 

 

. 4.? a

PA.‘ I
11,...
s . .

CL... xx.
..

,.._

t
1

h I
anagram
u,

I?
3.5.... a

. q

. f z.

Walnut}?

a. u.
..

mm ..

3h. .
ﬂ, Maﬁa.
kw

n...

1.9.!

1.55%». w}

. g (in! 4

4. §.: .4!- i 3?. II,
31.93. 1.

.i

r
I:

1: E ..
Ht ‘l

5.1;. ‘. . , w
a. dun...“ V. 5.91:2. .
‘. 1-? I r ..

. x 11‘ e

 

 

This is to certify that the
thesis entitled !

SPECIALIZATION AMONG SEX OFFENDERS

 

presented by

MOHAMMAD VAQAS ALI

has been accepted towards fulfillment
of the requirements for the

Masters degree in , Criminal Justice

  

 

aor B/cfessor's Signature
05 - 11- 04

Date

MSU is an Afﬁnnative Action/Equal Opportunity Instltution

via" ‘ P

, -—.~. 4r

 

 

LIBRARY
Michigan State
University

 

 

 

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

 

DATE DUE DATE DUE DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

6/01 c:/ClRC/DateDue.p65-p. 15

SPECIALIZATION AMONG SEX OFFENDERS
By

Mohammad Vaqas Ali

A THESIS

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

MASTER OF SCIENCE
School of Criminal Justice

2004

‘l’lli ‘llllf'l

ABSTRACT
SPECIALIZATION AMONG SEX OFFENDERS
By
Mohammad Vaqas Ali

In the past 15 years the Federal and state governments have enacted several
statutes that deal speciﬁcally with sex offenders. The contentions of such laws is that sex
offenders specialize and frequently recidivate, and have to be speciﬁcally dealt with. To
date criminologists have conducted few studies that analyze the veracity of this premise.
The primary objective of this thesis is to test this hypothesis i.e. whether sex offenders
specialize or not. Eight hundred males were selected from the population of arrestees
from a mid-size Michigan city who initiated their criminal careers between 1989 and
1999 as juveniles and had one arrest on a violet charge. The conviction records of the
subjects were also obtained from the Michigan Department of Corrections database for
the 10 years understudy. A bimodal approach (Survival and Negative Binomial methods)
was then used to measure two types of specialization. The results showed that sex
offenders were more likely to perpetrate sex offenses in their criminal careers as
compared to other violent offenses. However, sex offenders committed more assaults
during their criminal careers than sex offenses, but did not commit any other nonviolent
crimes before their ﬁrst sex offense. In contrast, other violent offenders committed
nonviolent offenses before their ﬁrst violent crime and with just a few exceptions had not
escalated to committing sex offenses. These characteristics of the sex offenders noted in
this study, render their behavior in explicable by any of the existing theories on the

matter.

Copyright by
Mohammad Vaqas Ali
2004

DEDICATION

Thank you Amma Jee and Abba Jee for always being there and for inspiring me with
your love and prayers.

iv

ACKNOWLEDGEMENTS

I would like to thank my committee members for their guidance, time and patience:
Dr. Christopher Maxwell (Chair), Professor John McCluskey and Dr. Vincent Hoffman.

TABLE OF CONTENTS

 

 

 

 

 

 

 

 

LIST OF TABLES VII
INTRODUCTION 1
THEORETICAL FRAMEWORK 3
THE SPECIALIZATION PERSPECTIVE ON SEXUAL OFFENDERS .......................................... 4
THE GENERALIST PERSPECTIVE ON CRIME .................................................................... ll
LITERATURE REVIEW 14
EMPIRICAL VALUE OF PSYCHOLOGICAL TESTS THAT PROPOSE TO PREDICT
SPECIALIZATION AMONG SEX OFFENDERS ..................................................................... 15
RECIDIVISM STUDIES ON SEX OFFENDERS ..................................................................... 25
EMPIRICAL LITERATURE ON SPECIALIZATION AMONG SEX OFFENDERS ........................ 30
RESEARCH OBJECTIVE 33
METHODS SECTION 33
DATA ............................................................................................................................. 34
VARIABLES .................................................................................................................... 37
Independent Variables. ............................................................................................. 37
Dependent Variables. ................................................................................................ 38
STATISTICAL ANALYSIS ................................................................................................. 39
Model I ...................................................................................................................... 39
Model II ..................................................................................................................... 40
RESULTS 43
UNIVARIATE RESULTS ................................................................................................... 43
BIVARIATE RESULTS ...................................................................................................... 48
MULIIVARIATE RESULTS ............................................................................................... 50
Model 1 ..................................................................................................................... 51
Model 11 ..................................................................................................................... 53
DISCUSSION 59
BIBLIOGRAPHY 69

