, management, and “not-so-total” institutional models. Crime & Delinquency, 41, 317-331.

38.

Morgan

R. D.

Flora

D. B.

(2002). Group psychotherapy with incarcerated offenders: A research synthesis. Group Dynamics: Theory, Research, and Practice, 6, 203-218.

39.

O’Keefe

M. L.

(2008). Administrative segregation from within a corrections perspective. The Prison Journal, 88, 123-143.

40.

Raudenbush

S. W.

Bryk

A. S.

(2002). Hierarchical linear models: Applications and data analysis methods (Vol. 1). Thousand, Oaks, CA: Sage.

41.

Sampson

R. J.

Groves

W. B.

(1989). Community structure and crime: Testing social-disorganization theory. American Journal of Sociology, 94, 774-802.

42.

Sampson

R. J.

Raudenbush

S. W.

Earls

(1997). Neighborhoods and violent crime: A multilevel study of collective efficacy. Science, 277, 918-924.

43.

Siddall

J. W.

Conway

G. L.

(1988). International variables associated with retention and success in residential drug treatment. International Journal of the Addictions, 23, 1241-1254.

44.

Sorensen

J. R.

Vigen

M. P.

Woods

S. O.

Williams

B. D.

(2015). Assaults on inmates and staff by perpetrators of intimate partner homicide: An examination of competing hypotheses. Journal of Interpersonal Violence, 30, 3367-3388.

45.

STATA. (2015). Fixed-effects, random-effects, & population-averaged negative binomial models. Retrieved from http://www.stata.com/manuals13/xtxtnbreg.pdf

46.

Steiner

(2009). Assessing static and dynamic influences on inmate violence levels. Crime & Delinquency, 55, 134-161.

47.

Steiner

Meade

(2014). The safe prison. In Cullen

F. T.

Jonson

C. L.

Stohr

M. K.

(Eds.), The American prison: Imagining a different future (pp. 129-150). Thousand Oaks, CA: Sage.

48.

Steiner

Wooldredge

(2008). Inmate versus environmental effects on prison rule violations. Criminal Justice and Behavior, 35, 438-456.

49.

Steiner

Wooldredge

(2009a). Individual and environmental effects on assaults and nonviolent rule breaking by women in prison. Journal of Research in Crime & Delinquency, 46, 437-467.

50.

Steiner

Wooldredge

(2009b). Rethinking the link between institutional misconduct and crowding. The Prison Journal, 89, 205-233.

51.

Steinke

(1991). Using situational factors to predict types of prison violence. Journal of Offender Rehabilitation, 17, 119-132.

52.

Stohr

M. K.

Lovrich

N. P

Jr. Menke

B. A.

Zupan

L. L.

(1994). Staff management in correctional institutions: Comparing Dilulio’s “control model” and “employee investment model” outcomes in five jails. Justice Quarterly, 11, 471-497.

53.

Sykes

(1958). The society of captives. Princeton, NJ: Princeton University Press.

54.

Toch

(1997). Corrections: A humanistic approach. Guilderland, NY: Harrow & Heston.

55.

Toch

Adams

Grant

J. D.

(1989). Coping: Maladaptation in prisons. New Brunswick, NJ: Transaction.

56.

Toch

Kupers

(2007). Violence in prison: Revisited. Journal of Offender Rehabilitation, 45, 1-28.

57.

Useem

Reisig

M. D.

(1999). Collective action in prisons: Protests, disturbances, and riots. Criminology, 37, 735-760.

58.

Welsh

McGrain

Salamatin

Zajac

(2007). Effects of drug treatment on inmate misconduct: A repeated measures analysis. Criminal Justice and Behavior, 34, 600-615.

59.

Wolff

Shi

Siegel

(2009). Understanding physical victimization inside prisons: Factors that predict risk. Justice Quarterly, 26, 445-475.

60.

Wooldredge

J. D.

(1998). Inmate lifestyles and opportunities for victimization. Journal of Research in Crime & Delinquency, 35, 480-502.

61.

Wooldredge

J. D.

Griffin

Pratt

(2001). Considering hierarchical models for research on inmate behavior: Predicting misconduct with multi-level data. Justice Quarterly, 8, 217-242.

62.

Worrall

J. L.

Kovandzic

T. V.

(2007). COPS grants and crime revisited. Criminology, 45, 159-190.

63.

Worrall

J. L.

Morris

R. G.

(2011). Inmate custody levels and prison rule violations. The Prison Journal, 91, 131-157.

64.

Wright

K. N.

(1991). The violent and victimized in the male prison. Journal of Offender Rehabilitation, 16, 1-25.