Content uploaded by Gerald G. Gaes
Author content
All content in this area was uploaded by Gerald G. Gaes on Feb 18, 2015
Content may be subject to copyright.
Prison Crowding Research Reexamined
Gerald G. Gaes
Federal Bureau of Prisons
Revised, January 1994
1
The opinions expressed in this paper are solely those of the
author and are not intended to represent the policies of the
Federal Bureau of Prisons or the Department of Justice.
Abstract
Dramatic increases in the United States' inmate population has
raised new concerns about prison crowding. Although growth in
prison capacity has lagged slightly behind that of the inmate
population, there is no consistent evidence that crowding is
associated with mortality, morbidity (defined as clinic
utilization), recidivism, violence, or other pathological
behaviors. This paper reviews the major areas in which prison
crowding has been examined. Conceptual, methodological, and
empirical criticisms are raised concerning prison crowding and the
areas of health, violence, and recidivism. The paper is divided
into five sections: I. the political and social context of prison
crowding research; II. ecological versus individual level
differences in crowding; III. theoretical and empirical problems
associated with violence and its relation to crowding; IV. an
analysis of inmate illness reporting and its relation to crowding;
and V. the degree to which the literature points to consistency in
results both within the prison and across other crowded settings.
Among several issues discussed in section I, the criteria
2
suggested by Thornberry and Call (1983) for deciding prison
crowding suits are evaluated. It is concluded that the first
criterion, that the courts first consider general evidence on the
relationship between crowding and debilitation, cannot be met,
based on the extant crowding literature. In the area of violence,
it is argued that most prison crowding studies do not clearly
examine the relationship of intervening mechanisms proposed by
theorists to account for the relationship between crowding and
violence, if and when a relationship is found. Furthermore, it is
suggested that one reason underlying the inconsistency in the
results of these studies is that researchers have failed to
examine the proximal causes of violence and the formal mechanisms
prison administrators use to control or limit violence. With
regard to health, it is argued that the most demonstrative
finding, that dormitories are associated with higher illness
reporting rates than are other types of housing, is probably an
artifact of selection bias. Furthermore, illness reporting is the
result of a complex set of circumstances that is affected as much
by psychological and sociological causes as by the health status
of the inmate. Thus, despite the prevailing sentiments about the
harmful effects of crowding, there is little consistent evidence
supporting the contention that short- or long term- impairment of
inmates is attributable to prison density. One reason for this may
be that researchers have failed to consider management
interventions under periods of high confinement, and have failed
3
to account for conditions other than crowding that affect inmate
debilitation.
4
The real-life impact of the social science literature that
treats the effect of prison crowding is undeniable. This
literature has been used by the legislatures, courts, prison
administrators, and others to shape policy that has profound
implications for both the incarcerated and for society as a whole.
Unfortunately, the oversimplification of problems and conclusions
in the literature itself reduces its usefulness and may lead to
less than optimal or even erroneous policy applications.
In addressing any problem area, one first must define the
terms or operational definitions one is using so that others can
determine whether they are studying the same phenomenon.
Unfortunately, the terms used in the crowding literature are often
unclear or confusing. The term "crowding" fails to differentiate
between the two operational definitions of density: social
density, which refers to the number of persons in a given area,
and spatial density, which refers to the amount of space
apportioned to each individual. Analysis is made more difficult by
the realities of prison administration. At the aggregate or
ecological level, crowding is usually defined as the ratio of the
number of inmates in a prison to its rated capacity. Rated
capacity usually combines spatial and social density so that a
change in one density is confounded with a change in the other
kind of density. Usually, increases in the number of prisoners
confined leads to decreases in the amount of space per person.
Furthermore, the effect of housing space on prisoners is typically
5
obscured by the highly variable amount of time spent in the
housing unit and by the availability and spaciousness of other
prison areas such as the prison library or recreation yard. I have
discussed in greater detail the many conceptualizations of
crowding in a prison setting in a previous review (Gaes, 1985).
In the that review, I concluded that there were two basic
findings to the emerging prison crowding literature: (1) open bay
dormitories were associated with higher clinic utilization and
elevated blood pressure; (2) prisons which greatly exceeded their
rated capacity, or contained dormitories, had higher assault
rates. Based on more recent evidence and a reconsideration of the
previous data, I now believe neither of these relationships has
much support.
There are five areas I consider in this paper: I. the legal,
political, and social context of prison crowding research; II.
theoretical differences between ecological and individual levels
of density; III. conceptual and empirical problems with the
relationship between prison violence and density; IV. the
relationship between density and inmate health, where health is
primarily defined as the rate at which inmates use the prison
health clinic
1
; and V. the degree to which there is consistency in
the crowding literature. A summary also appears at the end of the
paper.
Most attention is spent on the relation of crowding to inmate
health, both because there has been some consistency to the
6
research finding that dormitory housing produces increased
illness-reporting over single- or double-bunked housing, and
because if it can be shown that crowding is related to health
debilitation, this would merit serious attention by the criminal
justice community. However, the purported links between crowding
and health have not been critically evaluated in the light of a
rich literature that has been developed to account for clinic and
physician utilization. This literature is important because it
shows the many factors involved in the decision to seek medical
care and how these factors may or may not be influenced by
crowding. I do not consider some of the public health -
epidemiological literature on the relationship between crowding
and contagion. This is a complicated area involving knowledge
about epidemiology, immunity, and public health policy. This
literature should also be considered in the public health, prison
policy debate.
I also think the other four sections of this paper are
equally important in understanding the crowding literature. The
section on ecological versus individual differences in crowding
studies is most important for theorists who are primarily familiar
with individual data. The ecological fallacy (Robinson, 1950) of
inferring individual relations from aggregated data may be causing
further confusion in the crowding literature. The section on
crowding and inmate violence suggests that formal control
mechanisms used by corrections staff have been ignored by most
7
crowding researchers, and this may be the reason that there has
been such inconsistency in the crowding-violence studies. The
section on the presumed consistency across crowded settings is
intended to challenge the belief held by some researchers and
public policy practitioners that crowded environments produce
similar results in different settings. Although the scientific
merit of the crowding literature should stand or fall on its own
weight, there are non-research participants to this imbroglio, who
often have different public policy perspectives, which may or may
not be research-based. This is inevitable; however, the first
section tries to give some perspective to the status of the
crowding literature in relation to these different views.
I. The Legal, Political, and Social Context of Prison Crowding
Research
Quite apart from the research literature on prison crowding,
there is a multi-level, multifaceted context in which the crowding
research is evaluated and applied. The continued growth in the
confined population is one context. The legal context is another.
Directly related to the prison population growth, is the current
debate over determinate versus indeterminate sentencing and the
politicalization of crime policy. Finally, there is the political
context of prison overcrowding itself.
8
Recent Prison Population Growth
Increases in the confined population in the United States and
its territories have been viewed by some criminologists as
synonymous with a crisis in corrections resulting from overcrowded
prisons (see Gottfredson and McConville, 1987). From 1980 to June
1993, the combined State and Federal prison population climbed
from 329,821 to 925,247 (Bureau of Justice Statistics (BJS),
1993), an increase of 181 percent. Using each State's definition
of capacity, BJS found that in 1987, using a high estimate of
capacity, State prison systems were, on average, 5 percent
overcrowded, and using a low estimate of capacity, 19 percent
overcrowded. The State with the lowest reported density was Utah,
which was 18 percent below its capacity. Data published in 1992 by
BJS were comparable. The range of lowest to highest estimates of
crowding was between 3 and 13 percent. In 1992, Vermont reported
a high overcrowding level of 93 percent. For Federal prisons,
which represent about 8 percent of the entire prison population,
overcrowding was 37 percent.
The capacity of a prison system can be an elusive number
unless some care is used in systematically defining the living
arrangements and living space in an institution. One careful
analysis (Mullen and Smith, 1980) showed that by rigorously
applying the cell standards recommended by the American
Correctional Association (ACA), in March 1978 nearly two-thirds of
9
all prisoners were confined to below-standard units -- i.e., units
with less than 60 square feet per person. Every 5 years the Bureau
of Justice Statistics completes a prison census in which a
thorough analysis of capacity is assessed. The last census in 1990
showed that between 1984 and 1990 the inmate population rose 63
percent while the average amount of housing space declined from 57
to 56 square feet per person. In this same time period, however,
the number of correctional officers rose at a faster rate than the
inmate population. While there was a 63 percent increase in the
average daily inmate population from 1984 to 1990, there was a 70
percent increase in staff. Thus, despite the dramatic increase in
prison inmate population in the last 10 to 14 years, it appears
that both construction and staffing increases have kept pace with
the inmate growth.
The Legal Context of Prison Crowding
Another approach to describing the status of crowding and
corrections has been to cite the amount of litigation addressing
crowding and unconstitutional conditions in various jurisdictions.
According to an American Civil Liberties Union (ACLU) annual
report released in December 1988, there were 10 entire prison
systems, including Puerto Rico's, under court order or consent
decree in which overcrowding was a primary issue. In addition, 30
jurisdictions had a major prison under court order or consent
10
decree in which crowding was cited as a primary or major problem
(Corrections Digest, 1989). The primary issue in court
interventions has been the quality of conditions with regard to
(but not limited to) inmate safety, medical care, nutrition, and
sanitation. Ingraham and Wellford (1987) have pointed out that the
lower courts have treated crowding as the source of all problems
instead of being a contributing factor. The "totality of
conditions" test emphasized in Rhodes v. Chapman (1982) made it
clear that crowding and double bunking were not by themselves
unconstitutional. Crowding may exaggerate harsh and unpleasant
conditions to the extent that they become unconstitutional, but
the test is not the precise level of crowding, but whether a
prison system can deliver adequate care and protection. The Rhodes
standard requires that plaintiffs demonstrate there is an
objective prison condition that deprives them of life's basic
necessities.
Alexander (1993) has reviewed recent U.S. Supreme Court cases
relating to suits filed under Title 42, § 1983. In Wilson v.
Seiter (1991), an inmate alleged he was being subject to cruel and
unusual punishment due to overcrowding, excessive noise,
inadequate heating and cooling, improper ventilation, unclean and
inadequate restrooms, unsanitary dining facilities and food
preparation, and the close proximity of mentally and physically
ill inmates. According to Alexander, at issue in the Supreme Court
decision was whether "a plaintiff challenging conditions of
11
confinement must demonstrate a culpable state of mind on the part
of prison administrators(p. 107)." The U.S. Supreme Court, in a 5
to 4 decision ruled "that an inmate must show a culpable state of
mind and that the required state of mind is "deliberate
indifference" by prison administrators (p. 107)." Whether the
Wilson decision now adds a second standard, which is a "state of
mind" criterion to establish future claims under the eighth
amendment is beyond the scope of this paper. However, the clear
signal from the U.S. Supreme Court in recent decisions is that it
has made it more difficult for inmates to prevail in prison
conditions suits.
Call (1988) has reviewed the case law from 1979 to 1986 to
see whether lower courts required inmate-plaintiffs to demonstrate
the harmful effects of prison crowding. Prior to the more recent
stringent standards set forth in Wilson, inmate-plaintiffs won 73
percent of the reported cases
2
. In the 65 cases decided in
Federal courts that Call reviewed, 35 did not consider evidence
regarding harmful effects. According to Call's analysis, the
lower courts had not heeded the Supreme Court's "sermon about
giving greater deference to the decisions of corrections
officials" (p. 35). Thornberry and Call (1983) advocated an
approach to deciding prison crowding cases based on two criteria:
(1) the court first considers evidence on the general relationship
between crowding and debilitation; (2) the court then considers
evidence that crowding had caused specific problems at the
12
institution under consideration. Of the 65 Federal district or
appellate cases studied by Call, only 4 had used such rigorous
criteria.
There remains a great deal of confusion concerning how to
deal with prison crowding in relation to other conditions of
confinement. Clearly, the courts as well as social scientists
must acknowledge there is tremendous variability in the quality of
prison conditions independent of crowding levels. From my
perspective, the social science prison crowding literature is
fraught with theoretical and methodological problems.