 

vi

List Of Tables

 

 

 

 

 

 

 

 

 

 

 

 

TABLE 1A 43
TABLE 1B 43
TABLE 1C 43
TABLE 2 46
DESCRIPTIVE STATISTICS AND CODING .......................................................................... 46
TABLE 3 4s
CROSS TABULATION BETWEEN RACE AND TYPE OF 1‘r VIOLENT ARREST ..................... 48
TABLE 4 ‘ 49
COMPARISON OF MEANS BETWEEN TYPE OF 1‘r VIOLENT ARREST AND AGE AT 1"r
ARREST .......................................................................................................................... 49
TABLE 5 50
COMPARISON OF MEANS BETWEEN TYPE OF 1"" VIOLENT ARREST AND THE ARREST AND
CONVICTION MEASURES ................................................................................................. 50
TABLE 6 52
SURVIVAL MODEL FOR THE ARREST MEASURES WITH THE TYPE OF 15'r VIOLENT ARREST
AND RACE ...................................................................................................................... 52
TABLE 7 54
NEGATIVE BINOMIAL FOR THE ARREST MEASURES WITH THE TYPE OF 15'r VIOLENT
ARREST AND RACE ........................................................................................................ 54
TABLE 8 55
POISSON REGRESSION AND NEGATIVE BINOMIAL FOR THE CONVICTION MEASURES WITH
THE TYPE OF 15'r VIOLENT ARREST AND RACE ............................................................... 55
TABLE 9 57
GENERAL LINEAR MODEL FOR THE ARREST MEASURES WITH THE TYPE OF 1‘r VIOLENT
ARREST AND RACE ........................................................................................................ 57
TABLE 10 58
GENERAL LINEAR MODEL FOR THE CONVICTION MEASURES WITH THE TYPE OF 1‘r
VIOLENT ARREST AND RACE ......................................................................................... 58

vii

INTRODUCTION

Throughout the history of criminology and criminal justice, criminologists have
theorized the phenomenon of crime in order to rationally confront it. The very basic issue
that every good criminological theory has to address in order to be complete is to
determine the cause (or causes) of crime. For the past two decades the debate between the
proponents of general and typological theories of crime has often dwelled upon this
theme i.e. whether all crimes occur due to one universal cause or do different types of
deviant behaviors initiate from differing causes (Gottfredson and Hirschi, 1988;
Blumstein, Cohen, Roth and Visher, 1986; Mofﬁtt, 1993; Tracy and Kempf-Leonard,
1996). Gottfredson and Hirschi (1988), adhering to the generalist approach state that low
self-control introduces a criminal propensity in all offenders, which consequently induces
them to commit crimes. Criminals are more susceptible at a younger age to fall prey to
this propensity and its control diminishes as they reach maturity (Vold, Bernard and
Snipes, 2002; Gottfredson and Hirschi, 1988). Blumstein et. al. (1986) on the other hand,
asserts in the criminal career approach that different causal processes work at different
points in a life span and produce same or different types of deviant behaviors (Vold et.
al., 2002; Piquero, 2000). Further development of the criminal career approach
establishes, that different causal processes may also generate different types of deviant
behavior, hence leading to the concepts of specialization and escalation (Blumstein, et. al.
1986).

