The Use of Restrictive Housing on Gang and Non-Gang Affiliated Inmates in U.S. Prisons: Findings from a National Survey of Correctional Agencies

Abstract

Gangs present serious challenges to the management and order of prisons. Restrictive housing is viewed by correctional officials as one of the few effective responses to gangs, yet public officials and advocates continue to push for reductions in its use. Some evidence suggests gang affiliates are overrepresented in restrictive housing, although this research is limited to a few prison systems, and the reasons for this relationship remain speculative at best. This study examined the relationship between restrictive housing and gang affiliation based on administrative data gathered from 39 correctional agencies in 2016, collectively housing 73% of state prisoners. The relative risk of placement in restrictive housing was three times greater for gang affiliated than non-gang inmates. Over one-third of the inmates in restrictive housing were gang affiliated. While gang affiliates were overrepresented across all of the primary pathways into restrictive housing, the risk was greater for administrative purposes (6.3) than for disciplinary (3.1) or protective (2.6) purposes, although substantial variation existed across prison systems. The quest to reduce the footprint of restrictive housing in U.S. prisons cannot occur without accounting for one of the most difficult populations for correctional authorities to manage, that is, the 213,000 prisoners affiliated with gangs. The challenge will be greater in prison systems with larger custodial populations, a higher proportion of confirmed gang affiliates, higher rates of gang-related violence, and longstanding histories of gangs, where restrictive housing is more likely to be used disproportionately. Programs and practices aimed to reintegrate gang affiliates back into general population housing are deserving of research and evaluation owing to their implications for institutional and community corrections.

Full text

THE USE OF RESTRICTIVE HOUSING ON GANG AND NON-GANG AFFILIATED
INMATES IN U.S. PRISONS: FINDINGS FROM A NATIONAL SURVEY OF
CORRECTIONAL AGENCIES
DAVID C. PYROOZ
Department of Sociology
University of Colorado Boulder
MEGHAN M. MITCHELL
Department of Criminal Justice
University of Central Florida
forthcoming in Justice Quarterly
https://doi.org/10.1080/07418825.2019.1574019
This is the authors’ pre-print copy of the article. Please download and cite the post-print copy
published on Justice Quarterly’s website
Corresponding Author: David C. Pyrooz, Department of Sociology, University of Colorado
Boulder, Institute of Behavioral Science, 483 UCB, Boulder, CO 80309-0483 (email:
David.Pyrooz@colorado.edu)

Abstract
Gangs present serious challenges to the management and order of prisons. Restrictive housing is
viewed by correctional officials as one of the few effective responses to gangs, yet public
officials and advocates continue to push for reductions in its use. Some evidence suggests gang
affiliates are overrepresented in restrictive housing, although this research is limited to a few
prison systems, and the reasons for this relationship remain speculative at best. This study
examined the relationship between restrictive housing and gang affiliation based on
administrative data gathered from 39 correctional agencies in 2016, collectively housing 73% of
state prisoners. The relative risk of placement in restrictive housing was 3 times greater for gang
affiliated than non-gang inmates. Over one-third of the inmates in restrictive housing were gang
affiliated. While gang affiliates were overrepresented across all of the primary pathways into
restrictive housing, the risk was greater for administrative purposes (6.3) than for disciplinary
(3.1) or protective (2.6) purposes, although substantial variation existed across prison systems.
The quest to reduce the footprint of restrictive housing in U.S. prisons cannot occur without
accounting for one of the most difficult populations for correctional authorities to manage, that
is, the 213,000 prisoners affiliated with gangs. The challenge will be greater in prison systems
with larger custodial populations, a higher proportion of confirmed gang affiliates, higher rates
of gang-related violence, and longstanding histories of gangs, where restrictive housing is more
likely to be used disproportionately. Programs and practices aimed to reintegrate gang affiliates
back into general population housing are deserving of research and evaluation owing to their
implications for institutional and community corrections.
Keywords: gangs, gang members, solitary confinement, prisons, correctional policy and practice

1
The use of building tenders in the Texas prison system—inmates officially designated as
responsible for maintaining order and control—was declared unconstitutional in Ruiz v. Estelle
(1980). A consent decree resulted in sweeping reforms, including the dismantling of the building
tender system, beginning in 1982. It is widely understood that the rise of prison gangs was an
unintended consequence of prison reform (Crouch & Marquart, 1989; DiIulio, 1987). Prior to
1983, there was a single gang—the Texas Syndicate—with a presence in Texas prisons (Fong,
1990). By the end of 1985, eight prison gangs existed and the number of gang members increased
by a factor of 25. There were also 43 gang-related homicides over a 21-month period (Fong, 1990,
p. 37) that became known as the “war years” (Ralph & Marquart, 1991, p. 43). Texas was not the
only prison system to witness the emergence of gangs. Camp and Camp (1985) documented prison
gang activity in Arizona, Arkansas, California, Illinois, Missouri, Nevada, Pennsylvania, and the
Federal Bureau of Prisons (BOP). Like prison systems across the country, Texas sought to thwart
the rise of gangs. Their first attempt to regain control entailed transferring inmates and placing
gang leaders in restrictive housing
1
, that is, solitary confinement, but the violence continued. After
the prison administration instituted the placement of all gang affiliates in restrictive housing toward
the end of 1985, violent misconduct declined substantially (Ralph & Marquart, 1991). It is no
coincidence that the proliferation of gangs and expansion of restrictive housing in U.S. prisons
unfolded simultaneously over the last three decades. Yet, correctional agencies are now under
serious pressure to overhaul their use of restrictive housing (Frost & Monteiro, 2016).
It is against this backdrop that we consider the link between gang affiliation and restrictive
housing in U.S. prison systems. On any given day it is estimated that anywhere from 4.4% to 6.4%
1
Restrictive housing refers to the practice of housing inmates in cells that are separate from the general
population, over fixed or indeterminate periods, involving extended physical and social isolation. While there
are many terms for this practice, including solitary confinement, restrictive housing is the umbrella term used
by leading organizations to capture its various uses (American Corrections Association, 2014; Garcia, 2016).

2
of the custodial prison population is housed in restrictive housing (Baumgartel et al., 2015; Beck,
2015) and in some states, 40% of released inmates had spent at least one day in restrictive housing
(Labrecque & Smith, 2017). Inmates are generally placed in restrictive housing for disciplinary,
protective, or administrative purposes (Labrecque, 2016). Owing to their elevated levels of
misconduct, need for protection, and threat to institutional safety, inmates who affiliate with gangs
fit squarely into the logic and practice of restrictive housing (Pyrooz, 2016). Surveys of
correctional officials consistently view restrictive housing as one of the most effective responses
to managing gangs in prison (Mears, 2005; Winterdyk & Ruddell, 2010) and some have even
described it as the “silver bullet” (Vigil, 2006, p. 33). Thus, it is no surprise that some empirical
evidence indicates that gang affiliates are overrepresented—relative non-gang inmates—in
restrictive housing (Labrecque, 2015b; Pyrooz, 2016).
But the current state of the evidence on the relationship between gang affiliation and
restrictive housing is limited. First, this research is based on a small sample of prison systems.
There is considerable variation in gang policies and practices, as well as gang activity, across
prison systems. Some prison systems will place gang affiliates in restrictive housing based on their
status as gang members, while others do not (Butler, Griffin, & Johnson, 2013; Trulson, Marquart,
& Kawucha, 2006). Some prison systems have longstanding histories and high levels of gang
activity, particularly those where the research on this topic has been generated (e.g., California,
Texas), while others do not. If we are to reach sound conclusions on the relationship between gang
affiliation and restrictive housing, it is necessary to examine it across the population of U.S. prison
systems, or we run the risk of a few prison systems driving our understanding of the issue. Second,
if gang affiliates are indeed placed disproportionately in restrictive housing, the reasons for this
relationship remain speculative at best. It could be due to their problem behavior, need for

3
protection, or their preordained threat to institutional order and safety. If an empirical relationship
between gang affiliation and restrictive housing is observed, there is a need to uncover what gives
rise to it. Overall, it is essential to better understand the gang affiliation-restrictive housing link in
U.S. prisons, which has broader relevance to institutional and community corrections.
The purpose of this study is to examine the relationship between gang affiliation and
restrictive housing in U.S. prisons. Accordingly, we provide much-needed evidence that addresses
five research questions: (1) What is the frequency and proportion of the custodial population in
restrictive housing and classified as gang affiliates, respectively? (2) Are gang affiliates more
likely to be placed in restrictive housing relative to non-gang inmates? (3) Is the placement of gang
affiliates in restrictive housing primarily for disciplinary, protective, or administrative purposes?
(4) How much variation exists across prison systems in the use of restrictive housing on gang
affiliates? (5) What explains variation between prison systems in the use of restrictive housing on
gang affiliates? We answer these questions using data from the Gangs, Gang Members, and
Housing: 2016 survey, as well as data from the Bureau of Justice Statistics, where our unit of
analysis is the prison system. While recent studies have shed light on restrictive housing practices
nationally, including key correlates, the lack of information about gang affiliation remains a
serious void. This void takes on added significance given that restrictive housing is a core strategy
used to manage gangs. To the extent that correctional officials, policymakers, and advocacy groups
seek to reduce the footprint of restrictive housing in U.S. prisons (e.g., Haney, 2018; Obama,
2016), it is necessary to understand what role gang affiliation plays in the contemporary
correctional landscape of restrictive housing.
An Overview of Restrictive Housing in U.S. Prison Systems
While variation across prison systems exists, the practice of restrictive housing generally