Furthermore, since there is such inconsistency in the findings,
the literature cannot be the basis for establishing Thornberry and
Call's first criterion, that prison crowding is either a necessary
or sufficient condition for debilitation. Obviously, there can
still be Eighth Amendment tests based on the totality of
conditions criterion; however, alleviation of the problems may
require a combination of administrative responses including some
reduction in crowding, better classification of inmates, better
training of staff, and changes in the adjudication of misconduct.
In one of the most encompassing prison conditions suits, Martin
and Ekland-Olson (1987) have shown that all of these remedies were
imposed on the Texas Prison System.
Sentencing Policy and the Politicalization of Crime
13
Benekos (1991) has demonstrated the interrelationship of
sentencing policy, prison crowding, and public policy. He recounts
the history of the criminal justice public policy debate in
Pennsylvania during a period of proposed sentencing reform, high
growth in the prison population, and prison unrest resulting in a
significant prison riot. The debate in Pennsylvania was typical of
the public policy debate confronting many of the states. The
political rhetoric of crime control, which calls for tougher
sentencing policy, collided with the fiscal dilemma of funding the
additional prison beds that result from such a policy. In a recent
publication, the Advisory Commission on Intergovernmental
Relations published a report (1993) demonstrating the Federal
Government's role in the politicalization of crime control policy.
During the 1980's, Congress passed five crime bills and is
currently debating yet another. The result of this legislation at
the federal level has been to eliminate parole, decrease inmate
good time, establish mandatory minimum sentences for some crimes,
and lengthen prison time served for many crimes. The Advisory
Commission notes that the Federal influence on criminal justice
policy throughout the states has been substantial and that many
states have passed or proposed similar local legislation. In this
context, legislatures have the difficult problem of balancing
public demands for crime control and cost containment. While
corrections officials argue for funds to expand prison capacity,
there is a tension resulting from the burden this place on state
14
budgets. As the Advisory Commission report notes, the largest
growth in any component of state appropriations over the last 10
years has been in corrections.
As I have already noted, despite the dramatic growth in
the incarceration rate and the concomitant increase in prison
populations in the United States, it appears that in the
aggregate, construction and staffing have kept pace. This is not
to say that some prison systems may be understaffed or extremely
crowded, while others are not. The issues surrounding crowding,
prison reform, and sentencing reform are often argued, debated, or
discussed in the same forum. While these issues are interdependent
in the sense that policy decisions of one kind affect policy
decisions of another kind, these issues should also stand or fall
on there own merit. "Is determinate sentencing more equitable than
indeterminate sentencing?" is a separate issue from the question
of whether alternatives to incarceration and a broad range of
prison sentences are central to a rational sentencing scheme
(Morris and Tonry, 1990). Both of these issues are independent of
questions about prison crowding thresholds and manifestations of
crowding.
The Political Context of Prison Crowding
There is an interesting dilemma that arises from the
different perspectives of social scientists, policy makers, and
15
judges on the crowding issue. Each bring different criteria to
bear upon the importance or significance of the effects of
crowding. From a judicial/legal perspective, it would seem that
whether or not a general relationship exists between crowding and
the quality of inmate life is irrelevant to a specific finding
that a prison is in violation of the Eighth Amendment. Because
there will be exceptions at many diverse levels of crowding, a
finding of fact must be made for the specific prison under
consideration. If a finding of unconstitutional conditions
prevails, then a reduction in crowding may be the logical choice
to alleviate conditions. However, there may be other practical
interventions as well.
From the perspective of prison administrators and lawmakers,
correctional standards and practical limits on the number of
inmates a facility can hold are major concerns. With regard to
standards, administrators must be concerned about minimum levels
of inmate living conditions. Unfortunately, if they turn to
social scientists, it is rare that they will receive absolute
criteria with regard to crowding levels. Social scientists may
provide suggested standards; however, these must be weighed
against the needs of the public and the cost of designing and
constructing prisons. A very different set of questions arises
from the perspective of prison administrators than from that of
social scientists. Prison administrators, aside from their
concern about legal actions against them, want to know such
16
factors as the caseload limits of prison staff, the capacities of
prison dining facilities, or the capabilities of their medical
facilities. These issues constitute the "nuts and bolts" of
prison administration; however, such problems may be considered
too mundane by some social scientists, who are primarily
interested in the effects of crowding on inmates.
Finally, social scientists have traditionally concentrated on
what they have presumed to be the damaging effects of prison
crowding. This is because this area of study evolved from the
examination of subhuman species in which overpopulation led to
dramatic changes in mortality and other behaviors both in natural
environments (e.g., Christian and Davis, 1964) and laboratories
(Calhoun, 1962). Because of this orientation, the research
community has generally ignored other environmental and
psychological stressors; however, this seems to be changing.
Zamble and Porporino (1988) and, to some extent, Toch and Adams
(1989) have examined not only some of the other causes of
pathology and adaptation in prison, but have formulated the
research to suggest potential intervention strategies as well.
The reason there is a dilemma involving these different
perspectives is that there is no clear and consistent evidence
that levels of density are unequivocally (or even in all
probability) related to human health or pathology. If we could
say with some precision that a prison which confines more than
some fixed percent of its capacity will necessarily lead to
17
violence and illness, then this could be the basis for
establishing Eighth Amendment tests of inmate living conditions
that would satisfy judges as well as standards for correctional
institutions that would meet the needs of correctional
administrators. That kind of evidence may or may not be
achievable and, until it is, the criteria for judges and
administrators will shift with the context of situation.
Related to the different interests of judges, administrators,
and social scientists is the political context in which crowding
policy is decided. Bleich (1991) has argued that "there is no
active participant in the [prison crowding] "debate" with an
institutional interest in contesting the perception of crowding"
(p. 1127). Bleich extended his analysis to prison administrators,
reformers, prison staff, inmates, and legislators, showing how
each has a vested interest in supporting the conception that
prisons are overcrowded. After reviewing the history of crowding
litigation and some of the social science crowding literature,
Bleich came to three conclusions. First, courts and legislatures
"have an interest in having facilities declared overcrowded (p.
1128)." This is because the courts gain control over a domain
previously left to the executive branch and legislatures seize the
opportunity to appropriate more money for prisons based on claims
of overcrowding. Second, prison managers have an interest in
institutionalizing crowding. It can provide an excuse for
18
conditions that may be otherwise ameliorated by better prison
management. Furthermore, by operating prisons above their
capacity, corrections administrators can make the claim they need
increased funding to manage their systems. Lastly, Bleich notes
that the greatest danger of the prison crowding debate is that it
may obfuscate the more significant problems underlying prison
conditions. Citing conclusions I reached in an earlier review of
prison crowding research (Gaes, 1985) and other evidence, Bleich
reasoned that
"..., not only may the extent of crowding be
exaggerated, but also crowding may be institutionalized
in the political debate at these exaggerated levels.
This may result in a misallocation of scarce resources
among social institutions. Perhaps more disturbingly, it
may not improve conditions after all (p. 1179)."
If Bleich is correct in his conclusions, one might expect
that there would be a continued trend toward lower density
standards whether promulgated by the courts or by standards
associations such as the American Correctional Association (ACA).
In fact, the trend has been quite the opposite. The U.S. Supreme
Court decisions Rhodes v. Chapman and Wilson v. Seiter firmly
abandoned the one man - one cell requirement. More recently, the
American Correctional Association standards committee approved a
19
change effective in 1992 that allowed multiple occupancy housing
in both minimum and medium custody populations (ACA, 1992).
Previously, multiple occupancy housing had only been allowed in
minimum custody facilities. Furthermore, ACA reduced the amount of
unencumbered space requirement from 35 to 25 square feet in these
multiple occupancy housing units. However, ACA also added
requirements to meet compliance with these housing standards. For
example ACA now requires single cell housing for special needs
offenders such as those seeking protective custody or those
suffering from a mental disorder.
Bleich's analysis may have been more appropriate for the
period of the 1970's and early 1980's, the most progressive era in
modern prison reform. While there are still pressures toward
prison reform and stricter crowding standards, there clearly has
been a slow down, if not a reversal in that trend.
II. Ecological Versus Individual Crowding Effects
Because crowding has been studied at both the individual and
aggregate (ecological) levels of analysis, some confusion persists
over theoretical expectations at these different levels. Most
researchers have treated the different studies as if they were
reflections of the same process, when it may be the case that
crowding at an ecological level influences individuals in very
20
different ways than crowding at an individual level. An example
from a different, but related set of variables may help to clarify
this problem. A similar confusion has arisen over the relationship
between age and assault rates in prison. At an ecological level,
it is more useful to view age in terms of age composition or age
blending. Age blending refers to the policy that purposely houses
older inmates with younger inmates.
Proponents of age blending in a prison setting have argued
that older inmates tend to suppress the negative behavior of
younger, more volatile inmates. We might assume that the
unfettered relationship between age and assault rates at the
individual level is as it appears in figure 1 (i.e., the higher
the percentage of younger inmates, the higher the assault rate).
However, figure 2 shows the individual relationships at different
levels of the ecological variable, age composition. Figure 2
demonstrates how the contextual effect of age composition modifies
the individual effect of age with respect to assault rates. It
shows that the individual level relationship between age and
assault can change with different age compositions. To complicate
this example further, it is probably the case that the age
composition of an institution is also related to other ecological
characteristics such as staff-inmate ratios and available
programs. These variables might also affect the individual-level
effects of age with respect to assault rates.
21
In the same way, crowding may have very different effects at
both the ecological and individual levels of measurement. At an
ecological level, density may be related to economic costs and
benefits (i.e., economies of scale), the efficient or inefficient
delivery of services (also an economies-of-scale issue), and
custodial requirements concerned with monitoring and controlling
the inmate population. That is, density may affect an
institution's ability to control behavior and maintain health. At
the individual level, density could have very different effects on
stress, and consequently on behavior and health. In short, density
has different effects on institutions and individuals, and the
effects of density on institutions can modify effects on
individuals.
III. Crowding and Inmate Violence
Theoretical Problems
The relationship of crowding to violence has been studied
either in the absence of theory or with little or no effort to
conduct a strong test of the theoretical assumptions underlying
the assumed crowding-pathology relationship. Most of the
crowding-violence literature has examined the relationship between
some operational definition of crowding and some measure of
22
violence. There are few instances in this literature in which the
relationship of crowding to other prison variables (e.g. inmate-
staff ratios, institution size) has been studied - variables that
may clarify how crowding actually relates to violence. When other
variables have been used, it has typically been in the context of
controlling sources of error in the interpretation of the crowding
relationship. This shortsightedness may make many of the policy
implications of the prison crowding research irrelevant or very
limited in their application. The exceptions to this approach
(Ekland-Olsen, 1986; Ellis, 1984) have introduced rather broad
theoretical explanations of the possible link between crowding and
violence.
If prison administrators were to compile a "laundry list" of
all of the variables that may provoke violence among inmates, the
list would probably contain the following: prison drug
trafficking, homosexual relationships, predatory behavior, gang
confrontations, arguments over thefts and valued possessions, and
racially prejudicial behavior. These are the typical proximal
causes of violence. Social scientists have proposed that crowding
affects these proximal causes in two ways. The first is through
psychological mechanisms: crowding causes stress that elevates
the "arousal level" of the inmate and makes any behavior more
likely to occur or, secondly, unwanted interactions due to
increased density stress the individual over time. These are
stress-mediated mechanisms. The second set of mechanisms are
23
sociological. Social control theorists (Ekland-Olsen, 1986;
Ellis, 1984) argue that crowding creates uncertainty in the
control environment through transiency, a weakening in staff-
inmate attachments, and reduction in the certainty of punishment.
Ellis (1984) has noted the overlap in these approaches, and
Ekland-Olsen has argued that it is more likely social control
mechanisms explain serious violence (homicides) and some
combination of the psychological and sociological approaches
explain less serious misconduct. There have been very few
attempts to specify operational measures of stress or social
control constructs and then show how such measures 1) relate to
crowding and 2) mediate assaults or homicides. Ekland-Olsen,
however, has attempted to demonstrate how social control
mechanisms may have led to changes in the violence patterns during
resolution of the Ruiz suit involving the Texas correctional
system.
In summary, the crowding research offers little insight into
how crowding may interact with the proximal causes of violence.
