THESIS

MIU "Uru- p.-

lllllllllllllllllllllllllllllllllllllllll

3 1293 01780 674

LIBRARY
Michigan State
University

 

 

 

 

This is to certify that the
thesis entitled
Administrative Determinants of Inmate Violence:

A Multi—Level Analysis

presented by
Beth M. Huebner

has been accepted towards fulfillment
of the requirements for

M . S . degree in .CniminaLlustice

 

”a?“ D. 72%

Major professor

Date_M޴_1_._199.9___

0-7639 MS U is an Affirmative Action/Equal Opportunity Institution

 

PLACE IN RETURN Box to remove this checkout from your record.
To AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

DATE DUE

DATE DUE

DATE DUE

 

 

MAR atmlfllll 1

 

 

 

 

 

 

 

 

 

 

 

 

 

 

2/05 p:lCIRC/DaieDuemdd-p 1

 

 

ADMINISTRATIVE DETERMINANTS OF INMATE VIOLENCE:
A MULITI-LEVEL ANALYSIS
BY

Beth Marie Huebner

A THESIS
Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of

MASTER OF SCIENCE
School of Criminal Justice

1999

ABSTRACT

ADMINISTRATIVE DETERMINANTS OF INMATE VIOLENCE:
A MULTI-LEVEL ANALYSIS

BY

Beth Marie Huebner

This study examined the effect of remunerative and coercive
controls on inmate assault, net of traditional controls. The
sample included 4,168 male inmates nested within 185 state
correctional facilities. The results suggest a complex
relationship between remunerative controls and inmate assault.
Remunerative controls were not significant predictors of inmate-
on—inmate assault; however, prisoners involved in work programs
were significantly less likely to assault staff, net of control
variables. A significant relationship was not found between

coercive control and inmate assault.

TABLE OF CONTENTS

LIST OF TABLES

 

INTRODUCTION

 

REVIEW OF THE LITERATURE

 

Traditional Models
Administrative-Control

 

RESEARCH OBJECTIVES

 

METHODS AND DATA
Sampling

 

Dependent Variables

 

Contextual Variables

 

 

Independent Variables
Prison-Level Controls

 

ANALYTIC TECHNIQUE

 

RESULTS
Preliminary Models

 

Contextual Models

 

Administrative—Control and Prison Violence

 

 

Subculture of Violence Thesis

DISCUSSION

 

APPENDIX A DESCRIPTION OF VARIABLES:

 

APPENDIX B DESCRIPTIVE STATISTICS:

 

REFERENCES

 

iii

iv

N

10
11
12
12
14
18
22
23
23
26
29

31

LIST OF TABLES

Table 1 - Fixed Effects of Inmate-Level Variables on Dependent
Variables 16

 

Table 2 - The Effects of Prison—Level Variables on Inmate-on—
Inmate Assault 19

 

Table 3 - The Effects of Prison-Level Variables on Inmate-on
Staff Assault 21

 

iv

Introduction

What are the determinants of inmate misconduct? Researchers
have sought to answer this question for over a half—century and
three :models have emerged; deprivation, importation, and
administrative-control. Traditionally, researchers have focused
on the importation and deprivation models; however, recent
studies have included variables under the heading of
“administrative-control.” The deprivation model posits that
inmate behavior (e.g. misconduct) is a product of the prison
environment (Clemmer, 1938; Sykes, 1958). The importation
perspective asserts that an inmate's pre-prison socialization is
important and influences the inmate social structure and inmate
adaptation. to the jprison environment (Irwin, 1980; Irwin and
Cressey, 1962). Administrative-control theory departs from the
deprivation and importation perspectives and posits that prison
management is an important determinant of inmate misconduct
(DiIulio, 1987). Administrative-control theorists have
concentrated their attention on management's use of various
control mechanisms, including discipline and programming to curb
inmate misconduct.

While numerous studies have assessed the validity of the
importation and deprivation models, few researchers have tested
the hypotheses deduced from the administrative-control model.
The goal of this study is to provide such an‘assessment while
controlling for the variables traditionally found under the

importation and deprivation headings. Inmate assaults on staff

and inmates will be used as the outcome measures. The analysis
will use a subset of the 1991 Survey of Inmates of State

Correctional Facilities (U.S. Department of Justice, 1993).

Review of the Literature

Traditional Models

The deprivation model rests on the assumption that inmate
behavior is a function of the prison environment. Donald Clemmer
(1938), a pioneer in the field of prison research, argued that
inmates assimilate into an inmate social structure and culture
that condones deviant behavior. Gresham Sykes (1958), whose work
provided an extension of Clemmer's ideas argued that there are
certain “pains of imprisonment" inherent in prison, and these
pains define social group formation and behavior within prison
society. Pains of imprisonment include loss of liberty, loss of
material possessions, and lack of autonomy (Sykes, 1958).

Early studies of the deprivation model were based primarily
on examinations of inmate coping’ mechanisms (Flanagan, 1980;
Garabedian, 1963; Wheeler, 1961). Contemporary researchers have
focused on examination of absolute deprivation. Absolute
deprivation refers to the fixed institutional environment and has
been tested in relation to overcrowding, custody level, time
outside of cell, visiting patterns, and size of facility (Cao,
Zhao and VanDine, 1997; Ellis, Grasmick and Gilman, 1974;

Goetting and Howsen, 1986; Ruback and Carr, 1993). Of the

absolute deprivation variables, prison crowding has been tested
extensively; however, the direct relationship between absolute
deprivation and misconduct is not clear (see Ellis, 1984).
Importation theorists challenged the deprivation model and
proposed an alternative model based on pre-institutional

characteristics, culture, and environment (Irwin and Cressy,

1962) . Irwin (1980) argued that inmates did not enter prison
tabula rasa, instead inmates bring roles and cultural
expectations into prison. Pre-prison experiences and

socialization also influences how one copes and forms social
groups in prison (Irwin, 1980). The importation perspective has
also been extended to include the effects of macro-social events
and movements on behavior (Jacobs, 1977).