4
takes on similar features: inmates are segregated from the general population, typically in single-
cell housing, over fixed or indeterminate periods, and are isolated from other inmates. There are
three primary purposes of restrictive housing: disciplinary, protective, and administrative
(Labrecque, 2016). Restrictive housing for the purpose of punishing inmates who violate rules is
termed disciplinary segregation. Inmates commonly receive fixed sentences in restrictive housing
after being adjudicated guilty of a rule violation. Restrictive housing for the purpose of protecting
inmates who are vulnerable to victimization or self-harm is termed protective segregation. These
inmates often remain in protective custody until the vulnerability subsides. Variation also exists
within protective custody; some institutions may completely segregate inmates such as celebrities,
former police officers, or well-known criminal offenders, whereas other institutions may allow
inmates to congregate. Finally, restrictive housing due to the threat or risk posed by the inmate to
the institution is termed administrative segregation and is especially important in the context of
gang affiliates. Inmates often remain in administrative segregation for indeterminate periods
because something about the individual—such as escape risk or gang affiliation—is considered an
inflexible characteristic that is believed to create continued a threat or risk to the prison system.
While the use of restrictive forms of housing has a long tradition in U.S. prisons (Haney &
Lynch, 1997; Hinds & Butler, 2015; King, 1999), its practice has grown considerably over the last
several decades (Butler et al., 2013; King, 1999; Naday, Freilich, & Mellow, 2008). In 1984, one
state maintained a supermax prison facility designed for isolated inmate housing (Kurki & Morris,
2001). That figure increased to 32 states in 1996 (National Institute of Corrections, 1997) and 44
states by 2004 (Mears, 2005). While there is no source of data that has systematically tracked the
number of inmates, beds, or facilities using restrictive housing over an extended period of time,
several studies have produced estimates of the scope of restrictive housing. Naday et al. (2008)

5
reported an average of 26,177 inmates were housed in administrative segregation annually between
1997 and 2002 based on data from 51 agencies (excluding the BOP); these estimates equated to
2.2% of correctional populations using pooled agency-level figures (Harrison & Beck, 2003).
More recent data on the scope of restrictive housing are found in the work of Baumgartel
et al. (2015) and Beck (2015). Based on the National Inmate Survey, 2011-12, which consists of a
representative sample of 38,251 inmates in 233 state and federal prisons, Beck found that
approximately 4.4% of inmates reported spending “last night” in restrictive housing. This figure
is double that of the findings reported by Naday et al. (2008), although Beck’s estimates are based
on restrictive housing, broadly defined. Beck also found that nearly 20% of inmates reported
spending at least one day in restrictive housing in the last year, a figure that reveals a much greater
exposure to restrictive housing than daily snapshots. Baumgartel et al.’s findings are based on the
Yale University/Association of State Correctional Administrators national survey of state and
federal agencies. The study identified that approximately 6% of the custodial population in 34
jurisdictions was in restrictive housing in 2014. In terms of administrative segregation, where 39
states had valid data, Baumgartel et al.’s findings revealed that there were a total of 29,848 inmates
housed in administrative segregation. Overall, these figures indicate a continued and increased
reliance on restrictive housing in U.S. prisons.
It is expected that there is a relationship between gang affiliation and restrictive housing.
Indeed, correctional officials view this practice as an effective strategy for managing gangs in
prison. Mears (2005) found that 73% of the 416 wardens who participated in a national survey
believed that supermax confinement decreased the influence of gangs. Winterdyk and Ruddell
(2010) surveyed prison administrators in 37 prison systems about gang and security threat group
management strategies. The segregation or isolation of gang members was viewed as the most

6
effective strategy to manage gangs, with 75% of the respondents holding that is was “very
effective.” There are few effective options for managing gangs in prison, but restrictive housing
appears to one that is highly endorsed among prison officials.
The Logic and Practice of Segregating Gang Affiliates
Gang affiliates fit squarely into the logic and practice of restrictive housing (Pyrooz, 2016).
In terms of disciplinary uses, removing high-rate offenders from the general population and
placing them in restrictive housing units is believed to reduce individual misconduct and
institutional disorder in a manner that is consistent with deterrence, incapacitation, and
normalization theories (Mears & Reisig, 2006; Mears & Watson, 2006). Gang affiliates have long
been considered the “bad apples” in prisons (e.g., DiIulio, 1987), which places them within the
crosshairs of correctional policies as well as the most prominent theories of restrictive housing.
The removal of gang affiliates from the general population serves as a specific and general
deterrent by sending a message to the individual inmate and population at large that gang activity
will not be tolerated in prison. The use of restrictive housing also incapacitates a population that
is at high risk of engaging in institutional misconduct. In fact, studies consistently reveal that gang
affiliates are involved in higher rates of misconduct than are non-gang inmates (Gaes, Wallace,
Gilman, Klein-Saffran, & Suppa, 2002; Griffin & Hepburn, 2006; Huebner, 2003; Sheldon, 1991).
The gang-related group processes that lead gang affiliates to engage in misconduct are likely to
result in their placement in restrictive housing. With gang members placed into restrictive housing,
prison operations may be ‘normalized’ and institutional order reinstated as prisons to return to a
gang-free benchmark, where inmates can live in prison facilities and participate in programming
without the conflicts and intimidation associated with gangs.
In terms of protective uses of restrictive housing, there are numerous scenarios where gang

7
members would require secure housing due to victimization vulnerability. Gang affiliates could be
housed in a facility where rival gangs outnumber them, which may result in elevated tensions or
even violence. Additionally, gang dropouts may have outstanding debts or have violated gang
codes of conduct, particularly in the event that they have “debriefed” or informed correctional
officials of gang practices and structure (e.g., see Toch, 2007). There are also instances where gang
affiliates bring personal conflicts with fellow gangs and gang members from the street into the
prison. In fact, it is not uncommon for gang members to request to be housed in protective custody
(Fong & Buentello, 1991) or to be placed in protective custody upon debriefing (Fischer, 2002).
The administrative segregation of gang affiliates ranks among the most controversial uses
of restrictive housing. Unlike disciplinary and protective purposes, administrative segregation is
based on the potential for problem behavior rather than an inmate having “earned” or “needed”
restrictive housing. Also, placement is usually indeterminate, that is, until the threat wanes.
Because gang affiliation is the primary determinant of the threat, release to the prison’s general
population requires that gang members convince authorities that they are no longer affiliated with
a gang (Burman, 2012; Tachiki, 1995), which is a risky or uncertain endeavor. The decision to
place an inmate into restrictive housing is made administratively, which often entails wide
discretionary latitude. In cases of gang affiliation, the discretion that poses the greatest concern
involves the classification and validation of inmates as gang members (Toch, 2007).
2
Several reports have documented prison system policies associated with administrative
segregation, including Butler et al. (2013) and Lee and Jacobs (2012). Butler et al. (2013) reviewed
policies of 42 state prison systems in 2010 and determined that affiliation with a gang or STG was
2
The wholesale placement of gang affiliates in restrictive housing for indeterminate periods was the basis of the
Ashker v. Governor of California class-action lawsuit alleging eighth amendment violations of cruel and unusual
punishment, filed in 2012 and settled in 2015. We return to this point in this discussion section.

8
an adequate reason for placement in long-term supermax confinement in 15 of the 42 (36%) states.
As part of their investigative reporting for Mother Jones magazine, Lee and Jacobs obtained
information from 44 state agencies and identified 13 states where validation as a gang affiliate
could result in placement in restrictive housing. There was 69% convergence between the two
reports, which could be due to changes in policies and practices or the methods used to obtain the
information. Altogether 21 prison systems place gang affiliates in restrictive housing based on
their status as gang members rather than their behavior as inmates.
We hypothesize that gang affiliates are overrepresented in restrictive housing relative to
non-gang inmates. We also hypothesize that gang affiliates will be more likely to be placed in
restrictive housing relative to non-gang affiliates for disciplinary, protective, and administrative
purposes. We take no position on which specific pathway into restrictive housing is used more or
less on gang affiliates. Owing to the regional differences in the nature of prison gang activity and
the policies and practices of prison systems, we expect to observe variation across prison systems
in the overrepresentation of gang affiliates in restrictive housing, including specific pathways.
Evidence of the Overrepresentation of Gang Affiliates in Restrictive Housing
Several scholars have examined the relationship between gang affiliation and restrictive
housing in the United States.
3
Beck (2015) use of the National Inmate Survey (NIS), 2011-12,
sheds light on this relationship at the facility-level. Approximately one in six inmates reported that
there was “a lot” of gang activity in their facility over the last 12 months, which was positively
correlated with the concentration of inmates who have spent any time and 30 days or more in
restrictive housing. This shows a facility-level link between gang activity and restrictive housing.
3
Given the aims of our paper we reviewed literature based in the United States. Although we are not familiar with
research on this topic outside of the United States, the extent to which we could generalize such findings is
questionable. We thank an anonymous reviewer for raising this important point.

9
Labrecque (2015a, 2015b) conducted several individual-level analyses of the relationship
between gang affiliation and restrictive housing. First, three of the 16 studies in his meta-analysis
of restrictive housing correlates included (seven) effect sizes for gang affiliation, which was
correlated with restrictive housing at 0.15. Second, Labrecque observed a modest relationship
between a history of gang affiliation and any placement and length of placement in restrictive
housing in the Ohio prison system. More recently, using data from a Midwestern prison system,
Labrecque and Smith (2017) found that gang affiliates were twice as likely to be placed in
restrictive housing than non-gang inmates.
Pyrooz (2016) examined official reports of gang affiliation and restrictive housing
published by prison systems in California and Texas. Three-fourths of confirmed gang affiliates
in California and 35% of confirmed gang affiliates in Texas were in restrictive housing, compared
to 1.1% and 2.8% of non-gang inmates, respectively. The relative risk of placement into restrictive
housing was 71 times greater in California and 16 times greater in Texas for gang affiliates than
non-gang inmates.
Current Study
Both theory and research suggest a link between gang affiliation and restrictive housing.
However, the best evidence to date on this issue comes from a handful of states, such as California,
Ohio, and Texas. It is unclear if the findings observed in these few states will generalize to prison
systems throughout the United States. Consequently, we lack a national understanding—an
aggregate picture across the country—of the relationship between gang affiliation and restrictive
housing. And what is known about the relationship between restrictive housing and gang affiliation
reveals very little about the mechanisms that bring it about. The overrepresentation of gang
affiliates in restrictive housing could be due to discipline for misconduct, the need for protection