Instead, researchers in this arena study the relationship between
a crowding measure and some measure of inmate violence, and this
is taken as evidence that crowding causes violence. Such research
does not address the relative contribution of crowding to these
other proximal causes of violence, nor does it provide meaningful
results for policy intervention. The reduction of crowding from a
level of 150-percent overcrowding to 120-percent overcrowding may
24
have little or no impact on the proximal causes of violence. It
may be a waste of resources when other interventions are more
meaningful to the prison climate. With regard to violence, these
other interventions may include a better inmate security
classification system to reduce predatory behavior by separating
predators and more aggressive inmates from the other inmate
population, a more comprehensive random urinalysis program to
reduce drug usage and drug trafficking, more staff to increase
surveillance and cope with inmate problems, more concrete policies
regarding property ownership to reduce theft and an inmate
"underground economy," and more programs to promote meaningful
work and learning opportunities for the prison population.
These interventions are formal control strategies that are
practiced in varying degrees by administrators in most prison
systems. To simply focus on crowding to the exclusion of these
interventions adds little to the policy repertoire of prison
administrators. The prescription I am proposing for crowding
research as it relates to violence not only requires the
intervening constructs to be more clearly stated and measured, but
also requires the relative assessment of formal control
strategies. As Ruback and Innes (1988) have pointed out, most
crowding research has been conducted by social scientists who are
most focused on the processes underlying crowding, stress, and the
consequences of stress. Although their focus is on these
mediating constructs, they have failed to do strong tests of these
25
theoretical assumptions. Until crowding research also begins to
address the formal control strategies, it may be of limited value
to prison administrators, legislators, and jurists who must
balance issues of cost, humane treatment, and inmate and staff
security.
Crowding and Violence: Empirical Problems
There are also empirical reasons to question the
crowding-violence relationship. One of the few studies to find an
aggregate prison crowding and assault relationship (Gaes and
McGuire, 1985) also found that this relationship was inelastic
3
.
A 1-percent increase in social density resulted in a .3-percent
increase in the rate of inmates assaulting other inmates. In the
Federal prison system, where this data was collected, this meant
that the prison population could double with an increase of
between only 2 and 3 assaults per 10,000 inmates in a 1-month
period. Furthermore, Innes (1986) found that when looking at all
major prisons in the United States, there was no relationship
between crowding (defined as spatial density) and the assault
rates at the various State prisons. More recent studies by
Porporino and Dudley (1984) and Fry (1988) have not shown a
relationship between density and assault rates. Gaes and McGuire
(1985) reviewed the prison density-violence relationship.
26
Although the existing research results revealed various
inconsistencies, Gaes and McGuire argued that the inconsistency
was due to different specifications in the quantity and quality of
variables in the relationships and different levels of analysis.
Upon further consideration, it seems there might be fundamental
differences in the way individual Federal and State prison system
managers and staff respond to crowding. At an ecological level,
it may be possible to evaluate prison system differences in the
crowding-violence relationship; however, until there is a theory
relating the ecological influences to system responses, crowding
studies will continue to be aimless inquiries based on arbitrary
theoretical specifications.
Crowding and Violence: Prison Disturbances
In a Wall Street Journal article on August 18, 1981 ("Life in
Prison: More Riots Are Feared as Overcrowding Fuels Tension Behind
Bars"), a staff reporter addressed the Nation's swelling prison
population. In that article, the reporter introduced the topic by
describing the prison riot in the New Mexico State Penitentiary in
which 33 inmates died, some by decapitation or mutilation.
Although the reporter cited the State's attorney general's report
indicating the major problem was mismanagement, the report also
implicated chronic prison overcrowding. This article is typical
of many other newspaper pieces which have implicitly or explicitly
27
linked crowding with prison riots. Unlike the Wall Street Journal
article, which merely identified crowding as a problem area, most
newspaper accounts wrongly treat prison crowding as synonymous
with major disturbances.
This image of the crowded prison as a boiling cauldron of
inmate hostility, ready to erupt instantaneously into a riot has
also been used by prison administrators when they are asking for
funds to build more housing. To what extent is this image true?
It is very difficult to track prison disturbances in relation to
crowding, because riots occur rather infrequently. Journalists,
administrators, and other participants in the prison crowding
debate have used the suggested relationship between crowding and
riots for their own convenience.
Useem and Kimball (1989) define a prison riot as "...when the
authorities lose control of a significant number of prisoners, in
a significant area of the prison, for a significant amount of
time" p. 4). Useem and Kimball, who have studied some of the
major prison riots in the last two decades including Attica, Santa
Fe, and Joliet, contend that many of the sources of tension that
existed in these prisons prior to the riots also were present in
most prisons throughout the country. The major cause of riots in
these prisons was a breakdown in the administrative control and
operation of the prison, including an erosion of the security
system. The authors do not exclude objective deprivations (e.g.,
poor sanitation, violence, crowding) as a cause of prison riots;
28
however, they argue that many other prisons have suffered from the
same deprivations. The difference between besieged prisons and
the others was that inmates at the former perceived their
grievances to be legitimate. Additionally, they viewed the State
authorities responsible for the prison as likely to capitulate to
at least some of their demands.
While the previous discussion could be considered a broad
critique of the prison crowding literature and violence, the
following critique provides a more detailed analysis of the most
cited relationship between crowding and debilitation: the effect
of crowding on illness reporting.
IV. Crowding and Health (Illness Reports)
Two levels of analysis, aggregate and individual, have been
used to study the presumed deleterious effects of prison crowding
with regard to health. Aggregate studies have focused on indicia
of pathology such as rates of morbidity, mortality, and
psychiatric commitments (see Cox, Paulus and McCain, 1984).
Individual studies of prison crowding have examined a variety of
criteria including self-reports of stress, symptomatology,
hostility, and psychopathology measures of physical health such as
blood pressure, urine catecholamine production, and illness
29
reporting in a prison clinic. Illness reporting has been the most
frequently used criterion of health. Because of the focus on
illness reporting in the prison crowding literature and the
importance this measure has had in interpreting crowding's effect
on health, I will examine the theory and data underlying illness
reporting in some detail. Most of the crowding literature relying
on illness reporting does not recognize there is not necessarily a
strong correspondence between illness reporting and health.
Theories concerning crowding have dominated the explanations
for variations in prison illness reporting. This has occurred to
the exclusion of: (1) an understanding of the meaning and role of
illness reporting and (2) the consideration of other causes that
may prove to be more meaningful, more influential, and more
tractable to policy changes. One basic underlying model of
crowding which attempts to explain the variation in illness
reporting is that crowding antecedents such as high social density
results in a stress reaction which in turn leads to pathology
and/or psychopathology. This model has several variations (see
Cox, Paulus, and McCain, 1984; Gaes, 1985); however, illness
reports have been used as a primary criterion in drawing
conclusions about the relationship between crowding and inmate
health. In the following sections, I will examine in greater
empirical and theoretical detail the concept of illness reporting;
examine other research that has focused on antecedents of clinic
and physician utilization that have been overlooked in the
30
crowding-illness literature; and assess the relative contribution
of crowding to some of the other causes of illness reporting
variations.
Illness Reporting: Conceptual and Empirical Relationships
The process of illness reporting begins with a decision by
the inmate to go to the prison clinic during "sick call," or those
times when he is feeling acutely ill. The inmate is seen by a
physicians assistant, nurse, or physician and the results of the
medical intervention are recorded in the inmate's chronological
medical record. To measure illness reporting, an inmate's medical
record is culled for evidence related to the inmate's visit. The
data may consist of symptoms, diagnoses, lab tests, or suggestions
for further medical testing. The data are typically reported as
an incidence rate in a given time frame. In the prison crowding
literature, illness reports have been presumed to be a veridical
reflection of inmate health. For the remainder of this section, I
will analyze the nature of illness reports in the context of
prison crowding. The conclusions of this analysis are germane to
other research applications in which illness reporting is used to
assess health care issues such as those reported in the sociology
of medicine literature.
31
For clarity, we operationally define illness reporting as
the recording of a visit to the prison health clinic. The medical
record may contain a combination of the symptoms described by the
patient (e.g., pain, headache, anxiety), the medical indicia of
symptomatology (e.g., fever, rash, blood pressure level) and the
medical diagnosis (e.g., influenza, allergy, hypertension). As a
further refinement, I am restricting illness reporting to
voluntary visits, rather than a series of treatments (e.g., blood
pressure clinics) or required examinations (e.g., medical
histories and physical examinations).
Although there is a certain face validity to the assumption
that illness reports represent the underlying health of the inmate
population, there are parameters other than the health of the
inmate population which determine the extent of health service
utilization in a clinic setting. These other parameters are
related both to dispositional characteristics of the inmate
population and the administration of medical care in an
institution.
Table 1 indicates a large variation in individual illness
reporting (0-31 visits) over a six month period, as well as a
large variation in the incidence of reporting by institution (.016
- .044 visits per person per day). These data were collected by
Paulus, Cox, and McCain in six Federal prisons containing only
male prisoners (see Paulus, 1988). In examining the individual
level data, one can see that nearly one-third of the inmate
32
population never used the prison clinic in a 6-month period.
Furthermore, more than half of the inmates used the clinic once or
not at all in the 6-month period. A few inmates skewed the
distribution by using the clinic 11 to 31 times in the 6-month
time frame. Future research should identify these heavy users to
examine the etiology of such high usage. Turning to the
institution data, one can see that Site 3B's rate was 2.75 times
that of site 5. There does seem to be some stability to illness
reporting rates for the two sites (sites 2 and 3) that were
measured at different times. Clinic utilization can be extremely
high. As much as 67 percent of the inmate population may have
used the health clinic in a 1-month period (see McCain et al.,
1980). Table 1 also lists distributional information on illness
reports for a sample of 1,199 inmates in a 6-month period. It
also lists the different institutional incidence rates computed as
the number of reports per days at risk. This individual and
institutional variation has been attributed to levels of crowding;
however, there are many other factors that cause these variations.
Illness Behavior, Health Behavior, and the Sick Role
Analytical distinctions among the purposes of health-related
behavior have been the focus of work in the sociology of medicine
literature. Kasl and Cobb (1966) distinguished between health
behavior, illness behavior, and sick role behavior. Health
behavior is preventive, involving actions taken to insure health.
33
Illness behavior involves the actions of people who feel ill and
seek the definition, causes, and remedies for their illness. As a
consequence of the definition of illness, sick role behavior
includes activities while the individual is ill and attempting to
get better (Parsons, 1951).
A parallel theoretical development in the health education
literature has identified a conceptual model for understanding why
individuals engage in health-related actions. The health belief
model (see Becker, 1974) consists of attitudinal and behavioral
dimensions which focus on the individual's desire to avoid illness
and the belief that a specific health action will prevent or
ameliorate an illness (Janz and Becker, 1984). In their review of
the health education literature, Janz and Becker identified
perceived barriers as the most consistent dimension of the health
belief model in predicting health actions. Barriers refer to the
implicit cost analysis the individual undergoes when weighing the
disadvantages to a health action. The costs include medication,
side effects, other health action side effects, inconvenience, and
time away from work.
The second most common dimension in predicting health actions
was the individual's perceived susceptibility to contracting a
condition. The third most important dimension was the perceived
benefit of a health action. Thus, although an individual might
perceive a condition, no health action would ensue if he/she did
not anticipate that the health action would be beneficial.
34
Finally, the least consistent predictor of health actions was the
dimension of perceived seriousness of an illness. Janz and Becker
demonstrated that perceived seriousness is a more important
dimension for individuals who are already defined as sick rather
than individuals practicing preventive health actions.
Although illness reporting in a prison clinic is primarily
illness behavior, it also involves health behavior and sick role
behavior. This is because, unlike in the free community, prison
clinic and health care systems are an inmate's only opportunity to
acquire medical prophylaxis and receive medical attention. There
is little opportunity for "home remedies" and little opportunity
to seek alternative private or public sources of medical care.
An indication of the different functions of prison and free
community health clinics is the difference in the individual
utilization rates of inmates and citizens seeking health care.
Most estimates of the community clinic utilization rates place
individual usage rates at between 1 to 2 times per year (U.S.
Department of Health and Human Services, 1980). Estimates of
clinic utilization rates in a prison system average 6 to 7 times
per year and for those inmates who use the clinic, as much as 11
to 12 times per year for voluntary visits. There is the
possibility that some of the utilization differences are due to
differences in the health of free and imprisoned populations.