Tests of the importation perspective have provided mixed
results. Importation theorists have tested demographic variables
such as age, gender, military involvement, prior arrests,
employment, and marital status as predictors of misconduct (Cao
et al., 1997; Ellis et al., 1974; Thomas, 1977). Importation
theorists have also argued that inmate behaviors are a result of
cultural histories and traits (Akers, Hayner and Gruninger, 1977;
Harer and Steffensmeier, 1996). Harer and Steffensmeier (1996)
verified the “subculture of violence” thesis by establishing a
significant relationship between race and type of prison
misconduct. Black inmates were more likely to engage in violent
behavior and whites were more likely to engage in misconduct
related to alcohol and drugs (Harer and Steffensmeier, 1996).

These findings mirror misconduct rates typically found for these

groups in the general population (Wallace and Bachman, 1991;
Williams and Flewelling, 1988). Other studies have found little
evidence supporting the importation perspective (Gaes and
McGuire, 1985; Light, 1990).

Both the importation and deprivation perspectives have made
significant contributions to the study of prison misconduct. An
integrated model has also been used to predict inmate behaviors
and attitudes (Akers et al., 1977; Goetting and Howsen, 1986;
Thomas, 1977). Although traditional models have been moderately
successful in predicting inmate misconduct, there are limitations
associated with both the importation and deprivation models. The
importation. perspective has focused. solely’ on individual
characteristics, but individual-level research does not provide
guidance for prison managers in preventing prison misconduct when
managers rarely have control of the type of inmates they house
(McCorkle, Miethe and Drass, 1995:321). The deprivation
perspective includes both organizational and individual
determinants of prison misconduct; however, the deprivation
perspective ignores the role of management in mediating or
exacerbating levels of deprivation.

The goal of this paper is not to discount the validity of
the traditional models of prison misconduct. Because multi-level
models are being tested, it is inevitable that specific theories
within the deprivation and importation models will be tested
(e.g., subculture of violence). However, the primary goal of
this study is to examine what role administrative-control

variables play in predicting prison misconduct while controlling

for traditional variables.

Administrative-Control

Contemporary administrative—control scholars argue that the
deprivation and importation. models have ignored an important
element of prison misconduct and contend that “administration"
should be included as a determinant of inmate misconduct.
DiIulio (1987) studied three state prison systems: Texas,
California, and Michigan. From his studies, DiIulio (1987)
argued that prisons managed within a formal organizational
structure with a strict division of labor, detailed rules and
routines, and a consummate leader at the top of the organization
will have significantly less prison disorder. Useem and Kimball
(1989) also contend that control is an essential element of
prison. management; however, they argue that the relationship
between control and misconduct is more complex and that
“administrative breakdown” is the cause of collective misconduct.
Administrative breakdown occurs when the organizational structure
of the prison erodes or is otherwise weakened, security measures
are reduced, prison conditions and services deteriorate, and
inmate deprivation increases above tolerable levels (Useem and
Kimball, 1989). Collective prison misconduct is a result of
gradual organizational breakdown (see also Useem, Camp and Camp,
1996).

Administrative-control theory rests on the assumption that
control is a necessary element in corrections management;

however, there has been little consensus on the most appropriate

type of control (DiIulio, 1987; Reisig, 1998; Useem and Kimball,
1989). Two specific forms of control have been advanced.
Coercive control rests on the threat or application of physical
sanctions, the generation of frustration through control, or
restriction on achievement of personal needs through force.
Remunerative control is based on the allocation of material
resources and rewards (Etzioni, 1961:8-22). Remunerative
controls function as incentives in organizations and engender
commitment to cuganizational rules and goals; whereas, coercive
controls frustrate and alienate organizational members (Colvin,
1992; Etzioni, 1964). Research guided by the administrative-
control model has confirmed the importance of incentives.
Correctional institutions that maintained a balance between
coercive and remunerative controls were less likely to have high
incidence of misconduct when compared to institutions that used
coercion as the sole means of control (Colvin, 1992; Reisig,
1998; Wright 1994).

Research on the relationship between prison programming and
inmate behavior has been widespread. The majority of researchers
have focused their research on prison programming as a predictor
of recidivism (e.g. Matthews and Pitts, 1998; Turner and
Petersilia, 1996). Researchers have also centered evaluations on
specific types of programming such as HIV education, anger
management, or residential drug treatment (Accordino and Guerney,
1998; Eisenberg and Fabelo, 1996; Harrison et al., 1998; Matthews
and Pitts; 1998). This body of literature has provided

contradictory evidence on the effects of prison programming on

post-incarceration behavior (Eisenberg and Fabelo, 1996; Inciardi
et al., 1997; Ward and Baldwin, 1997). Few researchers have
examined the effects of prison programming on institutional
behavior. Research conducted in this area has indicated that
inmates involved in educational and vocational programming or
work are less likely to assault prison staff or inmates (Gaes and

McGuire, 1985; McCorkle et al., 1995).