10
from other inmates, or the risk or threat to institutional safety posed by gang affiliates. In other
words, there are sound reasons to anticipate why this relationship exists. Our aims are to document
the relationship, not to disentangle whether the disproportionality we hypothesize is a product of
bias on the part of prison systems. Our data are not positioned to address such a question, a point
we revisit in the discussion section. Accordingly, we address five research questions:
(1) What is the frequency and proportion of the custodial population in restrictive housing and
classified as gang affiliated, respectively, in the United States?
(2) Are gang affiliates more likely to be placed in restrictive housing than non-gang inmates?
(3) Is the placement of gang affiliates in restrictive housing primarily for discipline, protection,
or administrative purposes?
(4) How much variation exists across prison systems in the use of restrictive housing on gang
affiliates?
(5) What explains variation between prison systems in the use of restrictive housing on gang
affiliates?
These key questions are of critical importance to institutional corrections generally and restrictive
housing practices specifically. Understanding the sources of placement into restrictive housing has
been subject to considerable research in recent years (see e.g., Cochran, Toman, Mears, & Bales,
2018; Labrecque, 2015b). Despite the centrality of gangs to the national conversation on restrictive
housing (Reiter, 2016), as well as the social order of prisons (Skarbek, 2014), our knowledge of
the relationship between gang affiliation and restrictive housing remains woefully inadequate,
which is likely attributed to the absence of measures of gang affiliation integrated into the research
infrastructure on prisons. We aim to fill this void.
Methods
Data
The data for this study were drawn from the Gangs, Gang Members, and Housing: 2016

11
survey. The survey was administered to representatives of the 51 prison systems—50 states and
the Federal Bureau of Prisons—in the United States, who were asked to report population-level
information in their jurisdiction. Therefore, our unit of analysis is prison systems. Initial contact
was made with prison representatives in late December 2015, where they were emailed a brief
description of the project and a form-fillable PDF version of the survey. Contacts with prison
systems were established initially based on participation in the National Institute of Justice topical
working group on the use of restrictive housing in U.S. prisons (Marie Garcia, 2016). Additional
contacts were gathered through web-based searches and phone calls. Email reminders were sent
to non-responding prison systems one month later (January 2016). A third contact was attempted
in March 2016, which involved phone calls to prison executives and public information officers.
A final attempt to gather data from prison systems was made in early February 2017.
A total of 39 prison systems completed the survey, resulting in a 76% response rate. The
jurisdictions that completed the survey housed 983,211 inmates, representing 73% of the state
prison population in 2015 (Carson, 2016). Figure 1 details the prison systems that participated in
the study.
4
Non-participants included four jurisdictions that refused to participate in the study,
including the Federal Bureau of Prisons, six jurisdictions that never responded to email or phone
requests for participation, and two jurisdictions that verbally committed to participate but never
submitted a completed survey.
***Insert Figure 1 here***
As Frost and Monteiro (2016) and Pyrooz (2016) observed with respect to restrictive
housing and gang affiliation, both terminology and definitions vary across prison systems. To
4
Respondents held positions that ranged from security threat group coordinators, operations chiefs, prison directors,
and research analysts. The data collection process involved substantial correspondence with staff to determine the
appropriate contacts, encourage participation, and clarify questions and responses to the survey. The results are
nonetheless susceptible to the limitations of surveying correctional personnel and administrative data.

12
foster consistent definitions across agencies, we provided prison systems an operational definition
of restrictive housing, as well as its subtypes (i.e., discipline, protection, administrative, and other),
emphasizing (a) separation from the general population, (b) single-cell housing, (c) limited or no
contact with other inmates, and (d) confinement to cell for extended periods. These four
components are consistent with leading research on restrictive housing (e.g., Baumgartel et al.,
2015). We did not provide prison systems with a definition of a gang or security threat group, as
our focus was system-level reports of inmates who affiliate with what the prison system termed
gangs or STGs. Nor did we ask respondents to distinguish between types of gangs or security threat
groups, which is common in some larger prison systems (e.g., California and Texas). We did ask
respondents to distinguish between suspected/associated (hereafter, suspected) and
confirmed/validated (hereafter, confirmed) gang affiliates, noting that the inmate must have met
some or all of an agency’s criteria to be classified as a suspected or confirmed gang affiliate,
respectively. This is a notable advance in the research on gang affiliation and restrictive housing.
The Gangs, Gang Members, and Housing: 2016 survey contained 27 items within four
topical areas. First, the total custodial population of inmates as well as the number of suspected
and confirmed gang affiliates (3 items). Second, the total custodial population of inmates in four
subtypes of restrictive housing—disciplinary, protective, administrative, and other—along with
the number of suspected and confirmed gang affiliates in each (12 items). Third, a series of
questions pertaining to gang-related problems within the prison system was included (7 items).
Fourth, information about gang-related policies and practices was requested (5 items). Together,
these items paint a rich picture of population-level gang activity and responses in U.S. prisons.
Incomplete responses were not uncommon in the surveys that were returned. We sought
clarification when possible, but several prison systems lacked the technological infrastructure or

13
reporting practices to distinguish between subtypes of restrictive housing or gang affiliates. In
these cases, where it was impractical or unreasonable to request information across subtypes of
restrictive housing or gang affiliation, we asked respondents to report aggregate information. Of
course, it is preferable to secure any data from prison systems in these areas rather than no data,
especially given the current state of knowledge on the topic. Accordingly, all of our tables identify
the valid sample size when reporting any values relevant to our research questions. Also, national-
level estimates of restrictive housing and gang affiliation are presented with valid data and with
imputation data for only one analysis.
5
Measures
Restrictive housing. The survey queried respondents on the total number of inmates in
their prison system in four possible types of restrictive housing: disciplinary, protective,
administrative, and other. Disciplinary purposes were defined as inmate behavior, such as violence
or other forms of misconduct, leading inmates to be placed in restrictive housing. Protective
purposes were defined as inmates being placed in restrictive housing as a form of protection from
self-harm or harm from other inmates. Administrative purposes were defined as inmates being
placed in restrictive housing because they posed a threat or risk to staff or other inmates. Other
purposes was used as a catch-all for inmates who would not meet the definitions provided above
but were placed in restrictive housing, such as inmates on death row and those in temporary
confinement. Each of these four restrictive housing items are reported as frequencies and
5
Twelve agencies were completely missing, 19 agencies had complete data, and 20 agencies contained partial data
(missing 24%-55% of variables). Missing patterns were not randomly distributed according to a missing completely
at random test (2(171, N = 51) = 182.69, p = 0.26) (Li, 2013; Little, 1988). Missing data were multiply imputed using
chained equations in the mi suite in Stata 13 (StataCorp, College Station, TX). Variables with complete data included
BJS homicide and suicide rates, along with the 2015 custodial population. We generated five imputation data sets for
incomplete data on suspected and confirmed gang members, and the count of inmates in restrictive housing for
disciplinary, protective, administrative, and other reasons. Pooled parameter estimates were calculated using mi
estimate and used for only Table 2—national-level estimates of gang affiliation and restrictive housing.

14
proportions, both at the national-level (i.e., summed across prison systems with valid responses)
and prison system-level (i.e., means of prison systems with valid responses). This allows us to
identify the proportion of inmates in restrictive housing and its various subtypes, consistent with
prior research on the topic (e.g., Baumgartel et al., 2015).
Gang affiliation. Three measures of gang affiliation were included. All gang affiliates
refers to the total number of inmates in the prison system who are classified as having non-zero
levels of gang affiliation. This typically refers to inmates who meet one or more affiliation
identifiers or source items used to determine an affiliation with a gang (see Hill, 2009). Using
prison systems as the unit of analysis, 18.5% of the inmate population was classified as a gang
affiliate (see Table 1). Nested within this total are two items that distinguish between key
subgroups of gang affiliates, that is, inmates who are suspected and confirmed as gang affiliates.
We did not ask prison systems to tell us how this determination was made, but there is typically a
threshold of three items or a summative threshold of points associated with items to distinguish
between suspected and confirmed gang affiliates (Hill, 2009). Suspected gang affiliates,
comprising 9.8% of inmates across prison systems, refers to the proportion of inmates with non-
zero levels of gang affiliation, but not enough to be considered a confirmed gang affiliate.
Confirmed gang affiliates, comprising 10.3% of inmates across prison systems, refers to the
proportion of inmates who cross the validation threshold to be considered confirmed as gang
members. These proportions diverge from the proportion of all gang affiliates because not all
prison system reported on suspected (n = 31) or confirmed (n = 36) gang affiliates. As we have
established, very little is known about gangs in U.S. prisons, and the information we collected
updates and extends the work of Winterdyk and Ruddell (2010), which until now had been the
most current estimates of the proportion of gang affiliates across prison systems.