Table 2 represents information on the most frequently
occurring categories of complaints for the same sample of inmates
35
represented in table 1. The most commonly occurring contagious
illnesses were cold or flu symptoms (18.2 percent of all reported
symptoms). The most commonly occurring noncontagious symptoms
were neuroses/ psychoses (7.0 percent) and back pain (6.9
percent). As can be seen from table 2 most of the illness reports
concerned relatively minor, non-virulent symptoms.
Thus, although the analytical distinctions among health,
illness, and sick role behavior may be more important in a less
constrained environment, there are other analytical distinctions
that are more important in a prison. These are explained in the
following sections.
Factors Affecting Illness Reporting
Prison Housing, Crowding, and Illness Reports
Most of the evidence involving individual housing units
within a variety of Federal prisons has demonstrated that open
dormitory housing of 20 or more inmates produces twice as many
illness complaints as single- or double-cell housing (Paulus,
1988). While double-bunked cells or rooms are rated negatively by
inmates on other scales, such as degree of crowding relative to
singles, they do not consistently produce elevated illness
reporting rates. These findings concerning housing variations,
36
social density, and illness reporting have been interpreted as
reflecting the health consequences of the stress produced by
prolonged exposure to prison crowding (Cox, Paulus, McCain, and
Karlovac, 1980). I believe this presumption is in error and needs
to be examined more closely. The first step in dissecting this
presumption is to place the clinic visit in the broader context of
illness behavior.
One problem associated with the use of illness behavior is
that it involves a decision about one's health in addition to
judgments about the adequacy of diagnosis, the adequacy of
treatment, and the costs incurred in reporting health problems
(e.g., loss of wages, accusations of malingering). This is
particularly telling for crowding research if the decision to use
the clinic is associated with differences in the housing densities
of the inmates. As an example, in most prisons, inmates occupying
single cells have the most preferred housing in a prison.
Assignment to single cell housing is somewhat of a privilege,
based on a combination of tenure and good behavior. Inmates
housed in these preferred conditions may be less likely to
complain about a problem, including their health, in order to gain
the acceptance of prison administrators. Alternatively, inmates
assigned to single cells may be more likely to take better care of
themselves. In either case, assignment to single cells would be
confounded with health actions.
37
Illness Versus Psychological Problems
Many of the factors that produce variations in illness
reporting in settings other than prison are probably also
operating in prison. As Mechanic and his colleagues have pointed
out, illness behavior has an important psychological component
(Mechanic, 1980; Tessler and Mechanic, 1978; Tessler, Mechanic,
and Dimond, 1976). Patients respond to their overall sense of
well-being, not just to specific symptoms. Psychological distress
is just as potent a predictor of health care utilization as is the
patient's perceived health status. A clinic visit may represent
the inmate's attempt to alleviate distress that may have an
associated medical complaint, may be the cause of a medical
complaint, or may even be the result of a medical complaint. In
this sense, a clinic visit may represent the seeking of both
medical and psychological intervention.
There are other examples in the health literature in which
researchers have found psychological variables to be related to
clinic visits. Gortmaker, Eckenrode, and Gore (1982), in a
prospective study of women registered at a neighborhood health
center, found that stressful life events were related to the
probability of using the clinic. The Gortmaker et al. study also
replicated a study by Roghmann and Haggerty (1972) in which a
diary was used to measure daily stress in relation to clinic use.
38
Both studies found that daily stressors were predictive of clinic
use, while the Gortmaker et al. study also showed that the
contribution of daily stressors was independent of that of
negative life events.
Free-Community Determinants of Physician Utilization
In addition to psychological variables, there are many other
factors which determine illness reporting rates. Mechanic (1979,
1989) has reviewed both large-scale multivariate studies of
physician utilization and smaller scale studies. He found that in
the large-scale studies, those variables which measure the extent
of medical need, either actual or perceived, account for most of
the variation in physician utilization. Need is operationalized
as the extent to which people worry about their health, the
symptoms they are experiencing, or the number of bed days they
report. Smaller scale studies often find other factors such as
demographics, beliefs about medical care, and enabling
characteristics such as income and availability of the clinic to
be influential variables.
Determinants of Prison Illness Reporting
39
Suls, Gaes, and Philo (1989) measured negative life events
among a sample of 151 inmates in a Federal prison. Negative life
events were measured by a series of questions that asked inmates
about significant occurrences in their lives. Typical events
include the death of a loved one, financial loss, sickness in the
family, and separation from one's family. In addition to this
traditional set of questions, Suls, Gaes, and Philo developed a
set of questions specific to the prison environment. These items
included problems with parole, altercation with an officer,
problems with case appeals, and grievances against the prison
administration. Illness reports were measured at two different
time frames in this study: concurrent and prospective. The
concurrent time frame occurred in the same 3-month period as the
recall of significant life events. A prospective time frame was
the 3-month period immediately following the survey
administration.
Results of this study showed that negative life events
predicted both concurrent and prospective prison clinic
utilization rates as well as they predicted self-reported
symptoms. Thus, psychological distress is an important antecedent
of both clinic utilization and self-reports of symptoms. Suls,
Gaes, and Philo (1989) have also found that inmates have different
attitudes about their dependence on medical expertise. Adapting
the behavioral component of a self-treatment instrument developed
by Krantz et al., (1980), Suls, Gaes, and Philo developed
40
generalized and prison self-care subscales. The generalized
subscale elicits attitudes about a person's disposition to treat
medical problems without health care intervention. The prison
self-care subscale elicits the inmate's reliance on and
satisfaction with the health care he can receive at the prison
clinic. Only the prison self-care subscale predicted concurrent
and prospective clinic illness reports. These results demonstrate
that part of the decision to use the prison clinic depends more on
the specific trust in treatment of that particular clinic rather
than some generalized expectation about medical care and
expertise.
The failure to find a relationship between generalized
attitudes about health care intervention and clinic utilization
has also occurred in community studies of physician utilization
(Mechanic, 1979; Sharp, Ross, and Cockerham, 1983). Sharp et al.
(1983) demonstrated that neither a generalized attitude toward
physician visits nor symptoms by themselves predicted utilization;
however, symptoms that were rated as serious enough to warrant a
physician visit were related to the physician utilization
variable.
Schaeffer, Paulus, Baum, and Gaes (1988) measured
catecholamine levels (epinephrine and norepinephrine) in a sample
of inmates who volunteered to have their urine tested. Elevated
catecholamine levels have been associated with stressful
environments (Baum, Gatchel, and Schaeffer, 1982). Although this
41
study concluded that dormitories produced elevated levels of
catecholamine, the sample of inmates housed in single cells was so
small (n=9), it is difficult to be confident in the contrasts
between inmates housed in dormitories and double cells (n=84)
versus the inmates housed in single cells.
Self-reported symptoms and clinic illness reports were also
assessed for these same inmates. Disregarding type of housing,
the correlations among some of the variables measured in this
study appear in table 3. Time in unit was the number of weeks an
inmate reported he had spent in the particular housing unit he was
in at the time of the study. Time served was the amount of time
in weeks the inmate reported he had spent in his current
incarceration. Perceived crowding refers to the self-report of how
crowded the inmate felt in his particular housing unit (rated on a
5-point scale), and perceived control was a series of items
assessing the inmate's self-perceptions of control over the prison
environment.
Table 3 shows that neither the frequency nor the rate of
illness reports by inmates were related to self-reports of
symptoms in a concurrent time period. This lack of a relationship
between self-reports and clinic visits was also found by Suls,
Gaes, and Philo (1989). In this latter study, self-reported
symptoms were unrelated to the rate of clinic visits in both a
concurrent and prospective time period.
42
In Schaeffer et al., illness reporting was also unrelated to
biochemical markers of stress (epinephrine and norepinephrine
levels); however, illness reporting was related to time served in
the inmate's current incarceration--the longer the current
incarceration, the less likely it was that an inmate would make a
voluntary visit to the health clinic. This latter finding has
also been observed by Paulus with regard to time in housing with a
much larger sample. Illness reporting was related to perceived
control in the direction expected and marginally related to
perceived crowding in an unexpected direction. Inmates expressing
less control were more likely to use the prison clinic; however,
contrary to expectations, inmates reporting greater crowding used
the health clinic less often. Typically, perceived crowding is
unrelated to illness reports.
Together, the correlations in table 3 suggest that illness
reporting at the prison health clinic was more influenced by
psychosocial factors than it was by biological factors.
Furthermore, the biological stress markers were unrelated to both
self-report of symptoms and clinic visits, whereas these markers
were related to time in prison and perceptions of control and
crowding. If the stress markers were indicative of long-term
exposure to crowding, one would expect these variables would be
related to both the crowding and illness variables. This is one
of the few strong tests of the assumptions that crowding is
related to stress and that a biological component of stress is
43
related to health impairment. The data do not support this assumed
relationship. Instead, the data lend credence to an adaptation
effect. The longer an inmate has been incarcerated, the lower the
level of both biological stress markers and behavioral measures of
clinic use.
One problem in comparing the self-report of symptoms and
actual clinic visits, is that the former depends on recall
(typically 3-6 months), while the latter does not. Verbrugge
(1980) has reviewed the literature on health diaries and found
that the diary produces higher reporting rates for both acute and
chronic illnesses. This is taken as evidence that retrospective
self-reports are subject to recall errors typically under-
representing symptoms and disabilities. Thus, symptom reporting
is subject to greater error than health actions, unless the health
diary is used to record symptoms.
The Strength of the Crowding-Illness Reporting Relationship
Theoreticians and methodologists alike have cautioned against
using statistical variance explained as a criterion for assessing
the importance of empirical relationships (Lieberson, 1985;
Singer, and Marini, 1987). The purpose of this section is not to
dismiss variables based on their relative contribution to the
explanation of illness reporting, rather it is to point out what
may or may not be accomplished from a policy perspective. This
44
assumes that the structural relationships that have been studied
are well specified and do not omit important variables that are
correlated with illness reporting (Arminger and Bohrnstedt, 1987).
With well-specified relationships, we might then assess the degree
to which we can anticipate how manipulation of certain variables
might allow us to intervene in the health/illness reporting
process. It is the contribution of crowding relative to other
variables that concerns us, rather than the total amount of
variance explained.
In order to "benchmark" the strength of the crowding-illness
reporting relationship, two household density studies should be
mentioned. The first was conducted in Toronto (Booth, 1975), the
other in Chicago (Galle and Gove, 1972). Unfortunately, these
studies were discrepant in their results. The Chicago study
showed evidence of household density effects while the Toronto
study did not. The relationship in Chicago was weak; however,
Galle and Gove (1972) reported the effects of household density in
a hierarchical presentation by first demonstrating the variance
explained in their regression results when various background
factors were related to measures of illness and psychopathology.
The reduction in error gained by adding various household density
scales was on the order of 1- to 2-percent additional variance
explained.
In Mechanic's review of large-scale physician utilization
studies (1979) cited earlier, he reported that between 12 and 25
45
percent of the variation in physician utilization was explained by
the different variable sets in the various studies. Of this 12 to
25 percent, illness variables accounted for almost 85 percent of
the explained variance, with demograhic and other variables
accounting for the rest. Thus, the major determinant of a visit
to a physician was the degree to which an individual perceived his
illness to be serious enough to warrant a visit. Neither
demographic or enabling factors such as income or insurance
availability were as influential as the individual's perception of
seriousness. Although perceived seriousness seems to be the
critical factor for free community clinic use, this construct has
never been tested for its influence on prison clinic use.
Paulus (1988) has conducted an analysis of the data collected
on the 1,199 Federal inmates described throughout this paper in
which he looked at a number of variables in relationship to clinic
use. The relationship between different illness reporting
measures and back-ground variables was computed. Then the
additional variance contributed by inmate housing density was
computed. The background variables were parental occupation,
parents' high school graduation, home size, highest grade
completed, SAT score, Beta IQ score, size of hometown as a child,
size of hometown as an adult, prior prison commitments, duration
of prior commitments, current level of custody, months left to
serve, weeks in housing, weeks in prison, weeks in present
sentence, height, and weight. The multiple R-squared between
46
these variables and illness reporting was .03. When both age and
housing type were added, the multiple R-squared was .06.