Research Objectives

Due to the adolescence of the administrative-control model,
the perspective has not been subject to rigorous empirical
assessment. Existing studies of administrative-control have been
limited by two factors. First, the majority of research
conducted within this perspective has been qualitative and based
on case studies (DiIulio, 1987; Useem and Kimball, 1989; Useem et
al., 1996). Second, the scope of the current work has been very
limited. Many studies have also been based on only a small
sample of correctional facilities (e.g., Reisig, 1998).
Specifically, studies have not been performed to ascertain the
relationship between specific control (i.e., solitary
confinement) and incentive (i.e., work programs) measures on
prison misconduct. This study hopes to help fill this void left
in the administrative-control literature.

This study has been designed to overcome many of the
limitations associated with previous studies within the

administrative-control model. Unlike other studies performed,

this study will use a nation-wide sample of inmates to assess
simultaneously the effects of administrative-control variables on
prison misconduct, while controlling for traditional variables

such as age, race, and criminal history.

Methods and Data

Sampling

This study used a subset of the data provided in the 1991
Survey of Inmates of State Correctional Facilities (U.S.
Department of Justice, 1993). The sample data includes
information collected through personal interviews from 13,986
inmates within 272 state correctional facilities during June,
July, and August 1991.

Using a two-stage selection design, the sample was selected
from 1,239 institutions with 613,894 male and 34,538 female
inmates. The first sampling stage included selecting
institutions tn! population, census region, facility' type, and
security level. A systematic sample was then selected from each
sampling frame based on probabilities proportional to the size of
the prison. In the second stage, a systematic sample of inmates
was taken from each prison based on the daily roster of inmates.
Nine hundred inmates refused or were unable to be interviewed
resulting in a sample of 13,986 inmates. The overall non-
interview rate was 6%. .All inmates housed in community or
“unclassified” prisons were also deleted from the sample. Using

listwise deletion, a subset of data was selected consisting of

13,370 male and female inmates nested within 247 correctional
facilities. A representative subset of the initial data set was
then selected from inmates with complete dependent variable
information and included 5,026 inmates and 224 puisons. Women
were then deleted from the data-set resulting in a final sample
of 4,168 male inmates nested within 185 state correctional
institutions.

Prison-level data were constructed by aggregating individual
inmate data by institution. It can be assumed that the
individual inmate data are representative by prison because of
the detailed sample design employed by the U.S. Department of
Justice (1993). Inmates interviewed at each of the prisons
selected were assumed to be representative of that prisons total
population. Representativeness was assured (due to the complex
weighting and stratified sampling technique employed by the

Bureau of Justice Statistics.

Dependent Variables

Two dependent measures of inmate assault will be included in
the analysis. For each of the measures the inmate was asked,
“since your admission, how many times have you been found guilty
of”, (1) physical assault on staff or (2) physical assault on
inmate. Prior studies have relied only on dichotomous dependant
variables (e.g., Light, 1990). Count data is preferred to
categorical data because it allows the researcher to model how
many offenses will occur opposed to simply if misconduct occurs

(Gardner et al., 1995).

/
/

Contextual Variables

Three remunerative variables will be included as contextual
variables. The remunerative control variables are designed to
measure positive incentives provided by prison administration.
All remunerative control variables represent the percentage of
the prison population that have received remunerative controls at
some point during their current incarceration period.
Remunerative control measures include percent of inmates that
work outside the prison facility, work inside the prison and have
a paid work assignment.

Two coercive control variables will be included as

contextual variables to measure institutional punishments. A

[I 1:. I? u
hwtcoerc1ve control scale was constructed to measure loss of work or

‘d

f general privileges as a result of prison misconduct. The factor

included two measures: Percent of total inmate population that

(1) lost a work assignment or (2) lost general privileges as a

response to the most recent rule infraction (eigenvalue = 1.32;
factor loadings >0.70). A measure of solitary confinement will
also be included as a contextual variable. The solitary

confinement variable represents the percent of the total inmate
population that received solitary confinement as ea disciplinary

response to the most recent rule infraction.

10

Independent Variables

This study includes several measures of individual inmate
attributes. These variables represent individual pre-prison
characteristics and will be used as control variables.
Descriptions of all variables are provided in Appendix A and
descriptive statistics are included in Appendix B. Three dummy
coded measures of race were included: white, African American, or
other (Asian, Pacific Islander, Eskimo, or other). A separate

dummy measure of Hispanic ethnicity was also included.

Citizenship was also dichotomized into citizen (reference
category) and non-citizen. Age was included as measured as the
inmate's age in years (mean = 30.51; S.D. = 8.15) . Marital

status was dichotomized into married (reference category) and not
married. A measure of length of incarceration was included and

calculated based on number of years inmate had served for their

current sentence at the time of data collection (mean = 2.81;
S.D. = 2.56) .
Two scales were included as control measures. Gang

membership is an additive scale constructed using four survey
items: “Did you ever belong to a group that” (1) had members from
the same area, (2) have a turf or territory, (3) have a formal
membership, and (4) have a known leader? (Cronbach's alpha =
0.79). A four—item factor score was also developed to measure
criminal history. The factor score included four items: total
sentence in years, currently serving for a violent crime, age at

first arrest, and number of prior arrests (eigenvalue = 1.19).