Specialization and escalation are two essential elements of the criminal career
paradigm (Blumstein, Cohen and Farrington, 1988 and Britt, 1996). In the words of Britt

(1996), “Specialization refers to the tendency among criminal offenders to repeat the

same type of crime across their criminal careers, while escalation refers to the tendency
of some criminal offenders to commit crimes of an increasingly serious nature over the
span of their criminal careers”. Hitherto, almost all the studies conducted by
criminologists that have attempted to operationalize and measure the concepts of
specialization and escalation classify offence types as only “violent vs. non-violent” or
“physical vs. property” offenses (Bursik Jr. 1980 and Piquero, 2000). There are very few
empirical studies that get offence speciﬁc in specialization analysis. Taking the example
of sexual offenses, to date there has been only one study by a criminologist that has
compared the criminal histories of convicted sex offenders and convicted non-sex
offenders to determine whether sex offenders are more specialized (Simon, 1997 &
2000). The indolence displayed by the criminologists when countered by the eagerness
on the part of the legislature to deal decisively with the problem of sex offenses leaves a
lacuna for the academics from other ﬁelds to jump in and try to address the issue in
consistency with their particular disciplines; a path that may have lead to bad policy
making (Simon, 2000 and Palermo and Farkas, 2001).

Traditionally, sex offenders are epitomized as a class of delinquents that is
deemed qualitatively unique from all other varieties of criminals (Simon, 2000 and
Palermo and Farkas, 2001). Terms like sexual predators, pedophiles and rapists, signify a
societal precept of a sadistic and sexually perverted group of offenders that must
continually prey upon more and more victims to gratify their uncontrollable psychotic
impulses (Freeman-Longo, 2000 and Palermo and Farkas, 2001). Albeit, these
conclusions may be empirically justiﬁable among a sample of chronic sex offenders, but

generalizing them over the entire population of sex offenders would be erroneous (Simon

1997 & 2000). Nonetheless, such social precepts coupled with the erroneous
generalizations made by clinicians regarding sex offenders have often been translated
into laws by the legislature (Simon, 1997 & 2000 and Palermo and Farkas, 2001). The
contention of such laws being, that sex offenders ‘specialize’ in sex offenses and have to
be speciﬁcally dealt with as compared to other types of offenders (Simon, 1997 & 2000
& Palermo et. al., 2001). This statement can be practically validated from the fact that in
the past seven decades the Federal Government has enacted three statutes (Sexual
Predator Laws, 1990; Sex Offender Registration and Community Notiﬁcation Laws,
1994; Surgical Castration and Chemical Castration Laws, 1996) that deal speciﬁcally
with sex offenders (Palermo et. al., 2001). Furthermore, all of these federal statutes have
received sweeping support from the state legislatures (Palermo et. al., 2001). Thus far, no
notable empirical studies have been conducted, that analyzes whether these laws have
actually brought any signiﬁcant effect on the general rate of sexual offenses in the
country (Freeman-Longo, 2000).

The primary objective of this thesis is to empirically address the issue of
specialization among sex offenders (from a criminal justice perspective) using the Police
Arrest data from 1990 to 2000 for a mid-sized Michigan city. Before falling into the
speciﬁcs of the data and the statistical tests to be used, it is pertinent that the theoretical
framework underlying the study is established and the ﬁndings of the relevant empirical
literature be chronicled.

THEORETICAL FRAMEWORK
The two fundamental constructs that this thesis is going to be dealing with are l)

the specialization perspective regarding sexual offenders emanating from the criminal

career approach and the typological theories, and 2) the non specialist perspective propos
sexual offenders, stemming from the generalist theories of crime. Each theme has its own
theoretical history and shall be addressed to separately in the this sections.
The Specialization Perspective on Sexual Offenders

Early on in the history of criminology the notion was developing that crime it self
was not an isolated event and could be better understood if looked upon, from the
perspective of the criminals and their criminal histories (Park, Burgess and McKenzie,
1925; Shaw, 1931; Sutherland, 1937; Hughes, 1952 and Becker, 1957). In 1986,
Blumstein, Cohen, Roth and Visher organized and further developed this perspective into
what they called the “criminal career and career criminal career approach”. This
approach, instead of keeping the traditional focus on the aggregate crime rates, analyzed
the activity or careers of the individuals who commit criminal offenses (Blumstein et. al.,
1986). According to Blumstein et. al. (1986) the criminal career approach characterized
the phenomenon of crime into four key dimensions: Participation (the fraction of the
population that engages in criminal activity), Frequency (the rate of criminal activity
among those who participate), Seriousness (of the transgressions committed) and Career
length (time in between the ﬁrst and the last offense). This perspective however was not
presented as a theory of crime (Tracy and Kemph-Leonard, 1996). It was merely depicted
as a formula for arranging known facts, which if applied to different theories of crime,
might enhance their understanding and predictability. The criminal career approach itself
however does not explain the phenomenon of crime (Tracy and Kemph-Leonard, 1996).
Conversely, theories may also beneﬁt from the criminal career paradigm in the