Illustrating the comparatively small impact that housing type
had on illness reporting, Paulus reported results for other health
and subjective well-being criteria based on a series of
regressions that combined background and housing variables in
addition to the variables outlined below. In those regression
equations in which the housing variable was significant, Paulus
found the following R-squared values: perceived crowding, .33,
room rating, .29, diastolic blood pressure, .13, systolic blood
pressure, .07, self reporting of headaches, .23, and reported
crowding tolerance, .26.
These data indicate that when crowding is measured as a
contrast between dormitory and other kinds of prison housing, the
influence on illness reporting is rather weak. By contrast, the
negative life events scale used by Suls, Gaes, and Philo increased
the R-squared by .13 above the effects of race, age, time served,
disposition to use the prison health clinic, and self-reported
symptoms. Gortmaker et al. also found that negative life events
increased the R-squared by .14 after controlling for other
variables. Although it is inappropriate to compare these three
studies since the specifications for the regressions in each study
were different, there is consistency in the findings. The
contribution of crowding is much less important than the
measurement of negative life events and other psychosocial factors
47
examined in these two studies. The Mechanic review also points out
that the perceived seriousness of symptoms is more important than
other demographic or enabling factors. Future research should
identify whether the free and prison community factors determining
clinic use are the same. It is plausible that they are not the
same. The costs associated with free community clinic use are
probably greater than those associated with use of the prison
clinic. Public clinic use usually involves a long commute and long
waiting period once you are there. Neither of these factors enters
into the decision to use a prison clinic.
Adaptation Effects With Respect to Illness Reporting
Paulus (1988) has shown that differences in illness reporting
that are related to housing assignment begin to converge by the
6th week of confinement, and by the 4th month the illness
reporting rates among inmates housed in dormitories, double-bunked
cells, or single- bunked cells are practically identical (see
Paulus, 1988, table 4-10, p. 38). This data showed that for
inmates who remained in a dormitory for at least 3 months, the
rate at which they used the prison clinic is no different than
inmates in the other, more preferred types of housing, and was
actually lower than inmates in preferred housing who had recently
been assigned to that unit. Subjectively, inmates in dormitories
and double-bunked cells still rated their housing as more crowded
48
and rated the overall housing environment less favorably than
inmates in single-bunked cells. This evidence of an adaptation
effect with respect to illness reporting was cited before with
respect to biological stress markers. This adaptation effect
occurs for all of the housing types. Illness reporting for
inmates housed in single cells had a rate of .31 visits per week
in the first 5 weeks and dropped to .14 visits per week by the
16th week. For double-bunked cells, the respective rates were .38
and .16, and for dorms the rates were .81 and .18 . Clearly,
inmates use the prison health clinic more often when they are
first assigned to a housing unit.
Because of the problem of assignment bias (considered in
detail below), one must be cautious in interpreting the absolute
levels of illness reporting in relation to housing assignment.
Furthermore, this kind of longitudinal data involves a great deal
of retention bias due to the fact that the inmate sample size
decreases as time in housing increases. Inmates who are retained
in the dorm setting may be very different from inmates retained in
the more preferred housing. Reassignment from a dorm to more
preferred housing is typically associated with good conduct and
tenure, while reassignment to less preferred housing is associated
with misconduct or special treatment needs. Consistent with the
adaptation effects in Paulus' data are data reported by Ostfeld,
Kasl, D'Atri and Fitzgerald (1987). Ostfeld et al. were primarily
concerned with blood pressure measure over time; however, their
49
study not only shows an adaptation as measured by blood pressure
within hours and days of confinement (pages 97,98), it also shows
adaptation with respect to clinic use (page 155) over four weeks.
Ignoring the effect of housing for the moment, what are the
causes of this apparent adaptation effect for illness reporting?
There are at least two plausible reasons why inmates use the
prison clinic more often during the initial stages of their
confinement. First, many incoming inmates have health problems,
particularly drug-related illness. As reported by BJS (1988),
42.7 percent of all State prisoners admitted to using some kind of
drug on a daily basis in the month prior to their conviction
offense. Second, moving to a new housing unit or a recent
imprisonment may create psychological distress. Both of these
factors may be unrelated to the kind of housing to which an inmate
is assigned; however, to the extent dormitory inmates are more
likely to be recent arrivals than are double- or single- bunked
inmates, the observed relationship between housing and illness
reporting may be nothing more than an artifact of preferential
housing assignment.
Cross-Sectional Versus Longitudinal Studies and the Nature of
Selection Bias
Most of the data that have demonstrated increased illness
reporting in dormitories have involved cross-sectional
50
comparisons. In some instances, other variables have been
introduced to control for the differences in populations between
the dormitory- and single-cell housed inmates. Paulus (1988) has
claimed that the prison sites he and his colleagues chose were
selected for their variety in housing to insure contrasts between
single- and double-bunked cells and dormitory housing. By
optimizing the difference among types of housing, Paulus and his
colleagues may have inadvertently enhanced the possibility of
selection bias. In those sites where a variety of housing
existed, preferential assignment of housing was most likely to
occur. Paulus has acknowledged this possibility; however, he has
claimed that either assignment bias did not occur, or that such
bias could be evaluated independently. However, there is strong
evidence in Paulus' data that assignment bias did occur, and was
not adequately handled in either the methods of sampling or
statistical control.
Table 4-8 in Paulus (1988, 36) displays the results of
analysis of variance, contrasting characteristics for inmates
housed in singles, doubles, and dorms for the 1,199 prisoners on
whom the data were collected. Inmates housed in single-bunked
cells had spent 100-percent more time in any prison, 150-percent
more time in the current prison, and 50-percent more time in their
current housing than inmates housed in dormitories. Single-cell
inmates also had more time left to serve, a longer previous
confinement, and more previous commitments and were housed at a
51
lower level of custody than inmates housed in dormitories. The
picture that emerges from this data is that of inmates in single
cells who had "earned" preferred housing, had more extensive
criminal histories, and had by their behavior met requirements
that made them less prone to escape or assault than their
dormitory counterparts. These are relatively unmistakable signs
of self-selection or assignment bias.
Paulus and his colleagues (McCain, Cox and Paulus, 1980;
Paulus, 1988) have made an effort to statistically control
assignment through covariance or regression analysis. However,
covariance or regression analysis will not remove bias in two
different populations if there is a self-selected or assigned
difference in the populations and the variables used to adjust for
the bias are themselves affected by the assignment process.
Wainer's (1986) example of the distribution of SAT scores for
different States showed that when an adjustment is made with a
variable that is also affected by a treatment variable, (the
treatment variable in this case was the State in which the SAT
tests were taken), covariance adjustment does not remove the
assignment or selection bias. Analogously, adjusting for time in
housing or other variables, which may be a result of the housing
assignment process, will not produce an unbiased estimate of the
variable of interest.
Because of the strong relationship between time in housing
and illness complaint rates, the elevated illness complaining in
52
dormitories found by Paulus and his colleagues could be merely an
artifact based on the assignment of dormitory inmates, who have
spent much less time in the institution or housing unit. If we
are to have any confidence in future cross-sectional comparisons,
some quasi-experimental design or statistical control for
selection bias should be used. Either Heckman and Robb's (1986)
approach or the prospective method suggested by Rubin and
Rosenbaum (1984) using a propensity score will help control for
selection bias.
Because cross-sectional studies of prison housing suffer from
this methodological problem, longitudinal studies in which the
same inmates experience different housing or different density
conditions over time may prove to be more insightful. There have
been relatively few studies of this kind and the few that have
been done are very limited in their scope. Wener and Keys (1986)
examined two housing units which were nearly identical except for
their respective population levels. Housing unit A went from an
average daily population of 64 inmates to 56 inmates. Unit B went
from 48 inmates to 56 inmates. Although unit A did not show a
decline in sick call requests as would be expected if density was
causing elevated sick call, unit B did show an increase in sick
call requests. Pellisier (1988) examined sick call rates before
and after an institution increase of from 330 to 653 inmates. The
sick call rate declined after the population buildup. There were
some changes in the makeup of the population; however, the changes
53
in offense distribution and average security level should have had
no effect on sick call. The population actually aged somewhat and
this would have been expected to increase sick call rates.
Ostfeld, Kasl, D'Atri and Fitzgerald (1987) have written an
expansive description of their longitudinal and cross sectional
research. In addition to studying the relationship between
housing and blood pressure, these researchers have detailed an
analysis of occupational stress and blood pressure among
correctional officers. They also have written an extremely
insightful chapter on the health status of inmates. The health
status chapter assesses prevalence rates for intake disorders as
well as clinic use over time. Unfortunately, there is no
reference to the relationship between crowding, blood pressure,
and these latter measures of health status in this chapter. It is
unclear whether this was a purposeful omission, or whether the
authors simply found no relationship between blood pressure,
crowding, and/or clinic use over time.
Most of the book presents a detailed analysis of the
longitudinal assessment of blood pressure in relationship to many
covariates including obesity, age, race, education, marital
status, nativity, religion, religiosity, number of children, prior
conviction, residential history, parental relationship, housing
mode, work release, furloughs, other situational factors, and
psychological factors. I will spend some time on this study for
two reasons: (1) to my knowledge, there is no critical review of
54
the study, and (2) the study is already being cited in secondary
sources as evidence of at least short-term effects of crowding.
Bonta and Gendreau (in press) in a forthcoming review of the
effects of imprisonment, examined the effect sizes of the data
reported by Ostfeld and his colleagues as well as other research
results. Bonta and Gendreau concluded that overall, the research
data indicated a short-term effect of crowding, but not a long-
term effect. Inspection of the data shows that this conclusion
was primarily due to the effects reported by Ostfeld and his
colleagues.
Ostfeld et al. (1987) cite the following strengths of their
longitudinal study : (1) the participation rate was over 90
percent and thus reduced the possibility of bias introduced by
volunteering participants; (2) a longitudinal design allows one to
closely examine changes for the same inmate over time increasing
the likelihood of detection of an effect; (3) blood pressure is a
non self-report measure unlikely to be biased by other subjective
impressions. To this list of advantages, the authors cite one
major disadvantage, the longitudinal study was based on one, not
very crowded, medium security, white dominated prison consisting
of for the most part young, uneducated inmates.
To this one limitation, I would like to add some other
important limitations that from my point of view, severely
restrict the usefulness of the study. The major lesson to be
learned from the Ostfeld et al. data collection effort is that
55
although a researcher may begin with a clean, crisp, quasi-
experimental approach to a longitudinal design, the vagaries of
prisoner movement (death, escape, transfer, release) confound the
researcher's design. Although the researchers started with over a
90 percent participation rate, the rate shrank precipitously the
longer the time frame of the study. Ostfeld et al. presented a
figure (figure 6.1, page 77) showing the initial longitudinal
sample and the attrition rate of subjects over time. Of the
initial 568 inmates out of 612 eligible (92.8 percent), by the
second month there were 436 inmates remaining (71.2 percent of the
original eligible pool of 612) and by the sixth month there were
only 181 inmates (29.6 percent). In the longitudinal analyses,
that examined the same inmates from week one through two months,
there were only 165 inmates (27 percent) with complete data and
for a three-month interval there were only 116 inmates (19
percent). There is no attempt in the analysis either to control
for the differences in these censored observations relative to the
other inmates, or even depict the possible similarities or
differences between the participants and non-participants at each
stage.
A second problem with the study is the incompleteness of
their quasi-experimental design. Although the authors do not call
their design quasi-experimental, many of their analyses are
analyses of variance, especially those pertaining to housing and
blood pressure as if the study were quasi-experimental. The major
56
conclusions regarding crowding are based on changes in blood
pressure that result from a transition from one type of housing to
another. There were basically four types of housing studied by
Ostfeld et al.: cells, dormitories, work release dormitories and
work release cells. Table 4 depicts the possible combinations
resulting from a cross-classification of these four types that
would be necessary to assess the different housing transitions.
Only those cells containing an asterisk were actually examined in
this study. This was because it was prison policy to put all newly
admitted inmates into cells and transfer them to the "more
preferred dormitories" later. According to the authors, transfer
was based on good conduct, longevity, and whether there were
population pressures in the cell blocks. Thus, the transfer
decisions imply selection bias.