11

Prison—level Controls

Three prison~level control variables were included. Census
region and prison security levels were included as dummy—
variables. Census region reflected whether prison was located in
the South census region (reference category) or otherwise. Level
of prison security was also dichotomized into Maximum security
(reference category) and other security level. A measure of
prison population was included as final prison count at the date

of data collection (mean = 1503.04; S.D.= 1280.40).

Analytic Technique

Hierarchical linear modeling (HLM) was used as the primary
analytic technique (Bryk and Raudenbush, 1992). HUM is the most
appropriate technique for this analysis because it allows the
individual inmate data to be nested within prisons while studying
variation in incidence of inmate assault. HLM has many
advantages over traditional multivariate models. Typically
researchers were forced to choose between ignoring the multilevel
nature of the data or by aggregating all individual—level data to
the organizational level. These traditional modeling techniques
increase validity concerns. Ignoring the natural clustering in
data significantly increases the Type I error rate and
aggregating individual-level data discards important within group
variation (Kreft and De Ieeuw, 1998:10). HLM resolves many of
the problems with multi—level modeling by allowing the researcher

to model data within natural clustering, while simultaneously

12

modeling both within and between group variance.

In conventional HLM methodology Ordinary Least Squares
regression is used; however, OLS assumes that random errors are
independent, normally distributed, and have constant variance
(Bryk and Raudenbush, 1992:15). Because a majority of inmates
have not committed an assault, the dependent variables are
positively skewed. Significant overdispersion of time dependent
variables violates the OLS assumptions. An overdispersed
Poission model will be used instead of the default OLS model in
this analysis to compensate for overdispersion. Overdispersed
Possion models calculate a factor to correct the inferential
statistics; therefore, statistical reliability is not compromised
(Gardner et al., 1995).

Four general models will be estimated for each dependent
variable in this analysis. The One-way ANOVA model is the first
model estimated. The ANOVA model provides an estimate of within
and between prison-variance and a reliability estimate of the
sample mean. The second model, the fixed effects model, attempts
to measure variation in the dependent variable as a function of
inmate-level variables. The intention here was to explore how
inmate assault varies with. each, inmate-level predictor; The
third model, the means as outcomes model, assesses which
individual level slopes vary by mean prison incidence of assault.
The inmate and jprison-level. models are then combined into a
single hierarchical model. The final model evaluates the
probability that individual incidents of assault vary as function

of individual and contextual variables.

13

Results

Preliminary Models

Four models were constructed for each dependent measure to
determine the relationship between individual and contextual
variables on inmate assault. The modeling strategy used here is
consistent with established HLM methodology (Bryk and Raudenbush,
1992).

The One-way ANOVA model with random effects is the first
model constructed. This variance component model is used as a
baseline for future modeling and for the estimation of variance
that lies between and within prisons (Kreft and De Leeuw,
1998:63). An intra-class correlation coefficient is also
obtained from this model and illustrates the proportion of
variance in level two units (e.g. prisons) (Kreft and De Leeuw,
1998).

An ANOVA model was constructed for both the inmate and staff
assault variables. The intra-class coefficient from staff
assault was 0.33; therefore, 33% of the variance associated with
assault on staff lies between prisons. At 12%, the intra-class
coefficient for inmate assault was much lower than for staff
assault. This model also produces a reliability estimate of the
sample mean. The sample mean for staff assault is a moderately
reliable measure (>.57) of the true mean of staff assault. The
sample mean for inmate assault is also moderately reliable (>.49)
measure of the true mean of inmate assault.

The second model specified is the random coefficient model.

14

This .model tests if the relationship> between individual-level
variables and inmate and staff assault varies across the
population of prisons. In random coefficient models, slopes are
allowed to vary by institution. If variance in the slopes is
significantly different than zero, it is assumed that the
variation in the individual-level slopes is a result of
contextual effects (Kreft and De Leeuw, 1998:43). Each of the
individual-level variables is centered around the group (prison)
mean. Traditional Ordinary Least Squares regression was used in
place of the Overdispersed Poisson model due to data limitations
in the random coefficient models; however, the temporary change
will not negatively effect further models. Two random-
coefficient models were estimated, one for each dependant
variable.

The first random-coefficient model estimates the effect of
individual-level variables on inmate assault. Gang membership
and criminal history were the only variables found to vary
significantly by institution. Together the group of inmate-level
variables explains about 9% of total inmate-level variance.

The effect of inmate-level variables (n1 staff assault was
the second random-coefficient model estimated. The inclusion of

the inmate-level variables explains about 5% of the inmate-level

variance above the initial. model. Gang' membership, criminal
history, and African American significantly varied by
institution. The coefficients for age and education indicate a

negative relationship between age, education and staff assault

across prisons.

15

 

 

 

 

Table 1. Fixed Effects of Inmate-Level Variables on Dependent
Variables (N = 4,168)
Inmate-on-Staff Inmate-on-Inmate
Assault Assault

Independent Variable B SE b SE
Intercept —2.06*** 0.09 -O.51*** .05
African American 0.25* 0.11 0.18** .07
Hispanic 0.09 0.13 0.02 .09
Other Race 0.32 0.25 0.36* .17
Citizen -0.32 0.31 -0.38 .26
Age -0.03*** 0.01 -0.03*** .01
Education -0.06** 0.02 —0.05** .02
Married -0.11 0.15 -0.03 .10
Years Incarcerated -0.03 0.02 0.01 .01
Gang Involvement 0.11** 0.04 0.18*** .03
Criminal History 0.09** 0.03 0.07** .02
# of Cases Trimmed

to within 99th

Percentile 3 9
Within Prison

Variance Explained 5% 9%
*p<0.05 **p<0.01 ***p<0.001 (two-tailed tests)

16

All independent, individual-level variables will be
considered “fixed” in the final models. Because the individual-
level variables are used as controls in the final models, there
is little utility in modeling the variation in these variables
between prisons.