policymaking stage by employing selective incapacitation against those juvenile

offenders, who are thus predicted to become career criminals (Blumstein et. al., 1986 and
Blumstein, Cohen and Farrington 1988). Theories that utilize the criminal career
paradigm are classiﬁed as typological theories of crime and chiefly originate from “the
“Positive” school of criminology (Tracy and Kemph-Leonard, 1996 and Piquero, 2000).
Independent of the criminal career approach another theoretical perception was also
evolving, which evaluated criminal patterns and their likelihood of remaining invariable
or ﬂuctuating across distinct crime categories (Healy and Bronner, 1926; Schrag, 1944;
Cloward and Ohlin, 1960; Gibbons, 1965, and Wolfgang, Figlio, and Sellin, 1972). Later,
while developing the criminal career paradigm, Blumstein et. al. (1986) asserted that the
criminal careers may have a particular tendency or direction in seriousness or that the
offenders may specialize in particular types of Offenses. Specialization and escalation
thus were deemed to be consequents of the criminal career paradigm, which as stated by
Blumstein et. al. (1988) may or may not hold veracity depending on future empirical
evidence. So far the empirical evidence regarding the existence of specialization and
escalation has been mixed for the former and none for the later (Bursik, 1980; Klein,
1984; Britt, 1996 and Piquero, Paternoster, Mazerolle, Brame and Dean, 1999; Piquero,
2000; Piquero and Buka, 2002). The empirical validity of such studies is contingent upon
their focus on three factors, a) relationship of specialization with age, b) the development
of a typology of offenses and c) the method of measurement (Tracy and Kemph-Leonard,
1996). The methods part shall be dealt with in the sections following. Regarding age and
specialization it has been noted that age appears to bring about a decline in crime
versatility exhibited by offenders (Peterson, Braiker, and Polich, 1980; Gottfredson and

Hirschi, 1990 and Piquero et. al., 1999). According to criminologists this fact could be

best explained by unsuccessful and successful past experiences (Blumstein et. al., 1986),
ratio between controls exercised to control experienced (Tittle, 1995), gaining of personal
capital with age (Nagin and Paternoster, 1994) or the assertion that behavior tends to get
organized and patterned with experience (Loeber and LeBlanc, 1989). The development
of a typology of crime has been a major issue in the specialization literature (Tracy and
Kemph-Leonard, 1996). Mostly, criminologists use the typology of violent versus non-
violent (property) offenses in order to determine specialization (Piquero et. al. 1999;
Piquero, 2000 and Piquero and Buka, 2002). Some criminologists have also employed
different typology matrices to facilitate the testing of their particular research hypotheses
(Wolfgang et. al. 1972; Bursik, 1980 and Britt, 1996). In opposition of the above
mentioned views, Cornish and Clark (1987) in the rational choice theory proposed a
“crime speciﬁc model”, arguing that a classiﬁcation that group offenses cannot
satisfactorily identify offender involvement.

The existence of specialization among sex offenders is a research question that has
been theoretically and empirically dealt with in the clinical as well as the criminological
arena. However, clinicians have clearly dominated the ﬁeld (Simon, 2000). According to
the American Psychiatric Association (1994) sexual offenders belong to the group of
paraphilias that comprises of various kinds of sexual deviances including pedophilia and
rape (Palermo and Franks, 2001). According to clinicians a deviant paraphilas’ act of
forcibly possessing the victim is bound to leave him dissatisﬁed. Like drug addicts, the
act of sexual deviance renders only a momentary pleasure to him and will soon require
another ﬁx; therefore he is bound to recidiVate (Palermo and Franks, 2001). This

observation may be true with regards to the clinical ﬁeld where the population under

Study generally comprises of people already diagnosed with sex related disorders,
however in the context of criminology and criminal justice it is fallacious as the word
population under this discipline encompasses of the entire body of sexual offenders and
not just a few problem cases (Simon, 2000).