Although the authors measured a lot of background variables
in this study, it is not clear that in the analyses of housing
effects on blood pressure, the author's simultaneously tried to
control for these characteristics. There is a section in the book
(page 117) that describes an attempt to examine confounding
problems. The authors note that there were differences in the
average age and obesity of men assigned to the different housing
units (variables which have a dramatic influence on blood
pressure); however, Ostfeld et al. claim a covariance analysis
showed these influences did not account for the housing effects.
Other differences in the covariates among the inmates assigned to
57
different housing units were also addressed by the authors
individually; however, it is not clear they were addressed
simultaneously. Furthermore, as I have already argued, selection
bias cannot be controlled in this way.
Even the authors admit that the blood pressure results they
do find are rather small. I would add that given the
incompleteness of their design, the results are almost impossible
to interpret meaningfully. As can be seen from table 4, there is
no "group" which began in a dormitory and transferred to a cell or
remained in a dormitory. Although inmates showed a slightly
elevated systolic blood pressure when they transferred from a cell
to a dormitory, while inmates remaining in a cell or transferring
to a work release cell showed no change or a slight decline, there
is no evidence available to examine whether inmates initially
assigned to a dormitory and latter transferring to a cell versus
inmates remaining in a dorm would not have exhibited precisely the
same pattern of effects.
Finally, as I have already mentioned, Ostfeld and his
colleagues measured blood pressure in both a cross-sectional
design (D'Atri and Ostfeld, 1975) and a longitudinal design
(D'Atri, Fitzgerald, Stanislov, Kasl, and Ostfeld, 1981; Ostfeld,
Kasl, D'Atri and Fitzgerald, 1987). The same investigators get
dramatically different results when using cross-sectional and
longitudinal measures. The authors' explanation is that blood
pressure was measured in a more benign environment in the
58
longitudinal study suppressing the effects of crowding. One would
further have to accept an assumption that the suppression of blood
pressure would be more pronounced for inmates housed in cells than
inmates housed in dormitories, muting the differences.
I think a more realistic explanation is that the cross-
sectional results were at least partially attributable to
selection bias and the longitudinal design, although not
eliminating selection bias, at least observed the same inmates
when they switched housing modes. There are many problems with
the use of blood pressure as a measure of health (see Pickering,
Phil, James, Boddie, Harshfield, Blank, and Largh, (1987);
however, the magnitude of differences between the cross-sectional
and longitudinal effects are instructive. In the cross-sectional
design, dormitory inmates had a higher systolic blood pressure of
between 15 and 21 mmHg, and a higher diastolic blood pressure
between 7 to 11 mmHg. In the longitudinal design, there were no
differences in diastolic blood pressure when inmates changed
housing, and when inmates changed from a cell to a dorm, the
increase in systolic blood pressure was closer to 3.6 mmHg. In
the longitudinal design, Ostfeld et al. also reported on 42
inmates who transferred from a cell to a dorm and, although there
was an initial increase in blood pressure, after 1 month the
systolic blood pressure decreased by almost 50 percent. The
blood pressure results present a very confusing picture of density
relationships; however, one reason why the cross-sectional and
59
longitudinal results are so dramatically different is that the
cross-sectional results are probably attributable to selection
bias. Unfortunately, it will take a systematic study of a
combination of biological markers, illness complaints, and
objective health assessments before any meaningful assessment of
the density-health relationship can be made. The Ostfeld et al.
study is extremely useful in its systematic analysis of health and
background factors. Unfortunately, it also should serve as a
warning to researchers that a quasi-experimental, longitudinal
design will be extremely difficult to accomplish in a prison
setting. Clearly, one should start with a more stable population
than the one Ostfeld et al. had an opportunity to examine.
Summary of Crowding and Illness-Reporting
The use of illness reporting as a veridical indicator of
health may be misleading. The data previously reviewed indicate
there may not be a very good relationship between the occurrence
of illness and the choice to use the prison health clinic. If we
assume the self-reporting of illness to reflect actual health,
there is little or no relationship between self reported illness
and illness reporting in a prison clinic. Rather than assume that
the self-report of illness is the veridical indicator of health,
one approach may involve medical examinations of inmates by health
personnel "blind" to the inmates housing type. Booth's (1975)
60
household study included such an evaluation of Toronto residents.
The medical checkup also included a blood assay for biochemical
stress markers. There were no consistent findings resulting from
this examination; however, the level of crowding experienced by
Toronto residents may not be as deleterious as the level of
crowding experienced by prisoners.
This objective health assessment approach should establish
the level of health prior to residence in a particular housing
environment and remeasure health states after the inmate has lived
in the new environment. Furthermore, because time in prison and
time in housing are highly related to health measures and housing
assignment, an experimental design or quasi-experimental design in
which new commitments are assigned to different housing types may
be necessary to separate the influences of housing from these
time-related variables. Two quasi-experimental approaches to
control for the effects of selection bias have been proposed.
Rubin and Rosenbaum (1984) have shown how selection bias can be
controlled by prospectively selecting a comparison group using a
propensity score. The propensity score summarizes statistical
similarities (and dissimilarities) between a study group member
and the comparison group member. Heckman and Robb (1986) have
shown how selection bias can be controlled through statistical
modeling. Any longitudinal design should also include a follow-up
period long enough to determine if effects diminish over time.
61
The research by David Mechanic and his colleagues also shows
that illness reporting may be indicative of psychological distress
and that people mislabel a variety of psychological states as
illness. This may be especially true for inmates who are newly
imprisoned, or are new to a housing assignment, resulting in
increased somatic complaints. However, as Costa and McCrae (1987)
have shown, somatic complaints are not isomorphic with objective
indicators of health. Costa and McCrae's review of the
relationship between neuroticism and disease has implications for
the crowding-illness reporting literature. Costa and McCrae have
shown that neuroticism is associated with elevated somatic
complaints, but is not associated with increases in objective
measures of illness or disease. This implies that rather than
reflecting some personality dimension, elevated symptom reporting
in dormitories, if it is not an artifact of assignment bias, may
indicate only a heightened awareness of somatic complaints. Once
again, the methodological solution is to measure objective
indicators of disease or morbidity over time while controlling for
assignment bias.
Although Paulus (1988) has discussed other variables in the
context of crowding, these factors have been exogenous variables
used to control for potential differences among prisoners housed
in different housing units (race, parents' occupations, highest
grade completed, gender, ethnicity, criminal history); intervening
background variables that might explain reactions to crowding
62
(crowding tolerance, housing preference, home size); or
intervening prison variables that might explain reactions to
crowding (length of confinement, custody). The emphasis has been
on variables that modify the crowding-illness or crowding-
psychological relationships.
Paulus' approach is important if we are to thoroughly test
the effect of crowding. From this author's perspective, it is
time to move on to other prison stressors that probably have more
of an influence on the health of the inmate. Some of these
variables have already been mentioned: stressful life events, the
prisoner's disposition toward medical care, and the health care
delivery system. There are many other variables in a prison
setting that are crucial to corrections, yet have been ignored for
their relative contribution to inmate health and well-being.
These include prisoner classification, correctional officer
training and inmate-officer interactions, equity and justice in
inmate discipline, prison amenities such as a the library,
recreation yard, and other leisure time activities, and the basic
necessities of sanitation and nutrition.
Evans and Cohen (1987) have outlined a typology for
environmental stress which provides a theoretical vantage point
from which we can analyze the different levels of prison
stressors. Evans and Cohen discuss four types of stressors:
cataclysmic events such as war, floods, or imprisonment; stressful
life events such as the death of a loved one; daily hassles such
63
as irritating events at the office; and ambient stressors such as
chronic air pollution. Although from Evans and Cohen's
perspective, prison is a cataclysmic stressor, there are also more
mundane stressors encountered in prison, characteristic of the
other levels of stress. Some of these were discussed in relation
to the prison stressful life events subscale used by Suls, Gaes,
and Philo (1989).
One of the reasons why prison might be considered cataclysmic
is because it combines so many sources of stress. How are we,
then, to draw conclusions about the relative contribution of
crowding when so many other sources of stress converge in one
environment? To this point, most of the social science research
has concentrated on crowding to the exclusion of many other
important stressors. As a practical matter, a reduction in
crowding may have little or no effect on any of the other
stressors. In this paper, I have shown that when measures of
crowding are significantly related to measures of debilitation,
the relative percentage of variance explained by crowding is
rather low and that the relationship between crowding and
debilitation is inelastic. A note of extreme caution is in order.
None of these studies have been designed to contrast the relative
influence of different classes of stressors. Secondly, elasticity
and variance accounted for can themselves be rather slippery
statistics. These statistics depend on sound measurement,
homoskedastic relationships, unbiased estimates, and, in the case
64
of elasticity, on the level of the explanatory variable (usually
the mean) at which the relationship is evaluated.
V. Is There Consistency in Crowding Studies?
Paulus (1988) has argued that despite limitations in some of
the prison crowding studies there is consistency across a variety
of settings and disciplines. These settings would include human
laboratory studies, household density studies, and crowding
studies in other applied settings (ships, college dormitories).
Let us first examine the degree of consistency in the published
prison crowding studies, and then return to the question of
generalizability of data from other settings and disciplines.
Prison Crowding and Violence
The studies supporting a crowding-violence relationship were
Megargee (1974), Nacci et al. (1977), Jan (1980), Cox et al.
(1984), Ruback and Carr (1984), and Gaes and McGuire (1985).
Studies failing to find a relationship were Carr (1980),
Farrington and Nuttal (1980), Ekland-Olson et al. (1983),
Ekland-Olson et al. (1986), Dilulio (1987), Pellisier (1989),
Porporino and Dudley (1984), Innes (1987), and Fry (1988). These
65
studies vary a great deal in the units of analysis (aggregate
versus individual), the definitions of density, and attempt to
control for other influences on violence. A meta-analysis of
these studies would be inappropriate, because it would unfairly
weight those that are well done from those that are cursory
attempts at finding a relationship.
Prison Crowding and Health
Paulus (1988) reviews data he and his colleagues collected
examining the relationship between total population over time and
death and psychiatric commitment rates. These crude aggregations
showed a bivariate relationship over successive years between
population and the morbidity/mortality rates described. These
crude relationships have been criticized before (Gaes, 1985).
There are other similar analyses which contradict these findings
(Ekland-Olsen, Barrick, and Cohen, 1983; Carr, 1980; Lebowitz and
Pospichal, 1979; Greenfeld, 1982; and Fry, 1988). Furthermore,
there are too many other factors that are correlated with the
changing population over time that could be the cause of the
observed bivariate relationship. Although the burden of proof for
establishing what these concomitant changes might be falls on the
critics of these crude relationships, there are examples in the
prison crowding literature which show that what appears to be a
66
crude relationship at the aggregate level is an artifact of the
aggregation process. Ekland-Olsen et al. (1983) have shown that
the relationship between population changes and misconduct for an
entire prison system does not hold when the data are disaggregated
to the individual prisons. Fry (1988) has shown that what appears
to be a relationship between density and sick call rates over time
for an individual prison does not hold when the data are
disaggregated to the level of the housing unit. Furthermore,
policy changes that occurred in the prison over time completely
accounted for the prison and unit level changes in sick call
rates.
In addition to these aggregate data, Paulus (1988) has argued
that the individual data he and his colleagues collected,
primarily in six Federal prisons, fills in this picture of
consistency. These are the same data I have criticized at length
in this paper. Before we can have confidence in the housing
contrasts, it must be demonstrated that: (1) illness reports are
veridical indicators of the general health of inmates living in
different housing units and (2) the housing contrasts are not
merely a result of selection bias.
Prison Crowding and Recidivism
Farrington and Nuttall (1980) found an aggregate relationship
between prison density and recidivism. These data have been cited
67
as yet another density effect. Clayton and Carr (1987) have shown
that they could replicate a density-recidivism effect among
prisons in Georgia; however, further analysis showed that during
times of high density, Georgia practiced a policy of early release
to reduce crowding. This policy resulted in the release of a
higher proportion of younger offenders, who were more likely to
recidivate. Thus, the density-crowding relationship was spurious
and was attributable to an early release policy.