The third model is the fixed-effects model. This model is
similar to the random-coefficient model; however, in this model
individual variables are considered fixed. In the fixed-effects
models, individual-level slopes are constrained to be constant
across institutions (Table 1). Individual-level variables are
modeled against an estimate of the average intercept and slope
across all institutions.

This model indicates that African American gang members with
a high criminal history score are more likely to assault both
inmates and staff. Educated and older inmates were found to be
less likely to assault staff and inmates. Inmates of other races
are more likely to assault inmates, but no relationship between
other race and staff assault was found.

Finally, a "means as outcomes" model was constructed. In
this model (not shown), both the intercept and slope for inmate
and staff are modeled as a function of the prison-level controls
and remunerative and coercive control measures. To aid
interpretation, each level-2 variable is centered around its
grand mean.

The addition. of prison-level controls improves the
explanation of variation in inmate assault on inmate; however,

the inclusion of remunerative and coercive controls adds little

17

to the model. Maximum security, south census region, and inmate
population were all significant predictors of inmate assault. A
significant relationship> between. the remunerative and. coercive
control variables and inmate on inmate assault was not found.
Remunerative controls add significantly to the assault on
staff model; whereas, prison-level controls add little to the
model. Maximum security was the only prison-level control
variable that was a significant predictor of staff assault.

“M‘fi

Remunerative control measures were found to have a negative

.....I __ . .c

relationship to staff assault. Institutions with higher
percentages of inmates employed both within and outside the

facility had lower mean levels of“ staff _assault. Coercive

,.. 4.- o-t... . H-.. , ~..y -..a.-.....
- _.,.4-,.«. -J - "'

control measures had no effect on staff assault.

Contextual Models

Building on the preliminary models, two full hierarchical
models that include both contextual and individual-level
variables were estimated. The first model tests the relationship
between remunerative and coercive controls on prison assault, net
of individual controls. The second model estimated the
probability that individual incidents of assault vary as function
of individual and contextual variables. This model seeks to
examine. what relationship exists between remunerative and
coercive control and incidence of assault net of inmate and

prison-level controls.

18

Table 3. The Effects of Prison-Level Variables on Inmate-on-
Inmate Assault (N = 185)

 

 

 

Model 1 Model 2

Independent Variable B SE b SE

Intercept —0.51*** 0.05 -0.54*** 0.05
Remunerative Controls

Paid for Work -0.01*** 0.00 -0.00 0.00

Work Outside —0.02** 0.01 -0.01 0.01

Work Inside -0.01 0.00 -0.00 0.00
Coercive Controls

Solitary —0.00 0.00 -0.00 0.00

Coercive Control 0.03 0.05 0.06 0.05
Control Variables

Population 0.00*** 0.00

Southern Facility 0.31** 0.14

Maximum Security 0.52*** 0.10
Between Prison
Variance Explained 6% 30%

 

the: Estimates of prison-level coefficients control for African
American, Hispanic, other race, citizen, age, education, married,
sentence length, gang involvement, and criminal history.

*p<0.10 **p<0.05 ***p<0.01 (two-tailed tests)

19

Model 1 demonstrates a strong institutional effect on
inmate—on—inmate assault (Table 2). Both inmates employed
outside of the prison and inmates paid for work were
significantly less likely to assault inmates. Coercive factors
had no effect on inmate assault. The addition of the
remunerative and coercive control variables explains about 6% of
the total inmate—level variance.

When contextual variables were added in model 2 the
remunerative controls were no longer significant predictors of
inmate assault (Table 2). Not surprisingly, all contextual
variables added to the final model were significant predictors of
inmate assault. Inclusion of prison-level controls also adds
significantly to the model, explaining about 30% of the prison—
level variance. It appears from the models that prison
population, maximum-security facility, and South census region
mediates the effect of remunerative controls on inmate-on—inmate
assault.

Remunerative controls have constant, significant effects on
inmate-on-staff assault when compared to inmate-on-inmate assault
(Model 1 in Table 3). In the preliminary model staff assault
model, inmates employed both inside and outside of the facility
were significantly less likely to assault staff. No effect was
found for coercive control measures as predictors of staff
assault. Inclusion of the remunerative and coercive control

variables explains about 7% of the prison-level variance.

20

Table 3. The Effects of Prison—Level Variables on Inmate-on-
Staff Assault (N = 185)

 

 

 

Model 1 Model 2
Independent Variable B SE b SE
Intercept -2.08*** 0.09 -2.11*** 0.09
Remunerative
Controls
Paid for Work -0.00 0.00 -0.00 0.00
Work Outside -0.03** 0.01 -0.02* 0.01
Work Inside -0.02*** 0.01 -0.01** 0.00
Coercive Controls
Solitary 0.00 0.00 0.00 0.00
Coercive Control 0.05 0.10 0.05 0.10
Control Variables
Population -0.00 0.00
Southern Facility 0.20 0.25
Maximum Security 0.79*** 0.19
Between Prison
Variance Explained 7% 18%

 

the: Estimates of prison-level coefficients control for
African American, Hispanic, other race, citizen, age, education,
married, sentence length, gang involvement, and criminal history.