15
***Insert Table 1 here***
Gang-related correlates. A number of gang-related variables (see Table 1) are included in
this study to explain variability in the use of restrictive housing on gang affiliates across prison
systems. It is expected that prison systems with more gangs, a greater proportion of gang affiliates,
a longer history of gang activity, and higher rates of gang affiliated misconduct should rely more
heavily on restrictive housing. These are prison systems that likely have more gang conflict,
difficulty finding neutral housing arrangements, established gang rivalries and alliances, and
challenges in administering programs undisturbed by the presence of gangs; therefore, they may
rely on restrictive housing as a means to quell such challenges (Jacobs, 2001; Knox &
Tromanhauser, 1991; Winterdyk & Ruddell, 2010). Gangs/STGs refers to the total number of
gangs and/or security threat groups that are active in the prison system. Jurisdictions reported an
average of 204 gangs/STGs within their prisons (range = 4 – 1,534). History of gang emergence
measures when gangs emerged as an issue within a jurisdiction. We accounted for this by including
items for gang emergence in the 1960s/1970s (18%), 1980s (16%), 1990s (50%), 2000s (13%),
and never (3%). A series of questions about gang-related misconduct were included: murder,
assaults on inmates, and assaults on staff. Gang-related misconduct incidents were reported as
frequencies and then converted to rates per 10,000 inmates. Across prison systems with valid data,
there were 0.2 murders, 103 assaults on inmates, and 19 assaults on staff, per 10,000 inmates.
We also include three additional items related to pathways into and out of restrictive
housing for gang affiliates. Prison systems were queried about unique reasons for placing gang
affiliates in restrictive housing. Gang status segregation refers to the ability to place inmates in
restrictive housing based “solely on their gang affiliation” (“no” = 0, “yes” = 1), and existed in
13% of jurisdictions. Gang risk segregation, relied upon in 74% of jurisdictions, refers to the

16
ability to place gang or STG members in restrictive housing “if they were deemed a security risk
but haven’t engaged in behavioral misconduct” (“no” = 0, “yes” = 1). Prison systems were also
queried about programs for leaving gangs, which is closely related to administrative purposes for
placing gang affiliates in restrictive housing based on gang status. Gang exit programs refers to a
prison system having a program “for gang members that is designed to facilitate leaving their
gang” (“no” = 0, “yes” = 1). These programs were available in 37% of reporting agencies.
Non-gang-related correlates. Since it is unlikely that gang-related correlates alone will
explain the association between gang affiliates and restrictive housing, we also included several
non-gang-related correlates in our analysis. Prison population is positively correlated with
misconduct (Steiner, Butler, & Ellison, 2014) and believed to be associated with gang activity
(Skarbek, 2014). Therefore, we include a measure of prison population, which is the custodial
population, or the total number of inmates “on hand,” as reported by respondents in our survey.
The average population was 25,211 (range = 1,343 – 148,521). Data on inmate turnover,
overcrowding, homicide, suicide, and assault, were gathered from the Bureau of Justice Statistics
(BJS). Counts from the 2015 National Prisoner Statistics data were used to measure prison
overcrowding and changes in prison populations via incarceration and reentry. Overcrowding is
the percent of capacity within a prison system based on the custody population for the largest
capacity. On average, prison systems were operating at 97% capacity (range = 52% – 145%).
Turnover was created to represent the yearly variation in releases, relative to the correctional
population, computed by the yearly count of releases divided by the population, and multiplied by
100. The average turnover rate was 48% (range = 16% – 120%). Data from the Deaths in Custody
Reporting Program were used to account for the average homicide and suicide rates per 10,000
inmates at each institution from 2001 to 2014, which were 0.4 and 1.8, respectively. Finally, data

17
from the 2005 Census of State and Federal Adult Correctional Facilities were used to create an
assault rate, per 10,000 inmates, for each jurisdiction. The average number of assaults was 101 per
10,000 inmates.
Analytic Strategy
Our analysis proceeds in stages that follow our five research questions. First, we report
national-level estimates—using the country as the unit of aggregation—of the frequency and
proportion of inmates who are in restrictive housing and affiliates of gangs, respectively. That is,
we summed responses across the prison systems to produce frequencies and proportions at the
national-level. Estimates are provided for our sample with valid data (N = 39) and for the sample
with imputed data (N = 51).
Second, we report a series of bivariate statistics on the relationship between gang affiliation
and restrictive housing. We report the proportion of gang and non-gang affiliates in restrictive
housing. We do this at the national-level, as described above, and at prison system-level, where
we report central tendencies across prison systems. Then, we compute the relative risk of
placement into restrictive housing. Drawing from the Campbell Collaboration (Higgins & Green,
2011), we do this by computing a relative risk ratio nationally and relative risk ratios for each
prison system, where we divide the proportion of gang affiliates in restrictive housing by the
proportion of non-gang inmates in restrictive housing. At the national-level, we report an exact
value. At the prison system-level, we report a mean risk ratio.
Third, we further differentiate restrictive housing by the three leading purposes: discipline,
protection, and administrative. Similar to above, we do this at the national-level and the prison
system-level. We distinguish across subtypes of restrictive housing and report proportions of gang
and non-gang inmates in restrictive housing, as well as the risk ratios for gang affiliates.

18
Fourth, we examine variability across prison systems in their application of restrictive
housing and its various subtypes on gang affiliates. In addition to measures of dispersion, we
illustrate variability in a figure through the blinded rank-ordering of prison systems according to
relative risk ratios. Finally, we assess whether it is possible to explain variability in the use of
restrictive housing on gang affiliates. We do this across a series of bivariate OLS regression models
where we regress relative risk ratios for all forms of restrictive housing on the correlates noted
above. Since we have a finite sample of prison systems, we report unstandardized and standardized
regression coefficients in addition to null hypothesis significance tests.
At face value, the analyses may seem simplistic; however, this descriptive analysis
addresses a serious void in our understanding of restrictive housing. Together, this approach will
paint the most comprehensive and up-to-date picture of gang affiliation and restrictive housing,
and their interrelationship, in U.S. prisons.
Results
National Estimates of the Frequency and Proportion of Inmates in Restrictive Housing and
Inmates Classified as Gang Affiliates
National-level estimates based on the valid and imputed data are presented in Table 2.
Based on our valid data, around 967,000 inmates were incarcerated across reporting agencies (N
= 38),
6
and of those inmates, 61,706 were in restrictive housing, or 6% of the custodial population.
This estimate closely approximates prior prevalence estimates using daily snapshots, which range
anywhere from 4 – 6% of the prison population (Baumgartel et al., 2015; Beck, 2015). Turning to
our imputation data, where the proportions closely approximate our valid data, we find that 92,295
inmates were in restrictive housing. The most common reason for restrictive housing placement
6
These values diverge slightly from what we reported in the data section on the number of prison systems (N = 39)
and prisoners (983,211), owing to missing data on restrictive housing and gang affiliation in a single, but separate,
prison system for the respective values.

19
was for discipline—constituting around 30,595 inmates, or 2% of the custodial population. About
1% of inmates were segregated for protective reasons (n = 10,157), 1.6% for administrative reasons
(n = 24,574), and 1.3% (n = 20,535) for other reasons. Our administrative segregation estimates
closely approximate Naday et al.’s (2008), although they fall below Baumgartel et al.’s (2015),
findings. Nonetheless, these estimates are particularly useful because they disaggregate restrictive
housing into its various purposes.
Consistent with the prior work of Winterdyk and Ruddell (2010), our sample with valid
information revealed that a total of 148,185 inmates, or 15.4% of the correctional population, were
identified as gang affiliates. Suspected gang affiliates composed 7.9% (n = 56,111) of the custodial
population, while confirmed affiliates made up 9.5% (n = 88,357). The latter is slightly lower than
the 11.7% estimate provided by Winterdyk and Ruddell (2010).
7
Using imputed data, these
estimates slightly change. The imputation data estimates demonstrate a slightly lower prevalence
of gang affiliation—for all gang affiliates, as well as suspected and confirmed. Gang affiliates
made up 14% (n = 213,215) of the population in the imputed data, compared to 15.4% in the survey
data. Of those gang affiliates, 5.3% were identified as suspected and 8.4% as confirmed affiliates
using imputed data.
Collectively, these findings demonstrate that both the population of inmates in restrictive
housing and the population of inmates in gangs constitute a non-trivial minority of the custodial
population. Yet, these two populations account for many of the leading issues facing correctional
institutions, and are believed to intersect in important ways. We now shift our analysis to focusing
on the relationship between gang affiliation and restrictive housing in U.S. prisons.
National- and State-Level Estimates of the Relationship between Gang Affiliation and
Restrictive Housing, All Forms and Subtypes
7
Not all prison systems reported valid data for the confirmed and suspected gang affiliates, which is why the sum of
confirmed and suspected gang affiliates is not equivalent to the total number of gang affiliates.

20
Table 3 reports the proportion of gang and non-gang inmates in restrictive housing, as well
as the relative risk ratios, both at the national- and prison system-level, using the valid data. The
national estimates indicate that 12% of gang affiliates are in restrictive housing, compared to only
4% of non-gang inmates. Based on these numbers, the relative risk of placement into restrictive
housing for gang affiliates is 3.1 times higher than the risk of placement for non-gang affiliates.
Gang affiliates account for over one-third of the inmates in restrictive housing. While they
constitute a smaller minority of inmates in U.S. prisons, they comprise a much larger minority of
the inmates in restrictive housing.
8
We disaggregate this relationship further to evaluate the differential risk of placement for
suspected and confirmed gang affiliates. Whereas 10% of suspected gang affiliates are in
restrictive housing, 13% of confirmed gang affiliates are in restrictive housing, translating to a
relative risk that is 2.6 and 3.4 times higher relative to non-gang inmates. These findings suggest
that at the national-level, gang affiliates are at a greater risk of placement into restrictive housing
than non-gang affiliates.
Turning to the prison system-level estimates, we again confirm that gang affiliates are
overrepresented in restrictive housing. On average, the risk of placement in restrictive housing is
4 times greater for gang affiliates than non-gang inmates. But we are particularly interested in
whether gang affiliates’ risk of placement into restrictive housing is equal across purposes, that is,
for discipline, protective, administrative, or other reasons. Our findings suggest that gang affiliates
have an increased risk of placement in restrictive housing regardless of the purpose, but the
magnitude of those differences is not equal. For administrative reasons, the relative risk of
8
The share of inmates in restrictive housing who are affiliated with gangs is based on 34 prison systems with a
custodial population of 771,265 inmates, of whom 122,734 were classified as gang affiliated. The valid data
indicated that 17,401 of the 48,737 inmates in restrictive housing were gang affiliated, or 35.9%.