One theory that propagates the specialization of sex offenders is the Psychosis
Theory (Palermo and Franks, 2001). Sexually abusive behaviors are associated with
borderline Schizophrenic or psychotic conditions (Ryan, 1997). According to this theory
the sex offender is a self centered, callous and remorseless person lacking empathy and
the ability to form close relationships, a person who functions without the restraint of
conscience (Cleckley, 1941). As the centers of aggression and sexual arousal are in close
proximity in the brain, the coercive sexual act often leads to aggression and use of force
(Palermo and Franks, 2001). Such psychotic tendencies are deemed inherent in the
paraphilias, albeit social conditions may catalyze the process of sexual delinquency
(Palermo and Franks, 2001). Despite its popularity psychotic sex offenders only account
for 8% of the entire population of sex offenders (Knopp, 1984). Following the
assumptions of the psychotic theory i.e. that sex offenders are born rather than raised,
certain physiological theories also searched for neurological and hormonal factors that
might be the cause for sexual deviance (Ryan, 1997 and Palermo and Franks, 2001).
Brain functions especially neurotransmitters and their relationship to sexual aggression,
that relate to the emotional states have been the focus of much research (Ryan, 1997). To
date numerous psychiatric disorders like attention deﬁcit disorder, obsessive-compulsive
disorder, mood disorders, dissociative disorders have been identiﬁed as potential causes

for aggressive sexual behavior, yet no particular neurological explanation has been found

(Ryan, 1997). During the19803 research was conducted regarding the relationship
between post stress disorder and sexual aggression, stating that the experience of an
overwhelming trauma may result in causing permanent alterations in the way the brain
secretes certain chemicals in response to a subsequent event (Van der Kolk, 1986).
Although this line of research shows potential, it still lacks the backing of empirical
evidence (Ryan, 1997). In 1989, Ellis applied the arousal theory to explain the behavior
of rapists and claimed that it was genetically induced. According to him, rapists suffer
from a low arousal level, which necessitates strong emotional stimuli like envisioning or
actually committing a rape in order for them to ejaculate (Ellis, 1989, Palermo, 2001).
The most recent attention paid to hormonal levels and aggressive behavior relates to
testosterone (Vold et. al. 2002). Booth and Osgood (1993) established that a high level of
testosterone could lead to reduction in social integration and bring about higher levels of
adult and juvenile deviance. Anti testosterone drugs have been administered in treatment
of problem sex offenders in order to decrease their sexual arousal and in turn remove the
cause for deviance (Berlin and Meinecke, 1981) however the empirical value of such
treatments is still undecided (Ryan, 1997).

Albeit related to the theory of psychosis yet distinct within their domain are the
theories of intra-psychic conﬂict (Ryan, 1997). These theories basically emanate from
Freud’s theory of personality (Freud, 1954), which stated that people have two basic
instincts: sexual and aggressive; and the internal craving for gratiﬁcation of impulses is in
conﬂict with the external demand of being socially acceptable and it is the nature of this
conﬂict deﬁnes a personality. Developing upon his own theory Freud (1954) propagated

the ﬁrst intra-psychic theory i.e. the seduction theory. This theory stated that sexual

traumas in the childhood lay at the bottom of neurotic illnesses (Freud, 1954). In 1979,
Groth integrated the personality theory and the seduction theory by dichotomizing child
molesters into ﬁxated (whose sexual development ceazed after the childhood trauma thus
had sexual interest only in children) and Regressed (whose sexual interests had developed
into the adult stage but the trauma regressed them to an earlier stage which kept them
sexually involved with both children and adults). The idea that childhood traumas and
adult conﬂicts are causal of sexual deviance is not without substance; however there may
be other causal factors without the help of which the phenomenon of sexual deviance
cannot be explained in its entirety (Ryan, 1997). Regarding childhood traumas and the
developmental theories of Freud (1965), Piaget (1928) and Erickson (1963) also state that
a child develops his personality in the initial years of his life and a trauma sustained at
such period may trigger susceptibility towards deviance or sexual deviance. Recently,
Ward, Hudson, Marshall and Seigert (1995) studied the relation between sexually abusive
behavior and intimacy deﬁcit and hypothesized that such behavior is associated with an
abnormality in the early childhood relationships which results in the distortion of self
image and expectations in adult relationships throughout the life span. The
developmental—contextual theory on the other hand provides a more practicable approach
by focusing not on the individual’s developmental status, rather on the interaction of such
developmental status with the life experiences that are subject to constant change (Ryan,
1997).