*p<0.10 **p<0.05 ***p<0.01 (two—tailed tests)

21

In the final model inmates employed within and outside the

.. q, . . . - -b
flurqwmu . ~.-.a-.. ' ' r- I Jbl-i .“.A “'"

pri sons we re sill-((1....,....l.§.ss.._...like lywto....~assault-mstwaf f, ., ..desp.i 12.53. the

inclusion of prison-level controls (Model 2 in Table 3).
Maximum-security facility‘ was the only significant contextual
predictor of staff assault. Eighteen percent of the total prison-

variance is explained by the remunerative, coercive, and prison-

level control variables.

Administrative-Control and Prison Violence
The results from the full model indicate that there is a
significant relationship between remunerative controls and inmate
and staff assault; however, the relationship is more complex than
originally anticipated. It appears that not all forms of assault
are equivalent. Incidence of assault was determined by a variety
of factors. Determinants of assault also vary with the target of
the assault. This is evidenced from the findings from the full
HLM models. In the inmate assault models, both paid work and
work outside the facility were significant predictors of inmate
assault. However, when prison-level controls were added to the
model the effects of the remunerative control variables were
nullified. The staff assault models were not effected as
significantly by the prison-level controls. Emisoners employed
within or outside the prison were significantly less likely to
assault staff despite the inclusion of individual and prison-
level controls.
As important as the significance of remunerative control

variables on assault, is the lack of significance for the

22

coerciwe control variables. Traditional administrative-control
theorists have argued that control is essential for proper
corrections.:management (DiIulio, 1987; Kantrowitz, 1996). It
appears from these models that coercive controls do little to
mediate or exacerbate individual incidents of assault. Further
research should. be conducted to determine the true power of

coercive controls within correctional facilities.

Subculture of Violence Thesis

The relationship between subculture of violence and inmate
violence is also more complex than originally anticipated. In
the fixed-effects models African American was a significant
predictor of inmate-on-inmate and staff assault; however, in the
final model of inmate on staff assault African American was no
longer a significant predictor. African American remains
significant in all inmate-on-inmate assault models. In the same
light, South census region is a significant predictor of assault
in all inmate-on-inmate assault models, but has no effect in the
inmate on staff assault models. It is evident from the models
that African Americans have an increased inclination toward
inmate violence; however, when remunerative and prison-level
controls are included in the model the relationship between race

and inmate assault changes.

Discussion

Multi-level analyses of inmate violence have been rare in

23

corrections research. More exceptional is tflua examination and

development of administrative-control constructs as predictors of

prison misconduct. This study attempted to address these
limitations and to expand the knowledge within the
administrative-control perspective . The goal of this analysis

was not just to examine the role of incentive and punishments on
inmate assault, but to examine the role of remunerative and
coercive controls net of traditional variables. This study
provides additional support for traditional models of prison
misconduct; however, it also provides evidence that remunerative
controls are able to mitigate incidence of assault net of
traditional control variables.

Although the findings presented provide intriguing results,
two caveats should be noted. First, it difficult to evaluate the
validity of self-report data provided by inmates. Although the
accuracy of self-report data is always in question, researchers
have suggested that higher levels of rule infraction are reported
in self-reported studies when compared to official prison records
(Hewitt et al., 1984). From this analysis, it can be assumed
that self-report data may provide an expanded understanding of
inmate assault then would have been discovered from an analysis
of official records. In the same light, it is difficult to
ascertain biases that may effect ticketing for rule infractions
(Harer and Steffensmeier, 1996:353).

Because of the nature of the data, it was also impossible to
evaluate the relationship between specific types or length of

work programs on assault. It may be that certain forms of work

24

programs are more meaningful for prisoners or that long work
hours prevent assault; however, it was impossible to ascertain
the relative importance of specific forms of programming on
inmate and staff assault. Further research should be conducted
to examine specific forms of remunerative controls on inmate
assault and other forms of misconduct.

Despite the relative limitations of the data, it appears
that remunerative controls can mediate the effects of traditional
determinants of inmate assault. It remains unclear; however, why
remunerative control measures had varying effects (n1 different
forms of assault. Etzioni (1961) posits that effectiveness of
remunerative controls may be negatively related to amount of
alienation experienced. Because prisoners experience a great
deal of alienation, they may be less responsive to remunerative
controls. Inmate assault also has been associated with
traditional inmate culture and the subculture of violence;
therefore, remuneration may be nullified by inmate culture for
certain crimes (Harer and Steffensmeier, 1996; Irwin, 1980).
Inmate assault on staff is a serious crime and perhaps out of the
realm of cultural control; hence, remuneration may be more
effective in preventing staff assault. The results presented in
this study appear to suggest that there nay 1x3 an interaction
between remunerative control and culture; however, further
research is needed to understand the true nature of the

relationship.

25

Appendix A Description of Variables

 

Outcome variable

Physical Assault on Staff

Physical Assault on Inmate

Explanatory Variables

Inmate-Level Predictors

Age

Education

Criminal History

Citizen

Married

Gang Involvement

26

“Since your admission, how
many times have you been
found guilty of physical
assault on a correctional
officer or other staff
member?”

“Since your admission, how
many times have you been
found guilty of physical
assault on another inmate?”

Inmates’
years.

age measured in

Highest grade inmate attended
before admission to prison.