21
placement in restrictive housing is over 6 times greater for gang affiliates than it is for non-gang
inmates, whereas the risk ratios were comparably smaller—albeit still large in magnitude—for
disciplinary (RRR = 3.1), protective (RRR = 2.6) and other (RRR= 2.0) purposes.
9
A similar pattern emerges when disaggregating the reasons for placement by suspected and
confirmed gang affiliates. The risk of placement in restrictive housing for confirmed or suspected
affiliates was at least two times higher than the risk of placement for non-gang affiliates. These
estimates, however, varied by gang affiliate status. Overall, the risk ratio for placement in all forms
of restrictive housing was greater for confirmed gang affiliates (3.7) than suspected gang affiliates
(2.8). This effect appears to be driven by the large disparity in placement in restrictive housing for
administrative purposes. Indeed, the relative risk ratio for confirmed gang members was 7.3, while
about half the size (3.8) for suspected gang affiliates. Yet, it is notable that placement in restrictive
housing for disciplinary purposes was higher for suspected gang affiliates (3.7) than for confirmed
gang affiliates (2.7). The differences in the relative risk ratios for the remaining restrictive housing
measures—protective and other—were not substantively significant.
Overall, these findings confirm our hypotheses that gang affiliates fit squarely into the logic
and practice of restrictive housing in U.S. prison systems. Both suspected and confirmed gang
affiliates are overrepresented in restrictive housing for disciplinary, protective, and administrative
purposes. But confirmed gang affiliates bear the brunt of the restrictive housing policies and
practices, and especially for administrative reasons.
Variation in the use of Restrictive Housing on Gang Affiliates, All Forms and Subtypes
Figure 2 illustrates the variation across prison systems in the relative risk ratios for
9
Of course, the sample sizes change from estimate to estimate, but we confirm similar patterns when relying only
upon complete case data. It is also notable that the standard deviation suggests greater dispersion around the mean
for administrative purposes than the others, a point we return to in the next section.

22
placement of gang affiliates into restrictive housing, distinguishing between the various purposes
for placement. Substantial variation exists across prison systems. On average, the risk of placement
into restrictive housing for any reason was 4.0 (SD = 4.1, range = 0 – 19.2). All but three prison
systems had a relative risk score that exceeded one, that is, equal risk between gang and non-gang
inmates. The average risk of placement into restrictive housing for disciplinary reasons for gang
affiliates was 3.1 (SD = 2.0, range = 0 – 10.4). All but two of the prison systems (> 90%) had a
risk ratio score that exceeded one. For protective reasons, the risk of placement into restrictive
housing was 2.6 (SD = 2.8; range = 0 – 12.2) for gang affiliates relative to non-gang affiliates, and
most prison systems (> 75%) exceeded a risk score value of one. But the most substantial
difference in risk was for administrative placement into restrictive housing. Indeed, gang affiliates
were six times (SD = 8.6; range = 0 – 42.4) more likely than non-gang affiliates to be placed into
restrictive housing for administrative reasons. Only two prison systems maintained a relative risk
ratio that fell below one. This suggests that the overrepresentation of gang affiliates in restrictive
housing is much greater in some prison systems than in others.
Altogether, these findings confirm that both the relative risk of placement into restrictive
housing, as well as the purpose, varied substantially across prison systems. In some, the risk for
placement into segregated housing for administrative reasons was very low, whereas in others, risk
was as high as 42 times that of non-gang affiliates. It is plausible that these differences are a result
of the policies and practices found within each prison system. It is also plausible that these
differences have less to do with policies and practices than they do the gang and non-gang-related
characteristics that may give rise to them. We turn to the correlates of relative risk ratios to
understand variation in gang affiliates’ placement in restrictive housing.
Correlates of the Overrepresentation of Gang Affiliates in Restrictive Housing, All Forms

23
Table 4 presents the results of a series of bivariate OLS regression models where we regress
the relative risk of placement for gang affiliates into restrictive housing on factors believed to be
related to overrepresentation. Four correlates were related to the relative risk ratios at conventional
levels of statistical significance. First, a half-unit increase in the prevalence of confirmed gang
affiliates—representative of a marginal change from the minimum to the maximum value—
corresponded with an 8.86 ( = 1.82) increase in gang affiliates’ relative risk of place placement
in restrictive housing (p < 0.05). A similar finding was observed for the prison population, where
an increase in the custodial population by 1,000 inmates was associated with a 0.07 increase in the
risk ratio of placement ( = 2.23, p < 0.05). Third, the history of gang emergence was related to
risk ratios for those prison systems where gangs emerged during the 1980s (b = 4.92,  = 1.77, p
< 0.05). Finally, an increase in the rate of gang-related violence, measured by assaults on inmates,
corresponded with a 0.01 ( = 1.43) increase in the risk ratio for placement (p < 0.10). It is notable
that policies enhancing the likelihood of placement of confirmed gang affiliates in restrictive
housing did not correspond with the expected increase in the overrepresentation of gang affiliates.
Discussion
In 2009, Todd Ashker and Danny Troxell, both of whom were convicted murderers and
purported affiliates of the Aryan Brotherhood prison gang, filed a pro se complaint against the
California Department of Corrections and Rehabilitation (CDCR) for their indeterminate
placement in secure housing units (SHUs, i.e., restrictive housing). While it is common for inmates
to file pro se complaints against prison systems, this one was different. First, high-ranking
members of four gangs—including blacks, whites, and Latinos—orchestrated hunger strikes
protesting restrictive housing practices of gang affiliates in California in 2011 and 2013, the latter
of which included an upwards of 30,000 inmates and attracted international attention (see Reiter,

24
2016, pp. 194–203). Second, the complaint expanded into a federal class action lawsuit—Ashker
v. Governor of California—alleging Eighth Amendment violations of cruel and unusual
punishment was filed in 2012 on behalf of prisoners in SHUs at Pelican Bay. Third, and most
importantly, the lawsuit was settled in 2015, and resulted in sweeping changes to CDCR’s
restrictive housing practices. Whereas our introduction illustrated the historical origins of the use
of restrictive housing to manage gang affiliates in prison systems like Texas, the Ashker lawsuit
illustrated the intense scrutiny and possible reform of the practice. Indeed, leaders at all levels of
government, including President Obama, Supreme Court Justice Anthony Kennedy, and even
prison executives, have spoken out against restrictive housing (Frost & Monteiro, 2016).
This study aimed to determine the extent to which reform efforts to reduce the footprint of
restrictive housing must account for the challenges of gangs and gang affiliates.
10
We sought to
develop a national understanding of the relationship between gang affiliation and restrictive
housing in U.S. prisons. Since representative surveys of prison inmates do not include items about
gang affiliation (e.g., Beck, 2015), and since our current state of knowledge is driven only by a
handful of prison systems, the next-best option was to survey the 51 prison systems in the country.
We added much-needed nuance to the study of the link between gang affiliation and restrictive
housing by identifying the potential for overrepresentation by reasons for placement as well as by
gang affiliate subclassifications. Several key points bear further consideration for the future
research on this relationship as well as correctional policies and practices.
First, gang affiliates are overrepresented in restrictive housing across prison systems
throughout the United States. We found that the relative risk ratio for gang affiliates’ placement in
10
In addition to California, other incidents related have captured the national spotlight, including the execution of
the executive director of the Colorado prison system, Tom Clements, on his doorstep by an affiliate of the 211 Crew
who spent time in restrictive housing (Lin, 2017; Prendergast, 2014). Colorado has drastically reduced its use of
restrictive housing (Raemisch, 2018), including ceasing the placement of gang affiliates based on their gang status.

25
restrictive housing exceeded 1 in over 90% of the 34 prison systems for which we had valid data.
This means is that the overrepresentation of gang affiliates in restrictive housing is not limited to
California, Ohio, or Texas—the three prison systems that have come to represent our
understanding of this relationship. Rather, this disparity in the use of restrictive housing extends
across the vast majority of prison systems represented in our data. Overall, 12% of gang affiliates
in U.S. prisons are in restrictive housing on any given day, compared to just 4% of non-gang
inmates, translating into a relative risk that is over 3 times greater for gang affiliates.
Second, our results confirm that gang affiliates are overrepresented across each of the main
purposes of restrictive housing, consistent with the arguments introduced by Pyrooz (2016). We
found that gang affiliates are overrepresented in restrictive housing because they have “earned it”
(disciplinary RRR = 3.1), “needed it” (protective RRR = 2.6), or they were deemed a “threat”
(administrative RRR = 6.3). Of course, it is the last category, where the greatest disparity exists,
that is the most controversial. Indeed, placement in restrictive housing for administrative purposes
is generally indeterminate, which was a major source of consternation from the prison gang
members in California who united across racial, ethnic, and gang lines to organize the hunger
strikes (Reiter, 2016). The only option to return to general population, as Tachiki (1995) noted in
stark terms, was to snitch on the gang (i.e., debrief), parole to the community, or, ultimately, die.
As one gang affiliate in California told Hunt and colleagues (1993), he was “between a rock and
hard place” and ultimately decided to “do extra time” in restrictive housing (p. 402). Had this
inmate elected to debrief, it was likely that he would end up in protective custody. But since the
policies and practices of California are not in existence across all prison systems, and have not yet
been proven to result in the overrepresentation of gang affiliates in restrictive housing, it was
necessary to understand variation across prison systems.