Generalizability of Density Effects
Now, let us return to the issue of the generalizability of
data from other settings. It is clear that the major conclusion
to be drawn from the prison crowding literature is one of
inconsistency. How are we then to relate the prison crowding
literature to studies of crowded laboratories, dormitories,
households, and ships? Paulus (1988) has argued that these
studies have demonstrated adverse effects of crowding in
relatively benign environments, and that crowding in prisons -- a
relatively harsh environment -- should elicit demonstrative
effects. There are three problems with this argument. The first
is that the research on crowding in these other settings is not as
consistent as Paulus claims. This is especially true of household
density studies (Gaes, 1985). Secondly, there are many
68
differences between prisons and these other settings. The most
prominent difference is the nature of the people sharing the
prison environment. Finally, if, as Paulus claims, the combination
of the conditions of confinement and the antisocial nature of the
population should exacerbate the effects of crowding, why isn't
the evidence of the impact of prison crowding more compelling ?
Is it that the effect of crowding is minimized by other, more
important environmental conditions? Perhaps, further developments
in the theory of stressful environments offered by Evans and Cohen
will lead to testable hypotheses concerning the effects of
crowding in different settings. Until that time, we are left with
a perplexing set of results under circumstances that make it very
difficult to draw conclusions for policymakers.
Summary
There is no consistent evidence that crowding produces
short- or long-term impairment of inmates. There remain logical,
empirical, and theoretical problems with the assertions of some
social scientists who claim crowding effects have been demon-
strated beyond reproach. Crowding has proven to be neither a
necessary nor a sufficient condition for the short- or long- term
impairment of inmates. Crowding is not a necessary condition
because there are other stressors in the absence of crowding that
69
can elicit impairment. It is not a sufficient condition because
there is no proven absolute level of crowding that determines
debilitation.
In my review of the research in 1985, I suggested that there
were two basic conclusions that could be reached, based on the
prison crowding literature. The first was that large, open bay
dormitories elicit higher illness reporting rates than single- or
double-bunked housing arrangements. The second was that prisons
with more inmates than their rated capacity are more likely to
have higher assault rates. Based on a reevaluation of the data
and new evidence concerning prison violence, I now think that
neither of those conclusions is valid. The dormitory effects on
illness reporting are probably the result of selection bias, and
the data on aggregate measures of density and violence have shown
that it is more likely that some factor other than crowding (but
possibly correlated with density) is associated with changes in
violence.
Although the prison crowding research effort has not been
that fruitful, we have learned a great deal about how inmates
react to their environment. There are clear adaptation effects.
When inmates are first imprisoned and when they change housing
units, a number of indicators increase and then gradually subside
over time. The adaptation phenomenon can be measured by illness
reports, blood pressure, or other biological stress markers
(catecholamine). Inmates use the health clinic much more
70
frequently at these early levels of exposure. This result is also
consistent with misconduct data reported by Toch and Adams (1989)
for a large sample of New York State prisoners. Inmates in
general, and especially young inmates, were more likely to exhibit
misconduct early in their incarceration. During imprisonment, the
annual misconduct rate dropped dramatically for young inmates (age
21 or younger), dropped not quite as much for inmates of
intermediate age (22 to 30), and was rather constant for inmates
older than 30. Neither Paulus (1988) or Gaes (1982) have found
age to be related to illness reporting rates. Nevertheless, there
is a consistent picture beginning to emerge from different sources
on prison adaptation that deserves further attention.
An adaptation phenomenon has policy implications for both the
courts and prison administrators
4
. The claim that crowding is an
unconstitutional condition of confinement gathers most of its
weight from the implicit assumptions that both crowding and its
consequences are: (1) spread evenly throughout the population and
(2) constant or continuous in nature. Courts may tolerate a
condition that would otherwise be intolerable, if it can be shown
to affect only some prisoners very strongly, and/or with
decreasing effect beyond the first part of their sentences. If
crowding results are an artifact of assignment bias and merely
reflect the adaptation of inmates over time to their conditions,
then adding new inmates to a system will not only raise the level
of density but also the level of violence and clinic utilization.
71
This phenomenon will be transitory. As the population, on average,
increases its time in prison, the levels of violence and clinic
use will decline.
From a management perspective, an adaptation phenomenon
suggests that prison administrators adopt procedures at the
beginning of a sentence that help address the inmate's integration
into prison. This would include more than just the traditional
rule orientation. Orientation to prison might include greater
attention to the psychological and medical needs of the inmate
when he first arrives. There may be a practical limit to the
intervention that can be made when an inmate is first imprisoned.
If a prisoner has medical or nutritional requirements, they will
require attention. Psychologically, the shock of imprisonment may
be difficult to ameliorate.
In closing, it is difficult to review the prison crowding
literature without confronting pervasive, naive beliefs about
crowding. Although I have not systematically gathered evidence, I
believe that the prevailing intuition among people unfamiliar with
the research (and even some who are familiar) is that crowding, by
itself, is prima facie evidence that prison conditions will be
harsh and severe. This view ignores the potential responsiveness
of prison administrators and their supporting agencies to
increases in population. Camp and Camp (1989) have polled
administrators in the United States to ask them the extent of
their crowding and how they manage their prisons under crowded
72
conditions. Among their conclusions, Camp and Camp argue that
"managing a crowded institution is not substantially different
from managing a prison that is not crowded" (p 63). Sound
management is required in both instances. There are many
institutions that are crowded; however, for the most part, those
prisons that are orderly and meet basic inmate needs have avoided
litigation or have won conditions suits brought against them.
Researchers need to bear in mind that prison crowding research is
conducted in an organizational setting -- all the data are
collected in the context of an institutional "culture" and an
administrative "climate." How prison administrators react to
overpopulation may be as important as any variable yet to be
proposed in this domain.
73
References
American Correctional Association, (1992), The 1992 Standards
Supplement, Laurel Md.
Advisory Commission on Intergovernmental Relations (1993), The
Role of the General Government Elected Officials in Criminal
Justice, Washington, D.C.
Alexander, R., (1993), Slamming the federal courthouse door on
inmates, Journal of Criminal Justice, 21, 103-116.
Arminger, G. and Bohrnstedt, G.W. (1987). Making it count even
more: A review and critique of Stanley Lieberson's Making it
Count: The improvement of social theory and research, in C.C.
Clogg (ed.) Sociological Methodology, Vol. 17, 363-372.
Washington, D.C.: The American Sociological Association.
Baum, A., Gatchel, R.J., and Schaeffer, M.A. (1983). Emotional
behavior and physiological effects of chronic stress at Three Mile
Island. Journal of Consulting and Clinical Psychology, 51, 565-
572.
Becker, M.H. (ed). (1974). The health belief model and personal
health behavior. Health Education Monograph, vol. 2, 324-508.
Benekos, P.J., (1991), Public policy and correctional reform:
Politics of overcrowding., Annual Meeting of the American Society
of Criminology, San Francisco.
Bleich, J. (1989), The politics of prison crowding, California Law
Review, 77(5), 1125-1180.
Bureau of Justice Statistics (1993) Prisoners in 1992.
Bureau of Justice Statistics (1993), Advanced Press Release.
Bonta, J. and Gendreau, P. (in press) Re-examining the cruel and
unusual punishment of prison life, Law and Human Behavior.
Booth, A. (1976). Urban Crowding and Its Consequences, New York:
Praeger.
Calhoun, J.B. (1962). Population density and social pathology.
Scientific American, 206(1), 139-148.
74
Call, J.E. (1988). Lower court treatment of jail and prison
overcrowding cases: A second look. Federal Probation, vol. 52(2),
34-41.
Camp, G. and Camp C. (1989). Management of Crowded Prisons, U.S.
Department of Justice, National Institute of Corrections.
Carr, T.S. (1980). The Effects of Crowding on Recidivism,
Cardiovascular Deaths, and Infraction Rates in a Large Prison
System. Unpublished Ph.D. dissertation, Georgia State University.
Christian, J.J. and Davis, D.E. (1964). Endocrines behavior and
population. Science, 140, 1550-1557.
Clayton, O., and Carr, T. (1987). An empirical assessment of the
effects of prison crowding upon recidivism utilizing aggregate
level data, Journal of Criminal Justice, 15, 201-210.
Corrections Digest (1989), Washington Crime News Service, Fairfax
Va.
Costa, P.T. and McCrae, R.R. (1987). Neuroticism, somatic
complaints, and disease: Is the bark worse than the bite?, Journal
of Personality, 55, 299-316.
Cox, V.C., Paulus, P.B., McCain. G. and Karlovac, M. (1980). The
relationship between crowding and health, in A. Baum and J.P.
Singer, (eds.) Advances in Environmental Psychology, vol. 4,
Hillandale, N.J.: Erlbaum.
Cox, V.C., Paulus, P.B. and McCain, G. (1984). Prison crowding
research: The relevance for prison housing standards and a
general approach regarding crowding phenomena. American
Psychologist.
D'Atri, D.A., Fitzgerald, E.F., Kasl, S.V. and Ostfeld, A.M.
(1982). Crowding in prison: The relationship between changes in
housing mode and blood pressure. Journal of Psychosomatic
Medicine, 43, 95-105.
D'Atri, D.W. and Ostfeld, A.M. (1975). Crowding: Its effects on
the elevation of blood pressure in a prison setting. Preventive
Medicine, 4, 550-566.
DiIulio, J.J. (1987), Governing Prisons: A Comparative Study of
Correctional Management, New York: The Free Press.
Ekland-Olsen, S., Barrick, D. and Cohen, L.E. (1983). Prison
overcrowding and disciplinary problems: An analysis of the Texas
75
prison system. Journal of Applied Behavioral Science, 19, 163-
176.
Ekland-Olsen, S. (1986), Crowding social control and prison
violence: Evidence from the post-Ruiz years in Texas, Law and
Society Review, 20, 389-421.
Ekland-Olsen, S. (1986). Crowding, social control, and prison
violence: Evidence from the post-Ruiz years in Texas. Law and
Society Review, vol. 2O(3), 389-421.
Ellis, D. (1984). Crowding and Prison Violence: An Integration
of Research and Theory. Toronto: York University.
Evans, G. W. and Cohen, S. (1987). Environmental stress, in D.
Stokols and I. Altman (eds.) Handbook of Environmental
Psychology, 571-610. New York: John Wiley and Sons.
Farrington, D.P. and Nuttal, C.P. (1980). Prison size,
overcrowding, prison violence and recidivism. Journal of Criminal
Justice, 8(4), 221-231.
Franklin, B.J. and McClemore, S.D. (19 ). Factors affecting the
choice of medical care among university students. Journal of
Health and Social Behavior, 311-319.
Fry, L.J. (1988), Continuities in the determination of prison
overcrowding effects, Journal of Criminal Justice, 16, 231-240.
Gaes, G.G. (1982). The effect of spatial and architectural
housing variations on inmate clinic utilization rates.
Washington, D.C.: Office of Research, Federal Bureau of Prisons.
Gaes, G.G. (1985), The effects of overcrowding in prison, In M.
Tonry and N. Morris (Eds.), Crime and Justice: An Annual Review,
(Vol 6.), 95-146, Chicago: University of Chicago Press.
Gaes, G.G. and McGuire, W.F. (1985). Prison Violence: The
contribution of crowding versus other determinants of prison
assault rates. Journal of Research in Crime and Delinquency, vol.
22(1), 41-65.
Galle, O.R., Gove, W.R. and McPherson, J.M. (1972). Population
density and pathology: What are the relations for man?" Science,
176, 23-30.
Gortmaker, S.L., Eckenrode, J., and Gore, S. (1982). Stress and
the utilization of health services: A time series and cross
76
sectional analysis. Journal of Health and Social Behavior, vol.
23, 25-38.
Gottfredson, S.D. and McConville, S. (Eds.) (1987), America's
Correctional Crisis: Prison Populations and Public Policy, New
York: Greenwood Press.
Greenfeld, L.A. (l982). Prison Population and Death Rates.
Washington, D.C.: National Institute of Justice.
Heckman, J.J. and Robb, R. (1986), Alternative methods for solving
the problem of selection bias in evaluating the impact of
treatments on outcomes, In H. Wainer (Ed.), Drawing Inferences
from Self-Selected Samples, 63-107, New York: Springer-Verlag.
Ingraham, B. L. and Wellford, C.F. (1987). The totality of
conditions test in eighth-amendment litigation, in S.D.