A four item factor score
(eigenvalue = 1.20) including
total sentence in years,
currently serving for a
violent crime, age at first
arrest, and number of prior
arrests.

A dummy variable with 1 =
citizen and 0 = non-citizen.

A dummy variable with 1 =
married and 0 = not married.

A four item additive scale,
measuring participation is
varying levels of gang
organization including:
members of gang from same
area, gang territory, formal
gang membership, and known
gang leader (Cronbach’s Alpha
= 0.79)

Appendix A Description of Variables

 

African American

Other

Hispanic

Length of Incarceration

Prison-Level Predictors
Remunerative Controls

Outside Work Assignment

Prison Employment

Paid Work Assignment

Cocrcivc Controls

Solitary Confinement

27

A dummy variable with 1 =
African American and 0 = non-
African American.

A dummy variable with 1 =
other race and O = not other
race.

A dummy variable with 1 =
Hispanic and 0 = non-
Hispanic.

Inmates' length of
incarceration calculated from
year that inmate entered
prison.

Percent of total inmate
population that is employed
outside prison grounds.

Percent of total inmate
population that is employed
within the prison.

Percent of total inmate
population that is paid for
work.

Percent of total inmate
population that received
solitary confinement as a
disciplinary response to the
most recent rule violations.

Appendix A Description of Variables

 

Coercive Control

28

Two item factor score
(eigenvalue = 1.32; factor
loadings > 0.79) including:
Percent of total inmate
population that lost a work
assignment as a response to
the most recent rule
infraction and percent of
total inmate population that
lost general privileges as a
response to the most recent
rule infraction.

Appendix B Descriptive Statistics:

 

Variable Mean SD Minimum Maximum

 

Outcome Variables

Physical Assault on 0.20 0.92 0.00 15.00
Staff
Physical Assault on 0.74 1.93 0.00 20.00
Inmate

Explanatory Variables
Inmato-Lovol Predictors

Age 30.51 8.15 14.00 73.00
Education 10.61 2.17 0.00 18.00
Black (Black = 1) 0.50 0.50 0.00 1.00
Other Race (Other = 1) 0.04 0.18 0.00 1.00
Hispanic (Hispanic = 1) 0.16 0.37 0.00 1.00
Citizen (Citizen = 1) 0.97 0.16 0.00 1.00
Married (Married = 1) 0.14 0.35 0.00 1.00
Length of Incarceration 2.81 2.56 0.00 33.00
Gang Membership 0.81 1.22 0.00 4.00
Criminal History 0.29 2.56 0.00 33.00

Prison-Laval Predictors

Population 1543.04 1280.41 109.00 7971.00
South 0.38 0.49 0.00 1.00
Maximum 0.30 0.46 0.00 1.00

Remunerativo Controls

Outside Work Assignment 6.52 8.71 0.00 47.90
Prison Work Assignment 64.83 15.33 29.30 98.10
Paid Work Assignment 70.64 38.79 0.00 100.00

29

Appendix B Descriptive Statistics:
(continued)

 

Variable Mean SD Minimum Maximum

 

Coercive Controls
Solitary Confinement 48.80 24.50 0.00 100.00

Coercive Factor Score -0.02 0.93 -1.10 5.87

 

3O

REFERENCES

Accordino, M. and B. Guerney. 1998. "An Evaluation of the
Relationship Enhancement Program with Prisoners and their
Wives." International Journal of Offender Therapy and
Comparative Criminology 42:5-15.

 

 

Akers, R.L., N.S. Hayner and W. Gruninger. 1977. "Prisonization
in Five Countries." Criminology 14:527-554.

 

Bryk, A. and S. Raudenbush. 1992. Hierarchical Linear
Models. Newbury Park, CA: Sage Publications.

 

Cao, L., J. Zhao, and S. VanDine. 1997. “Prison
Disciplinary Tickets: A Test of Deprivation and Importation
models.” Journal of Criminal Justice 25:103-13.

 

Clemmer, D. 1940. The Prison Community. Boston: The Christopher
Publishing House.

 

Colvin, M. 1992. The Penitentiary in Crisis: From Accommodation
to Riot in New Mexico. Albany: State University of New York
Press.

 

 

DiIlulio, J. 1987. Governing Prisons: A Comparative Study of
Correctional Management. New York: Free Press.

 

 

Eisenberg, M., T. Fabelo. 1996. "Evaluation of the Texas
Correctional Substance Abuse Treatment Initiative: The
Impact of Policy." Crime and Delinquency 42:296-305.

 

Ellis, D. 1984. “Crowding and Prison Violence: Integration of
Research and Theory.” Criminal Justice and Behavior 11:277-
308.

 

Ellis, D., H.G. Grasmick, and B. Gilman. 1974. “Violence in
Prisons: A Sociological Analysis.” American Journal of
Sociology 80:16-43.

 

 

Etzioni, Amitai. 1964. Modern Organizations. Englewood Cliffs,
NJ:Prentice-Hall.

 

Etzioni, Amitai. 1961. A Comparative Analysis of Complex
Organizations: On Power, Involvement, and Their Correlates.
New York: The Free Press.

 

 

Flanagan, T.J. 1980. “Time Served and Institutional Misconduct:
Patterns of Involvement in Disciplinary Infractions among
Long-term and Short-term Inmates.” Journal of Criminal
Justice 8:357-367.

 

31

Gaes, G.G. and W.J. McGuire. 1985. “Prison Violence: The
Contribution of Crowding Versus Other Determinants of
Prison Assault Rates.” Journal of Research in Crime and
Delinquency 22:41-65.