26
Third, the prison systems where gang affiliates are more overrepresented in restrictive
housing are those with larger custodial populations, higher proportions of confirmed gang
affiliates, higher rates of gang-related inmate assault, and experienced gang emergence in the
1980s. These results are perhaps not surprising. Larger prison systems are also those with
longstanding and violent histories of gang activity, particularly when paired with compositional
changes in inmate demographics and overcrowding, as theorized by Skarbek (2014). Further, the
1980s represented the early and significant stages of prison gang growth much like it did in the
proliferation of street gangs (Klein & Maxson, 2006). The fact that higher proportions of
confirmed gang affiliates and gang-related inmate assaults is related to overrepresentation is
indicative of gang activity in the institutions—as it increases, pressures are placed on prison
administrators to respond. Since prison administrators endorse restrictive housing as an effective
solution to manage gangs (Winterdyk & Ruddell, 2010), and with the boom of restrictive housing
cells coinciding with the rise in mass incarceration (Frost & Monteiro, 2016; King, 1999; Reiter,
2016), it is no wonder that these are correlates of overrepresentation.
There are several key issues of fundamental importance that shape the research and policy
implications associated with the use of restrictive housing to manage gangs and gang affiliates in
U.S. prisons.
First, the quest to reduce the footprint of restrictive housing in U.S. prisons cannot occur
without accounting for one of its main drivers, that is, the estimated 213,000 inmates who were
classified as gang affiliates by prison authorities. Given the urgency placed on reducing the use of
restrictive housing in the United States, and the large numbers of gang affiliates in restrictive
housing, a top priority will involve accomplishing this goal without inflaming tensions that give
rise to gang-related misconduct and violence. To this end, there is a need to take stock empirically

27
of whether restrictive housing actually reduces inmate misconduct and prison disorder as
anticipated (Pyrooz, 2016). Like others, it is necessary to note the observation that, despite decades
of its application, restrictive housing has not appeared to have lessened the existence or threat of
prison gangs in U.S. prisons (e.g., Steiner & Cain, 2016). And even if researchers should uncover
that restrictive housing reduces prison violence and disorder, such benefits must be balanced
against research on the potential psychological and social harms of imprisonment and restrictive
housing (Garcia, 2016; Haney, 2018; Morgan, Robert D et al., 2016; Pyrooz, Gartner, & Smith,
2017). This same logic applies to any problem population overrepresented in restrictive housing.
Second, while we have established a nationwide portrait of the relationship between
restrictive housing and gang affiliation, we would urge researchers at the prison system-level to
drill down into their data to be able to explain the reasons why gang affiliates are overrepresented
in restrictive housing. This study focused on disproportionality rather than bias. Prison authorities
contend that their application of restrictive housing to manage gangs is justified, as the
overrepresentation of gang affiliates is a consequence of discipline, protection, or threat. Whether
that is truly the case is in need of further empirical scrutiny. The challenge will be to “explain
away” the effect of gang affiliation on restrictive housing. Whether those explanations are found,
if at all, in legal or extra-legal factors would be highly useful in shaping the future of restrictive
housing policy and practice. Our work provides a glimpse into possible explanations, but it will be
necessary to draw on data with rich information on individual and situational factors that can track
inmates before, during, and after restrictive housing placement to explain this relationship.
Conducting this analysis with multiple rather than single prison systems should be a priority, lest
we continue to report a piecemeal representation of this relationship.
Third, there is an urgent need to develop or integrate measures of gang affiliation into

28
correctional research generally, but also restrictive housing particularly. Any study that aims to
understand the correlates of restrictive housing but lacks a measure of gang affiliation is missing
an important driver of placement and runs the risk of misspecification. This not only applies to
restrictive housing in general, but also placement for the purposes of discipline, protection, and
administrative control. Our research design was motivated by the absence of measures of gang
affiliation in representative studies (e.g., NIS). Therefore, pushing to develop or integrate repeated
and standardized surveys of prison system-level reporting, representative samples of prison
facilities, or representative samples of inmates will go a long way toward understanding the
frequency and proportion of the overrepresentation of gang affiliates in restrictive housing, but
also trends over time (Pyrooz, 2016, p. 150).
Fourth, the policies and practices underlying the use of restrictive housing to manage gangs
and gang affiliates largely remain in a black box. We have demonstrated that confirmed gang
affiliates are overrepresented in restrictive housing compared to suspected gang affiliates. One of
the most vocal criticisms of restrictive housing is related to gang classifications practices that may
result in the placement of gang affiliates in restrictive housing. These practices are deemed as
highly discretionary, lacking due process, and operating outside of the domain of the general
public, much less researchers with access to prisons (Toch, 2007). Prison systems generally rely
on a fixed set of source items—such as self-reports, legal documents, photographs, tattoos, and
associations—to identify inmates as suspected or confirmed, not unlike the criteria employed by
police agencies (Barrows & Huff, 2009). Whereas police department practices have been subject
to (some) reliability and validity testing, this is not the case in prison systems. Are gang affiliation
classification practices reliable and valid? Given that identifying an inmate as a confirmed gang
affiliate can be the difference between spending a prison sentence in general population and

29
restrictive housing, it is imperative for prison administrations and researchers alike to find out.
Lastly, the reintegration of inmates into the general population and into communities must
account for the added burden that comes with gang affiliation and restrictive housing. Related to
prison reintegration, prison systems throughout the country are developing strategies to reduce
their reliance on restrictive housing, and most now have programs (e.g., step-down) and procedures
(e.g., no indeterminate placement) in place to promote inmates’ exit from restrictive housing
(Chammah, 2016; Ghafar, 2017). Since gang affiliation is the driver of placement into restrictive
housing there is a programmatic need to facilitate disengagement from gangs. Just over one dozen
prison systems maintain a type of program or procedure designed for inmates to leave gangs (see
Pyrooz & Mitchell, 2019). The Texas prison system, for example, has a segregation diversion
program for gang affiliates at admission to prison and the equivalent of a step-down program for
gang affiliates in restrictive housing (see Burman, 2012; Pyrooz, 2016). However, we know very
little about the efficacy of these efforts to lead inmates out of gangs or to lead gang affiliates out
of restrictive housing.
Related to community integration, having spent time within restrictive housing may
exacerbate or facilitate a mental health illness (Kapoor, 2016). The implications of restrictive
housing on an ex-prisoners’ ability to successfully reintegrate back into their community remains
unclear. Some researchers have found that inmates who have been placed in restrictive housing
recidivate at a higher rate than individuals who did not spend time in restrictive housing (Ward &
Werlich, 2003), although once controlling for selection biases many of the significant differences
disappear (Mears & Bales, 2009; Smith, Gendreau, & Labrecque, 2015). Reentry success is also
challenging for gang members, regardless of whether they were place in restrictive housing
(Fleisher & Decker, 2001; Griffin, 2007). The combined effect of gang membership and restrictive

30
housing placement remains unclear. This is troubling due to the fact that 95% of current prisoners
are released from prisons (Hughes & Wilson, 2004), and our estimates suggest that around 14%
of those inmates are likely to be gang affiliated. Consequently, the ability of prison systems to
reduce their reliance on restrictive housing and successfully integrate returning ex-prisoners is at
least partially dependent upon responding to the unique needs of gang affiliates.
In summary, this study examined the relationship between gang affiliation and restrictive
housing in U.S. prisons. Relying on administrative records provided by prison systems, we
identified that over 200,000 inmates in U.S. prisons are affiliated with gangs, of whom about 12%
are housed in restrictive housing units on any given day—a proportion three times greater than
inmates who are not affiliated with gangs. We have provided initial explanations for the
overrepresentation of gang affiliates in restrictive housing and have identified the characteristics
of prison systems where gang affiliates are more likely to be overrepresented. Given the strong
link between gang affiliation and restrictive housing in U.S. prison systems, there is a continued
need to examine the factors that give rise to this relationship, as well as to understand the impact
of overhauling policies and practices (e.g., California, Colorado), will reshape institutional and
community corrections.

31
References
American Corrections Association. (2014). Standards committee meeting minutes. Presented at
the ACA Winter Conference, Tampa, FL. Retrieved from
http://www.aca.org/ACA_PROD_IMIS/Docs/Standards%20and%20Accreditation/SC%2
0Minutes_Tampa_Jan2014.pdf
Barrows, J., & Huff, C. R. (2009). Gangs and public policy: Constructing and deconstructing
gang databases. Criminology & Public Policy, 8(4), 675–703.
Baumgartel, S., Guilmette, C., Kalb, J., Li, D., Nuni, J., Porter, D. E., & Resnik, J. (2015). The
ASCA-Liman 2014 national survey of administrative segregation in prison. New Haven,
CT: The Liman Program, Yale Law School and the Association of State Correctional
Administrators.
Beck, A. J. (2015). Use of restrictive housing in U.S. prisons and jails, 2011–12 (No. NCJ
249209). Washington, DC: Bureau of Justice Statistics. Retrieved from
http://www.bjs.gov/content/pub/pdf/urhuspj1112.pdf
Burman, M. L. (2012). Resocializing and repairing homies within the Texas prison system: A
case study on security threat group management, administrative segregation, prison gang
renunciation and safety for all. The University of Texas at Austin, Austin, TX.
Butler, H. D., Griffin, O. H., & Johnson, W. W. (2013). What makes you the “worst of the
worst?” An examination of state policies defining supermaximum confinement. Criminal
Justice Policy Review, 24(6), 676–694. https://doi.org/10.1177/0887403412465715
Camp, G. M., & Camp, C. G. (1985). Prison gangs their extent, nature, and impact on prisons.
National Institute of Justice. Retrieved from
https://www.ncjrs.gov/pdffiles1/Digitization/99458NCJRS.pdf

32
Carson, E. A. (2016). Prisoners in 2015. Washington, DC: Bureau of Justice Statistics.
Chammah, M. (2016, January 7). How to get out of solitary — one step at a time. The Marshall
Project. Retrieved from https://www.themarshallproject.org/2016/01/07/how-to-get-out-
of-solitary-one-step-at-a-time
Cochran, J. C., Toman, E. L., Mears, D. P., & Bales, W. D. (2018). Solitary confinement as
punishment: Examining in-prison sanctioning disparities. Justice Quarterly, 35(3), 381–
41. https://doi.org/10.1080/07418825.2017.1308541
Crouch, B. M., & Marquart, J. W. (1989). An appeal to justice: Litigated reform of Texas
prisons. Austin, TX: University of Texas Press.
DiIulio, J. J. (1987). Governing prisons. New York, NY: Free Press.
Fischer, D. R. (2002). Arizona department of corrections: Security threat group (STG) program
evaluation. Washington, DC: National Institute of Justice.
Fleisher, M. S., & Decker, S. H. (2001). Going home, staying home. Corrections Management
Quarterly, 5(1), 65–77.
Fong, R. S. (1990). The organizational structure of prison gangs: A Texas case study. Federal
Probation, 54(1), 36–44.
Fong, R. S., & Buentello, S. (1991). The detection of prison gang development: An empirical
assessment. Federal Probation, 55(1), 66–70.
Frost, N. A., & Monteiro, C. E. (2016). Administrative segregation in U.S. prisons. In Maria
Garcia (Ed.), Restrictive housing in the U.S. issues, challenges, and future directions (pp.
1–47). Washington, DC: National Institute of Justice.