The learning theory renders a psychosocial perspective on sex offenses, by stating
that although the capacity to sexuality is inborn the manifestation of such capacity is

learned (Abel, Becker, Murphy and Flanagan, 1975; Freeman-Longo, 1982 and Ryan,

1997). According to this theory the learning process begins prior to experience, thus
repeated exposure to a negative stimuli (in this case sexual intercession and deviance)
tends to propagate acceptance and attraction towards it (Bandura, 1977). However, this
theory does not assume specialization of sex offenders and learning is considered
instrumental only in the pre-experience stage and after the experience the behavior has to
be reinforced repetitively in accordance with the actor’s personal inclinations in order to
establish specialization (Pavlov, 1927; Ryan, 1997). Further developing on the repetitive
aspect of sexual offending the Cognitive theory states in order to reinforce sexually
abusive behavior the offender allows himself to imagine that his or her behavior is
socially acceptable or harmless (Yochelson and Samenow, 1976). Alternatively, the
Addictive theory concludes that, “the sexual behaviors become unmanageable or out of
control because of the offender’s preoccupation, ritualism, compulsivity and subsequent
despair. The offender sees his behavior as beyond his control” (Cames, 1983 and Ryan,
1997).

Mildly related to the Learning theory is the Feminist perspective on sexual offenses.
According to this theory, “rape is an act of male aggression resulting from an atavistic
male dominance and female submission in a political and economic system that
perpetrates social inequalities between the two sexes (Brownmiller, 1975). In the same
spirit as the Feminist perspective the Evolutionary theory perceives the natural masculine
derive to copulate with different partners and the concomitant desire to impregnate large
numbers of women as a cause for rapes (Oliver, 1989). Both the Evolutionary and
Feminist theories, like the Psychotic, Physiological and Psychological perspective,

assume specialization among sex offenders.

10

Lastly, the sexual abuse cycle (Lane, 1991) attempts to explain the working of a
juvenile sex offender’s brain by identifying different states of mind involved in the sexual
abuse process. The cycle begins with a trigger (event); which distorts the juvenile’s
perception of future (negative anticipation); to counter his hopeless situation he resorts to
evading the issue (avoidance); as his attempts at avoidance fail he becomes defensive and
begins to assert power on others in non-sexual ways (power/control); this stage leads him
to an indulgence in power related sexual fantasies (fantasy); which results in the act of
sexual abuse; the gravity of his behavior and the fear of the consequences steers him into
fugitive thinking; and ﬁnally he manages to assimilate his behavior through a series of
thinking errors (reframing), putting him back in the same state as he was before the
triggering episode (Lane, 1997).

The Generalist Perspective on Crime

The debate between the Criminal Career paradigm and the Generalist perspective
can be traced back to the historical contention between the Classical and Positive schools.
The Criminal Career approach although not a theory itself works on the assumption that
the criminal element in the society has distinct characteristics either emanating from
external or internal stimulus or both, that will induce them to recidivate and that it is the
criminal or the criminality within the criminal that has to be addressed in order to counter
crime (Blumstein et. al. 1986, Tracy and Kempf-Leonard, 1996). As this assumption is
part and parcel to this paradigm mostly the social control theories and some biological
theories that adhere to the same assumption utilize it (Cullen and Agnew, 1999). The
social control theories are sequentially offshoots of the Positive School of criminology.

On the other hand, the Classical School of Criminology focuses on the instance of crime

11

instead of the criminal (Vold et. al. 2002). According to the Classical School, criminals
possess no speciﬁc criminal propensities; they merely follow the universal tendency to
enhance their pleasure (Bentham, 1998, Cullen and Agnew, 1999). Hence crime is to be
prevented through consequences painful to the individual ( Bentham, 1998). Such are
also the assumptions of the control theories and the generalist perspectives that originate
from the Classical School (Gottfredson and Hirschi, 1990 and Cullen and Agnew, 1999).
For the purposes of this study only two main generalist theories are relevant and need to
be elaborated upon i.e. Mofﬁtt’s “Taxonomy” and Gottfredson and Hirschi’s “General
Theory of Crime” (Mofﬁtt, 1993 and Gottfredson and Hirschi, 1990).