 

 

Garabedian, P.G. 1963. “Social Roles and Processes of
Socialization in the Prison Community.” Social Problems 11:
139-152.

 

Gardner, W., E. Mulvey and E. Shaw. 1995. "Regression Analysis of
Counts and Rates: Poisson, Overdispersed Poisson, and
Negative Binomial Models." Psychological Bulletin 118:392-
404.

 

Goetting, A. and R.M. Howsen. 1986. “Correlates of Prisoner
Misconduct." Journal of Quantitative Criminology 2:49-67.

 

Harer, M.D. and D.J. Steffensmeier. 1996. “Race and Prison
Violence.” Criminology 34:323-55.

 

Hewitt, J.D., E.D. Poole, and R.M. Regoli. 1984. “Self-Reported
Observed Rule Breaking: A Look at Disciplinary Response.”
Justice Quarterly 1:437-447.

 

Harrison, L., C. Butzin, J. Inciardi and S. Martin. 1998.
"Integrating HIV-Prevention Strategies in a Therapeutic
Community Work-Release Program for Criminal Offenders." The
Prison Journal 78:232-243.

 

Hewitt, J.D., E.D. Poole, and R.M. Regoli. 1984. “Self-reported
and Observed Rule—breaking: A Look at Disciplinary
Response.” Justice Quarterly 1:437-47.

 

Inciardi, J., S. Martin, C. Butzin, R. Hooper, and L. Harrison.
1997. "An Effective Model of Prison-Based Treatment for
Drug-Involved.Offenders." Journal of Drug Issues 27:261-
278.

 

Irwin, J. 1980. Prisons in Turmoil. Boston: Little, Brown.

 

Irwin, J. and D. Cressey. 1962. “Thieves, Convicts and the Inmate
Culture.” Social Problems 10:142-155.

 

Jacobs, J.B. 1977. Stateville: The Penitentiary in Mass
Society. Chicago: University of Chicago Press.

 

Light, S.C. 1990. “The Severity of Assaults on Prison Officers: A
Contextual Study.” Social Science Quarterly 71:267-284.

 

Kantrowitz, N. 1996. Close Control: Managing a Maximum Security
Prison. Guilderland, NY: Harrow and Heston.

 

32

Kreft, I. and J. De Leewu. 1998. Introducing Multilevel Modeling.
Thousand Oaks, CA: Sage Publications.

 

Matthews, R. and J. Pitts. 1998. "Rehabilitation, Recidivism, and
Realism: Evaluating Violence Reduction Programs in Prison."
The Prison Journal 78:390-401.

 

McCorkle, R.C., T.D. Miethe and K.A. Drass. 1995. “The Roots of
Prison Violence: A Test of the Deprivation, Management, and
“Not-so-total” Institution Models.” Crime and Delinquency
41:317—331.

 

Reisig, M.D. 1998. “Rates of Disorder in Higher-custody State
Prisons: A Comparative Analysis of Managerial Practices.”
Crime and Delinquency 44:229—244.

 

Ruback, R.B. and T.S. Carr. 1993. “Prison Crowding over Time: The
Relationship of Density and Changes in Density to
Infraction Rates.” Criminal Justice and Behavior 20:130-
148.

 

Sykes, G.M. 1958. The Society of Captives: A Study of a Maximum
Security Prison. Princeton, NJ: Princeton University Press.

 

 

Thomas, C.W. 1977. “Theoretical Perspectives of Prisonization: A
Comparison of the Importation and Deprivation Models.” The
Journal of Criminal Law & Criminology 68:135-145.

 

Turner, S. and J. Petersilia. 1996. "Work Release in Washington:
Effects on Recidivism and Corrections Costs." The Prison
Journal 76:138-150.

 

U.S. Department of Justice, Bureau of Justice Statistics.1993.
Survey of Inmates of State Correctional Facilities, 1991:
[United States] [Computer File]. Conducted by U.S. Dept.
of Commerce, Bureau of the Census. ICPSR ed. Ann Arbor, MI:
Inter-university Consortium for Political and Social
Research [producer and distributor], 1993.

 

Useem, B. and P. Kimball. 1989. States of Siege: U.S. Prison
Riots, 1971-1986. New York: Oxford University Press.

 

 

Useem, B., C.G. Camp and G.M. Camp. 1996. Resolution of
Prison Riots: Strategies and Policies. New York: Oxford
University Press.

 

 

Wallace, J.M. Jr. and J.G. Bachman. 1991. “Explaining
Racial/Ethnic Differences in Adolescent Drug Use: The
Impact of Background and Lifestyle.” Social Problems
38:333-341.

 

33

Ward, A. and S. Baldwin. 1997. "Anger and Violence Management
Programs in Correctional Services: An Annotated
Bibliography." The Prison Journal 77:472-488.

 

Wheeler, S. 1961. “Socialization in Correctional Communities.”
American Sociological Review 26:697-712.

 

Williams, K. and R. Flewelling. 1988. “The Social Production of
Criminal Homicide: A Comparative Study of Disaggregated
Rates in American Cities.” American Sociological Review 53:
421-431.

 

Wright, K.N. 1994. Effective Prison Leadership. Binghampton,
NY: William Neil.

 

34

MICHIGAN sran UNIV. LIBRARIES
i“WNWHI“WlililllillllmllilllWWHI
31293017806674