33
Gaes, G. G., Wallace, S., Gilman, E., Klein-Saffran, J., & Suppa, S. (2002). The influence of
prison gang affiliation on violence and other prison misconduct. The Prison Journal,
82(3), 359–385. https://doi.org/10.1177/003288550208200304
Garcia, Marie (Ed.). (2016). Restrictive housing in the U.S. issues, challenges, and future
directions. Washington, DC: National Institute of Justice.
Ghafar, M. (2017). Exiting solitary confinement: A survey of state correctional policies. UCLA
Law Review, 64, 508–547.
Griffin, M. L. (2007). Prison gang policy and recidivism: Short-term management benefits, long-
term consequences. Criminology & Public Policy, 6(2), 223–230.
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(4), 419–466.
Haney, C. (2018). The psychological effects of solitary confinement: A systematic critique.
Crime and Justice, 47(1), 365–416. https://doi.org/10.1086/696041
Haney, C., & Lynch, M. (1997). Regulating prisons of the future: A psychological analysis of
supermax and solitary confinement. NYU Rev. L. & Soc. Change, 23, 477–570.
Harrison, P. M., & Beck, A. J. (2003). Prisoners in 2002. Washington, DC: Bureau of Justice
Statistics. Retrieved from
http://www.ncjrs.gov/App/abstractdb/AbstractDBDetails.aspx?id=200248
Higgins, J. P., & Green, S. (Eds.). (2011). Cochrane handbook for systematic reviews of
interventions. London, UK: The Cochrane Collaboration. Retrieved from
www.handbook.cochrane.org
Hill, C. (2009). Gangs/security threat groups. Corrections Compendium, 34(1), 23–37.

34
Hinds, M., & Butler, J. (2015). Solitary confinement: Can the courts get inmates out of the hole.
Stanford Journal of Civil Rights & Civil Liberties, 11, 331–372.
Huebner, B. M. (2003). Administrative determinants of inmate violence: A multilevel analysis.
Journal of Criminal Justice, 31(2), 107–117. https://doi.org/10.1016/S0047-
2352(02)00218-0
Hughes, T., & Wilson, D. J. (2004). Bureau of Justice Statistics reentry trends in the U.S.:
Highlights. Retrieved January 11, 2019, from
https://www.bjs.gov/content/reentry/reentry.cfm
Hunt, G., Riegel, S., Morales, T., & Waldorf, D. (1993). Change in prison culture: Prison gangs
and the case of the pepsi generation. Social Problems, 40(3), 398–409.
Jacobs, J. B. (2001). Focusing on prison gangs. Corrections Management Quarterly, 5(1), vi–vii.
Kapoor, R. (2016). Mental health effects of restrictive housing. In Robert Trestman (Ed.),
Restrictive housing in the U.S. issues, challenges, and future directions (pp. 199–232).
Washington, DC: National Institute of Justice.
King, R. D. (1999). The rise and rise of supermax: An American solution in search of a problem?
Punishment & Society, 1(2), 163–186. https://doi.org/10.1177/14624749922227766
Klein, M. W., & Maxson, C. L. (2006). Street gang patterns and policies. New York, NY:
Oxford University Press.
Knox, G. W., & Tromanhauser, E. D. (1991). Gangs and their control in adult correctional
institutions. The Prison Journal, 71(2), 15–22.
https://doi.org/10.1177/003288559107100203
Kurki, L., & Morris, N. (2001). The purposes, practices, and problems of supermax prisons.
Crime and Justice, 28, 385–424.

35
Labrecque, R. M. (2015a). Effect of solitary confinement on institutional misconduct: A
longitudinal evaluation. University of Cincinnati, Cincinnati, OH. Retrieved from
https://www.ncjrs.gov/App/Publications/abstract.aspx?ID=271153
Labrecque, R. M. (2015b, October). Who ends up in administrative segregation? A meta-analytic
review. Presented at the National Institute of Justice Topical Working Group on the Use
of Administrative Segregation in the United States, Arlington, VA.
Labrecque, R. M. (2016). The use of administrative segregation and its function in the
institutional setting. In Maria Garcia (Ed.), Restrictive housing in the U.S. issues,
challenges, and future directions (pp. 117–164). Washington, DC: National Institute of
Justice.
Labrecque, R. M., & Smith, P. (2017). Reducing institutional disorder: Using the inmate risk
assessment for segregation placement to triage treatment services at the front end of
prison sentences. Crime & Delinquency, Online First.
https://doi.org/10.1177/0011128717748946
Lee, J., & Jacobs, R. (2012). Maps: Solitary confinement, state by state. Retrieved October 15,
2015, from http://www.motherjones.com/politics/2012/10/map-solitary-confinement-
states
Li, C. (2013). Little’s test of missing completely at random. The Stata Journal, 13(4), 795–809.
Lin, J. (2017). Program evaluation in the context of supervision regime change: Motivational
interviewing in Colorado. Justice Quarterly, 0(0), 1–24.
https://doi.org/10.1080/07418825.2017.1367027
Little, R. J. (1988). A test of missing completely at random for multivariate data with missing
values. Journal of the American Statistical Association, 83(404), 1198–1202.

36
Mears, D. P. (2005). Evaluating the effectiveness of supermax prisons. Washington, DC:
National Institute of Justice. Retrieved from
https://www.ncjrs.gov/app/Publications/abstract.aspx?ID=233437
Mears, D. P., & Bales, W. D. (2009). Supermax Incarceration and Recidivism*. Criminology,
47(4), 1131–1166. https://doi.org/10.1111/j.1745-9125.2009.00171.x
Mears, D. P., & Reisig, M. D. (2006). The theory and practice of supermax prisons. Punishment
& Society, 8(1), 33–57.
Mears, D. P., & Watson, J. (2006). Towards a fair and balanced assessment of supermax prisons.
Justice Quarterly, 23(2), 232–270.
Morgan, Robert D, Gendreau, Paul, Smith, Paula, Gray, Andrew L., Labrecque, Ryan M.,
MacLean, Nina, … Mill, Jeremy F. (2016). Quantitative syntheses of the effects of
administrative segregation on inmates’ well-being. Psychology, Public Policy, and Law.
http://dx.doi.org/10.1037/law0000089
Naday, A., Freilich, J. D., & Mellow, J. (2008). The elusive data on supermax confinement. The
Prison Journal, 88(1), 69–93.
National Institute of Corrections. (1997). Supermax housing: A survey of current practice.
Longmont, CO: National Institute of Corrections.
Obama, B. (2016). Why we must rethink solitary confinement. Washington Post. Retrieved from
https://www.washingtonpost.com/opinions/barack-obama-why-we-must-rethink-solitary-
confinement/2016/01/25/29a361f2-c384-11e5-8965-0607e0e265ce_story.html
Petersilia, J. (2003). When prisoners come home: Parole and prisoner reentry. New York, NY:
Oxford University Press.

37
Prendergast, A. (2014, August 21). After the murder of Tom Clements, can Colorado’s prison
system rehabilitate itself? Retrieved August 2, 2017, from
http://www.westword.com/news/after-the-murder-of-tom-clements-can-colorados-prison-
system-rehabilitate-itself-5125050
Pyrooz, D. C. (2016). Gang affiliation and restrictive housing in U.S. prisons. In Maria Garcia
(Ed.), Restrictive housing in the U.S. issues, challenges, and future directions (pp. 117–
164). Washington, DC: National Institute of Justice.
Pyrooz, D. C., Gartner, N., & Smith, M. M. (2017). Consequences of incarceration for gang
membership: A longitudinal study of serious offenders in Philadelphia and Phoenix.
Criminology, 55(2), 273–306.
Pyrooz, D. C., & Mitchell, M. M. (2019). The hardest time: Gang members in total institutions.
In B. M. Huebner & N. A. Frost (Eds.), The collateral consequences of sentencing and
punishment decisions. New York, NY: Routledge.
Raemisch, R. (2018, January 20). Why we ended long-term solitary confinement in Colorado.
The New York Times. Retrieved from
https://www.nytimes.com/2017/10/12/opinion/solitary-confinement-colorado-prison.html
Ralph, P. H., & Marquart, J. W. (1991). Gang violence in Texas prisons. The Prison Journal,
71(2), 38–49.
Reiter, K. A. (2016). 23/7: Pelican bay prison and the rise of long-term solitary. New Haven,
CT: Yale University Press.
Ruiz v. Estelle, 503 F.Supp. 1265 (United States District Court, S. D. Texas, Houston Division
1980).

38
Sheldon, R. G. (1991). A comparison of gang members and non-gang members in a prison
setting. The Prison Journal, 71(2), 50–60. https://doi.org/10.1177/003288559107100206
Skarbek, D. (2014). The social order of the underworld: How prison gangs govern the American
penal system. New York, NY: Oxford University Press.
Smith, P., Gendreau, P., & Labrecque, R. M. (2015). The impact of solitary confinement on
inmate behavior: A meta-analytic review. Presented at the North American Correctional
and Criminal Justice, Ottawa, Canada. Retrieved from
http://media.wix.com/ugd/7fc458_2efc5654e7ea4d27a9d45a64f331fec5.pdf
Steiner, B., Butler, H. D., & Ellison, J. M. (2014). Causes and correlates of prison inmate
misconduct: A systematic review of the evidence. Journal of Criminal Justice, 42(6),
462–470.
Steiner, B., & Cain, C. M. (2016). The relationship between inmate misconduct, institutional
violence, and administrative segregation: A systematic review of the evidence. In Maria
Garcia (Ed.), Restrictive housing in the U.S. issues, challenges, and future directions (pp.
165–197). Washington, DC: National Institute of Justice.
Tachiki, S. N. (1995). Indeterminate sentences in supermax prisons based upon alleged gang
affiliations: A reexamination of procedural protection and a proposal for greater
procedural requirements. California Law Review, 83(4), 1115–1149.
Toch, H. (2007). Sequestering gang members, burning witches, and subverting due process.
Criminal Justice and Behavior, 34(2), 274–288.
https://doi.org/10.1177/0093854806296663
Trulson, C. R., Marquart, J. W., & Kawucha, S. K. (2006). Gang suppression and institutional
control. Corrections Today, 68(2), 26–31.