Mofﬁtt’s taxonomic framework divides criminal offending into two separate
categories (Piquero et. al. 1999). The ﬁrst group labeled as “Life Course Persisters”,
relatively smaller between the two, is comprised of individuals that commence offending
at a very early stage in life and continue to engage in such behavior throughout their life
span. Such deviant behavior generates from neuropsychological deﬁcits that form before
or shortly after birth, however adverse social condition may catalyze the onset of criminal
activity. The crimes committed within this group are diverse and of a serious nature
(Mofﬁtt, 1993, Piquero et. al. 1999). The second and comparatively populous group
classiﬁed as the “Adolescence Limited Group”, includes socially well-adjusted
individuals that indulge in crime only during the adolescence phase of their life spans.
Crimes committed within this group are less diverse and less severe in relation to the ﬁrst
group (Mofﬁtt, 1993). Mofﬁtt (1993) identiﬁes inadequate socialization as a predictor of
deviance within both the groups: inadequate socialization is inversely related to onset age

and severity of offense; inadequate socialization is positively related to offense variety;

12

thus onset age and severity is inversely related to the variety of offenses committed
(Piquero et. al. 1999).

Gottfredson and Hirschi’s “General Theory of Crime” (1990) signiﬁes the opposing
pole in comparison to the criminal career criminal paradigm and shall be deemed as the
representative of the generalist perspective for the purposes of this research. According to
this theory one’s self-control prevails as the focal impediment against indulgence in
criminal activity and that low self-control is ﬁgurative of the criminal propensity in all
criminals (Gottfredson and Hirschi, 1990). In the words of Piquero et. al. (1999), self-
control is positively related to the onset age; self.control is inversely related to the variety
of offenses committed; and onset age is inversely related to the severity of offenses.
Hence, a person with very low self—control is likely to onset offending activity at an
earlier age and should exhibit the most versatility in offending patterns (Piquero et. al.
1999). Gottfredson and Hirschi (1990) attribute ineffective child rearing as a major cause
of low self-control. Hence, the direct control exercised by parents early on in childhood,
in terms of discipline, affection and supervision determines the criminal propensity
existent in the individual later on in life (Patterson, 1980, Gottfredson and Hirschi, 1990).
Gottfredson and Hirschi (1990) also introduce a situational aspect to their theory by
stating that low self-control or impulsivity shall only be roused in the presence of a
lucrative criminal opportunity. Once a person has experienced ineffective child rearing
the resulting criminal propensity is likely to exist in him throughout the life Span, albeit
reducing the criminal opportunities present within the society can still prevent crimes
(Gottfredson and Hirschi, 1989). This is essentially a Classical attribute, which proposes

prevention of crime through the principles of severity, celerity and certainty (Vold et. al.,

13

2002). The General Theory of crime has been chieﬂy criticized on two accounts: a) that
low self-control cannot explain the effects of other social forces (e. g. differential
association and social learning) on crime (Evans, Francis, Velmer, Dunaway and Benson,
1997) and b) that family life cannot be meaningfully separated from the social forces that
effect it from the outside (Cunie, 1985).

LITERATURE REVIEW

This section shall be more problem speciﬁc in dealing with only empirical
literature that pertains speciﬁcally to the issues of sex offender specialization and
specialization in general. The review of such literature shall bear two purposes i.e.
analyzing different approaches that are being applied in analyzing sex offender
specialization in the light of different disciplines (clinical psychology and criminal
justice) and also to evaluate different statistical methods used in these studies and other
empirical studies that deal with specialization in general or offense speciﬁc specialization
in order to ﬁnd the best suited statistical methodology for the present study.

In this section Sex Offender specialization studies shall be divided mainly into
three distinct categories. First, the studies that explain and assess the empirical value of
different tests developed under clinical psychology and psychiatry, which are proposed to
predict sex offender specialization. However, it should be pointed out here that all such
studies only test the empirical value of the proposed tests and not the specialization
hypothesis; therefore they assume the existence of sex offender specialization. This
assumption in itself is wrong to the extent that sex offender specialization has not yet
been established as an empirical fact, hence rendering the scope of such studies weak.