Gottfredson and S. McConville (eds.) America's Correctional
Crisis, 13-36. New York: Greenwood Press.
Innes, C.A. (1987), Population density in State prisons, Bureau of
Justice Statistics: Special Report.
Jan, Lee-Jan (1980). Overcrowding and Inmate Behavior: Some
Preliminary Findings. Criminal Justice and Behavior, 7, 293-301.
Janz, N. K. and Becker, M.H. (1984). The health belief model: A
decade later. Health Education Quarterly, vol. 11(1), 1-47.
Kasl, S.V. and Cobb, S. (1966), Health behavior, illness behavior
and sick role behavior, Archives of Environmental Health, 12, 246-
266.
Krantz, D., Baum, A. and Wideman, M. (1980) Assessment of
preferences for self-treatment and information in health care,
Journal of Personality and Social Psychology, 39, 977-990.
Lebowitz, H. and Pospichal, T. (1979). Federal and State Inmate
Deaths, 1972-1978. Unpublished manuscript. Washington, D.C.:
Federal Bureau of Prisons.
Lieberson, S. (1985) Making It Count:The Improvement of Social
Research and Theory, Los Angeles: University of California Press
Martin, S.J. and Ekland-Olsen, S. (1987). Texas Prisons: The
Walls Came Tumbling Down. Austin, Texas: Texas Monthly Press.
77
McCain, G., Cox, V. and Paulus, P. B. (1980). The Effect of
Prison Crowding on Inmate Behavior. Washington, D.C.: Law
Enforcement Assistance Agency.
Mechanic, D. (1979). Correlates of physician utilization: Why do
major multivariate studies of physician utilization find trivial
psychosocial and organizational effects? Journal of Health and
Social Behavior, vol. 20, 387-396.
Mechanic, D. (1980). THe experience and reporting of common
physical complaints, Journal Of Health and Human Behavior, 21,
597-608.
Mechanic, D. (1989). Medical sociology: Some tensions among
theory, method and substance. Journal of Health and Social
Behavior, vol. 30(2), 147-160.
Megargee, E.I. (1974). Population density and disruptive behavior
in a prison setting, in J. H. Cullen (ed.) Experimental Behavior:
A Basis for the Study of Mental Disturbance. Dublin: Irish
University Press.
Morris, N. and Tonry, M. (1990) Between Prison and Probation:
Intermediate Punishments in a Rational Sentencing System, New
York: Oxford University Press
Mullen, J. and Smith, B. (1980) American Prisons and Jails. Vol
3.: Conditions and Costs of Confinement. Cambridge: Abt.
Nacci, P.L., Teitlebaum, H.E. and Prather, J. (1977). Population
density and inmate misconduct rates. Federal Probation, 41, 26-
31.
Ostfeld, A. M., Kasl, S.V., D'Atri, D.A., and Fitzgerald, E.F.
(1987) Stress, Crowding, and Blood Pressure in Prison, Hillsdale,
N.J. Lawrence Earlbaum Assoc, Inc.
Parsons, T. (1951) Illness and the role of the physician: A
sociological perspective American Journal of Orthopsychiatry, 21,
452-460.
Paulus, P.B. (1980) Crowding. In Psychology of Group Influence.
Hillsdale, N.J.: Erlbaum.
Paulus, P.B., McCain, G. and Cox, V.C. (1978). Death rates,
psychiatric commitments, blood pressure, and perceived crowding as
a function of institutional crowding. Environmental Psychology
and Nonverbal Behavior, 3(2), 107-116.
78
Paulus, P.B., McCain, G. and Cox, V.C. (1981). Prison standards:
Some pertinent data on crowding. Federal Probation, 45(4), 48-54.
Paulus, P.B. (1988) Prison Crowding: A Psychological Perspective,
New York: Springer-Verlag.
Pellisier, B. (1989), The effects of a rapid increase in a prison
population: A pre- and post-test study, Under Review at Criminal
Justice and Behavior, Department of Justice, Federal Bureau of
Prisons.
Pickering, T.G. Phil, D., James, G.D. Boddie, C., Harshfield,
G.A., Blank, S., and Largh, J.H. (1988), How common is white coat
hypertension ?, Journal of the American Medical Association, 259,
225-228.
Porporpino, F.J. and Dudley, K. (1984), An analysis of the effects
of overcrowding in Canadian penitentiaries, Ministry of the
Solicitor General of Canada.
Robinson, W. S. (1950) Ecological correlations and the behavior of
individuals. American Sociological Review, 15, 351-357.
Roghmann, K.J. and Haggerty, R.J. (1972). Family stress and the
use of health services. International Journal of Epidemiology,
vol. 1(3), 279-286.
Ruback, R.B. and Carr, T.S. (1984). Crowding in a women's prison:
Attitudinal and behavioral effects. Journal of Basic and Applied
Psychology.
Ruback, B. and Innes, C. (1988), The relevance and irrelevance of
psychological research: The example of prison crowding. The
American Psychologist, 43, 683-693.
Rubin and Rosenbaum (1984), Reducing bias in observational studies
using subclassification on the propensity score. Journal of the
American Statistical Association, 79, 516-524.
Schaeffer, M.A., Paulus, P.B., Baum, A., and Gaes, G.G (1988),
Architecturally mediated effects of social density in prison.
Environment and Behavior, 20, 3-20.
Sharp, K., Ross, C.E. and Cockerham, W.C. (1983). Symptoms,
beliefs, and the use of physician services among the
disadvantaged. Journal of Health and Social Behavior, vol. 24(3),
255-263.
79
Singer, B. and Marini, M.M. (1987). Advancing social research: An
essay based on Stanley Lieberson's Making it Count, in C.C. Clogg
(ed.) Sociological Methodology, Vol. 17, 373-391. The American
Sociological Association, Washington, D.C.
Stokols, D. and Altman, I. (eds). (1987). Handbook of
Environmental Psychology, Vol. 1. New York: John Wiley and Sons.
Suls, J., Gaes, G.G., and Philo, V. (1989) Life stress and illness
behavior in prison: The role of self-care, attitudes and race,
Unpublished manuscript.
Tessler, R. and Mechanic, D. (1978). Psychological distress and
perceived health status. Journal of Health and Social Behavior,
19, 254-262.
Tessler, R., Mechanic, D. and Dimond, M. (1976) The effect of
psychological distress on physician utilization: A prospective
study. Journal of Health and Social Behavior, 17, 353-364.
Thornberry, T.P. and Call, J.E. (1983). Constitutional challenges
to prison overcrowding: The scientific evidence of harmful
effects. Hastings Law Journal, vol. 35, p. 23.
Toch, H. and Adams, K. with Grant, J.D. (1989). Coping:
Maladaptation in Prisons. New Brunswick: Transaction Publishers.
Useem, B. and Kimball, P. (1989). States of Siege: U.S. Prison
Riots, 1971-1986, Oxford University Press: New York.
U.S. Department of Health and Human Services (1980), Health of the
Disadvantaged, Office of Health Resources Opportunity.
Verbrugge, L.M. (1980), Health diaries. Medical Care, 18, 73-95.
Wainer, H. (1986) The SAT as a social indicator: A pretty bad
idea, 7-22, in H. Wainer (ed.) Drawing Inferences from Self-
Selected Samples, Springer-Verlag : New York.
Wall Street Journal (1981), Life in Prison: Vol CXCVIII, No. 34.
Wener, R. and Keys, C. (1986). The Effect of Changes in Jail
Densities on Perceived Crowding, Spatial Behavior and Sick Call:
Absolute and Contrast Effects. New York: Polytechnic Institute
of New York.
Zamble, E. and Porporino, F. J. (1988) Coping, Behavior and
Adaptation in Prison Inmates. New York: Springer-Verlag.
80
Cases Cited
Wilson v. Seiter, 111 S. Ct. 2321 (1991)
Rhodes v. Chapman, 452 S. Ct 337 (1982)
1
A and B refer to two data collection efforts involving
different inmates at different time periods in the same
institution.
81
Table 1. Summary data on illness reports collected in 6 Federal
prisons in a six month period.
Basic Information
Percentage
Complaint Range 0 to 31
Total Complaints 3,426
Total Inmates 1,199
Total Inmates Using the Clinic 788 65.7%
Total Inmates Not Using the Clinic 411 34.2%
Number Reports Per Inmate 2.86
Number Reports Per Inmate Using
the Clinic 4.35
Number of Clinic Visits per Days
at Risk .032
Distribution of Illness Reports
Number Complaints 0 1 2 3 4 5 6 7 8 9 10 11-31
Number Inmates 411 186 144 127 82 50 48 23 20 21 25 62
Illness Reporting Rate by Institution
Site: Site 1 Site 2A
1
Site 2B Site 3A Site 3B Site 4 Site5 Site 6
Rate: .037 .023 .027 .038 .044 .035 .016 .029
2
The values indicate the percentages of these inmates
who reported a particular complaint. Of the entire
sample 35% of the inmates had no clinical visits.
82
Table 2. Examples of the most frequent contagious and
noncontagious illness symptom classifications and their percentage
of occurrence among inmates.
2
CONTAGIOUS % NONCONTAGIOUS %
Venereal Disease 1.8 Neuroses, Psychoses 7.0
Eye, Ear, Nose, Throat 11.9 Headache 5.2
Infections/Problems Circulatory Heart Problems 2.1
Cough, Cold, Flu 18.2 Joints, Bursitis, Arthritis 4.6
Gastrointestinal, Stomach 8.3 Neck, Shoulder, Hip Pain 5.9
Skin, Subcutaneous 11.3 Chest Pain 2.8
Virus, Chills, Fever 5.2 Back Pain 6.9
83
Table 3. Correlations Among Self Reports, Illness Reports, Catecholamine
Levels and Time in Prison Variables
PCr PCn SS CV UE UN TH TP
Perceived Crowding (PCr)
Perceived Control (PCn) -.25*
Self Reported Symptoms (SS) .24* -.37*
Clinic Visit Rate (CV) -.16 -.24* .12
Urinary Epinephrine (UE) .30* -.15 .12 .03
Urinary Norepinephrine (UN) .31* -.25* .09 .09 .62*
Time in Housing Unit (TH) .01 .18* -.05 -.10 -.10 -.10
Time in Prison (TP) -.18* .15 .01 -.20* -.20* -.29* .64*
* significant at p < .05, n=87
84
Table 4. The Potential Cross Classification of Housing Types in the
Longitudinal Study Reported by Ostfeld, Kasl, D'Atri and Fitzgerald (1987).
Time 2 Housing
---------------------------------------------------------------------
| Work Release Work Release
Time 1 Housing | Cell Dorm Cell Dorm
---------------------------------------------------------------------
Cell | * * * *
---------------------------------------------------------------------
Dorm |
---------------------------------------------------------------------
Work Release Cell |
---------------------------------------------------------------------
Work Release Dorm |
---------------------------------------------------------------------
Note: Only those cells with an asterisk were included in the design used by
Ostfeld et al. The remaining cross classifications were not possible
because of the prison policy on assignment of inmates to housing units.
85
1. This paper does not consider the epidemiological basis of
problems associated with crowding. There are clearly unique
health problems associated with institutionalized persons,
including those in prisons, nursing homes, and shelters. The
relationship between infectious disease transmission and
environmental factors such as crowding and ventilation is also
confounded by such issues as proper medical screening, treatment
protocols, the prevalence of disease in the population, and
policies and procedures related to hygiene.
2. Apparently, reported cases represent only a fraction of all
District Court cases. The decision to publish a case rests with
the presiding judge. The extent to which published cases are
representative of unpublished cases is a factor that might be
taken up by other researchers.
3. The concept of elasticity deals with the percent change in a
dependent variable that is associated with a 1-percent change in
an independent variable. A relationship is inelastic when a good
deal of change in the independent variable is needed to evoke a
change in the dependent variable. An elastic relationship is one
in which little change in the independent variable is needed to
affect the dependent variable. More specifically, if the percent
of change in a dependent variable is less than 1 percent, the
relationship is considered inelastic. When the percent change is
greater than 1-percent, the relationship is considered elastic.
4. My thanks to Charles Logan, who suggested these implications
of an adaptation effect after reading a draft of this paper.
Footnotes