39
Vigil, D. A. (2006). Classification and security threat group management. Corrections Today,
68(2), 32–34.
Ward, D. A., & Werlich, T. G. (2003). Alcatraz and Marion: Evaluating super-maximum
custody. Punishment & Society, 5(1), 53–75.
Winterdyk, J., & Ruddell, R. (2010). Managing prison gangs: Results from a survey of U.S.
prison systems. Journal of Criminal Justice, 38(4), 730–736.
https://doi.org/10.1016/j.jcrimjus.2010.04.047

40
Table 1
Prison System-Level Descriptive Statistics for the Study Variables
Mean/%
SD
Min
Max
n
Inmates in restrictive housing
6.2%
1.1%
14.7%
38
Inmates classified as gang affiliates
18.5%
0.1%
47.3%
38
Suspected gang affiliates
9.8%
0.0%
35.8%
31
Confirmed gang affiliates
10.3%
0.0%
41.0%
36
Gangs/STGs
204.3
(384.3)
4.0
1,534.0
38
History of gang emergence
Never
2.6%
0.0%
100.0%
38
1960s/1970s
18.4%
0.0%
100.0%
38
1980s
15.8%
0.0%
100.0%
38
1990s
50.0%
0.0%
100.0%
38
2000s
13.2%
0.0%
100.0%
38
Gang misconduct rates per 10,000
Homicide
0.2
(0.3)
0.0
1.1
26
Assaults on inmates
103.2
(105.2)
0.0
350.6
23
Assaults on staff
19.3
(29.1)
0.0
115.1
23
Pathways into and out of segregation
Gang status segregation
13.2%
0.0%
100.0%
38
Gang risk segregation
73.7%
0.0%
100.0%
38
Gang exit programs
36.8%
0.0%
100.0%
38
Prison population
25,210.5
(32753.6)
1,343.0
148,521.0
39
Bureau of Justice Statistics data
Turnover rate
47.8%
(19.2)
15.5%
119.7%
47
Overcrowding
97.06%
(16.83)
51.7%
145.1%
49
Homicide rates per 10,000
0.4
(0.3)
0.0
1.4
51
Suicide rates per 10,000
1.8
(0.9)
0.5
4.5
51
Assault rates per 10,000
101.1
(219.6)
7.2
1519.6
47
Abbreviations: SD = standard deviation; STGs = security threat groups.

41
Table 2
National-Level Estimates of Restrictive Housing and Gang Affiliation in U.S. State Prison Systems Including Valid and Imputed
Estimates
Responding Prison Systems
All Prison Systems
(Valid Data)
(Imputation Data)
n
Total Custodial
Population
Frequency
% Custodial
Population
N
Total Custodial
Populationa
Frequency
% Custodial
Population
Restrictive housing
All forms
38
967,217
61,706
6.4%
51
1,526,792
92,295
6.0%
Disciplinary
32
775,317
21,742
2.8%
51
1,526,792
30,595
2.0%
Protective
30
745,521
7,719
1.0%
51
1,526,792
10,157
0.7%
Administrative
30
702,328
12,515
1.8%
51
1,526,792
24,574
1.6%
Other
33
900,662
13,296
1.5%
51
1,526,792
20,535
1.3%
Gang affiliation
All forms
38
962,796
148,185
15.4%
51
1,526,792
213,215
14.0%
Suspected
31
712,522
56,111
7.9%
51
1,526,792
80,932
5.3%
Confirmed
36
934,841
88,357
9.5%
51
1,526,792
128,565
8.4%
Note. The unit of aggregation for all frequencies and proportions is the nation (i.e., summed across prison systems). All forms for restrictive housing and gang
affiliation are not equivalent to the sum of the subtypes because these estimates are reported separately by agencies. Total custodial population for all forms is not
identical because different agencies are included in the 38 responses.
aData source: Bureau of Justice Statistics.

42
Table 3
National- and Prison System-Level Relative Risk of Placement in Restrictive Housing for
Inmates and Gang Affiliates, by Reason for Placement
Percent in Restrictive
Housing
Mean of Relative Risk
Ratiose
n
Gang
Non-Gang
n
Mean
SD
National-Level Estimatesa
All gang affiliatesb
12%
---
3.1
(---)
Suspected gang affiliatesc
10%
---
2.6
(---)
Confirmed gang affiliatesd
13%
---
3.4
(---)
Non-gang affiliatesb
---
4%
---
(---)
Prison System-Level Estimates
All forms
34
15%
5%
34
4.0
(4.1)
Disciplinary
29
7%
2%
27
3.1
(2.0)
Protective
28
4%
1%
25
2.6
(2.8)
Administrative
28
2%
1%
22
6.3
(8.6)
Other
31
1%
1%
21
2.0
(3.8)
Suspected gang affiliatesc
All forms
27
14%
---
27
2.8
(2.5)
Disciplinary
22
8%
---
21
3.7
(3.4)
Protective
21
2%
---
19
2.6
(3.5)
Administrative
22
3%
---
16
3.8
(2.8)
Other
23
2%
---
17
2.3
(4.1)
Confirmed gang affiliatesd
All forms
31
14%
---
31
3.7
(3.8)
Disciplinary
27
7%
---
25
2.7
(1.6)
Protective
26
2%
---
23
3.1
(3.3)
Administrative
26
4%
---
21
7.3
(9.7)
Other
28
1%
---
19
2.3
(4.4)
Note. Sub-categories do not combine to equal “all forms” due to missing data. Response n’s between percent in
restrictive housing and relative risk ratios are not identical because relative risk ratios cannot be computed for states
with prevalence estimates of 0%.
Abbreviation: SD = standard deviation.
aUnit of aggregation is the country, where proportions and risks ratios are computed across the aggregated values for
the 39 prison systems.
bDenominator for percentage = all gang members or all non-gang members.
cDenominator for percentage = all suspected gang members.
dDenominator for percentage = all confirmed gang members.
eMean of relative risk ratios = (total gang in restrictive housing/total gang)/(total non-gang in restrictive
housing/total non-gang).

43
Table 4
Bivariate OLS Regression Models at the Prison System-Level for the Relative Risk of Placement
into All Forms of Restrictive Housing for All Gang Affiliates
Gang Member Risk of
Restrictive Housing Placement
Relative Risk Ratio
n
b
(SE)

t-value
Model
Intercept
Prevalence of gang affiliates
Suspected
27
4.89
(6.65)
0.55
0.74
3.12*
Confirmed
31
17.72*
(4.92)
1.82
3.60
0.74
Prison populationa
34
0.07*
(0.02)
2.23
3.23
2.43*
Population in restrictive housinga
34
0.29
(0.32)
0.66
0.92
3.56*
Gangs/STGsb
34
0.22
(0.23)
0.68
0.95
3.63*
History of gang emergencec
1960s/1970s
34
-1.86
(1.84)
-0.72
-1.01
4.30*
1980s
34
4.92*
(1.82)
1.77
2.71
3.25*
1990s
34
-0.37
(1.42)
-0.19
-0.26
4.16*
2000s
34
-1.09
(2.00)
-0.39
-0.54
4.14*
Gang misconduct rates per 10,000
Homicide
26
2.17
(2.03)
0.75
1.07
3.31*
Assaults on inmates
23
0.01#
(0.01)
1.43
1.90
2.38*
Assaults on staff
23
0.00
(0.03)
0.12
0.15
3.75*
Official data estimatesd
Homicide rates per 10,000
34
-0.10
(2.37)
-0.30
-0.42
4.33*
Suicide rates per 10,000
34
0.01
(0.82)
0.01
0.01
3.95*
Assaults rates per 10,000
31
-0.00
(0.00)
-0.44
-0.55
4.16*
Turnover rate
31
0.01
(0.03)
0.17
0.25
3.29#
Overcrowding
32
0.15
(0.04)
0.26
0.43
2.15
Pathways into and out of segregationc
Gang status segregation
34
2.78
(1.95)
1.00
1.42
3.57*
Gang risk segregation
34
-0.43
(1.56)
-0.20
-0.27
4.28*
Gang exit programs
34
1.04
(1.46)
0.51
0.71
3.58*
Note. Standardized and unstandardized coefficients are presented.
Abbreviations. STGs = security threat groups.
aOne unit increase = 1,000 inmates.
bOne unit increase = 100 gangs.
cDichotomously coded.
dData source: Bureau of Justice Statistics.
# p < 0.10, * p < 0.05

44
Figure 1
Responses to survey solicitation

45
Figure 2
Relative Risk Ratios for Placement into Restrictive Housing across Prison Systems for all Reasons and by Reason for Placement
Note. Each dot represents a relative risk ratio for a blinded prison system. Risk ratios are the proportion of gang affiliates in restrictive housing divided by the
proportion of non-gang inmates in restrictive housing. Solid lines = even risk of 1; dotted lines = mean risk ratios. Risk ratios exceeding 10 are labeled.
Administrative (n = 22)Protective (n = 25)All (n = 34)
Jurisdictions
Discipline (n = 27)
12.8
15.8
19.2
4.0
0
5
10
15
20
25
30
0 1 2 3 4 5 6 7 8 9 10
Risk Ratios
10.4
3.1
0 1 2 3 4 5 6 7 8 9 10
Risk Ratios
10.5
10.6
42.4
6.3
0 1 2 3 4 5 6 7 8 9 10
Risk Ratios
2.6
12.2
0 1 2 3 4 5 6 7 8 9 10
Risk Ratios