Secondly, the studies that endeavor to measure the recidivism rates of sex offenders.

l4

These studies albeit will be useful to look at but shall fail to entirely address the issue at
hand, as specialization and recidivism are two very different constructs and cannot be
used in substitution. Lastly, the studies that measure specialization among sex offenders
shall be reviewed. A review of these studies will not only elucidate the empirical standing
of the sex offender specialization hypothesis in the contemporary criminological
research, but will also help in identifying the best suited statistical methodology for the
proposed study. For this purpose other studies that deal with specialization in general or
from an offense speciﬁc perspective shall also be looked at.
Empirical Value of Psychological Tests that propose to predict Specialization among Sex
Offenders

The clinical psychology ﬁeld chieﬂy deals with the phenomenon of sex offender
specialization form a risk assessment perspective (Prentky and Burgess, 2000). However
even in this ﬁeld the body of work available on the subject is fairly recent and hardly
extensive ( Prentky and Burgess, 2000). On the other hand, from a social science
perspective the entire treatment of the sex offender specialization issue in the clinical
psychology ﬁeld has been essentially unscientiﬁc (Kuhn, 1996 and Singleton and Straits,
1999). In social sciences after a phenomenon or problem has been identiﬁed through a
theory, it is initially subjected to the exploratory research process in order to establish its
existence and once the existence of such a phenomenon or problem is empirically
ascertained only then can the research process transcend into the descriptive, explanatory
and treatment stage (Singleton and Straits, 1999). In the above-mentioned case however,
clinicians started off by identifying the sex offender specialization problem in different

theories (which have been explained in chapter 1) and instead of going through the

15

exploratory process, hopped straight onto the descriptive, explanatory and treatment
stages, under the unfounded assumption that sex offender specialization existed. After
scrutinizing the clinical psychology literature on sex offender specialization a very small
number of studies were observed to address the specialization question, out of these one
will be analyzed here (as it is written purely from a clinical psychology perspective) and
the rest shall be reviewed with other recidivism studies (Hanson, Heather and Steffy,
1995). Hanson et. al. (1995) partially addressed the specialization question by conducting
a 15 to 30 year follow up study on 191 child molester and 188 non-sex offenders and
comparing the recidivism rates of the two groups by using the proportional survival
hazard analysis. The results indicated an 83.2% recidivism rate for non-sex offenders and
a 61.8% recidivism rate for child molesters. Although the study showed a high level of
recidivism among child molesters the empirical value of this study is weak. First, the
entire sample was extracted from a pool of child molesters in maximum-security prison,
and to generalize the results as representative of the entire sex offender population is
empirically wrong. Second, the study used specialization and recidivism in substitution,
which under criminal justice are two separate constructs and cannot be used alternatively.
Lastly, recidivism was only considered if an offender re-offended in the same type of
offense. Thus, all subjects that did not re-offend or recidivated into a different kind of
offense (e. g. a child molester recidivating into a non-sexual offense) were censored,
which in turn falsely established the two categories as mutually exclusive and created a
design in favor of specialization (Hanson et. al., 1995).

Relying on the unsubstantiated assumption of specialization, clinical

psychologists have identiﬁed a number of personality traits that are deemed to be

16

common in the sexually deviant mentality and have devised specialized tests that are
proposed to measure such characteristics and possibly predict future deviance (Palermo
and Farkas, 2001). Before evaluative literature on speciﬁc tests is reviewed it is necessary
to establish a generalized perspective as to the different factors and approaches in the sex
offenders risk assessment process.

According to clinicians there are two major types of risk assessment factors, static
(ﬁxed) and dynamic (changeable) (Hanson, 1998). Static risk factors, like childhood
maladjustment or prior offenses, can indicate long-term trajectories of propensities to
engaging in deviant behavior. Static factors, however, cannot predict recidivism, nor can
they determine whether offenders have substantially reduced their likelihood of re-
offending e. g. response to treatment etc. (Hanson, 1998). Dynamic Risk factors, on the
other hand, can not only predict recidivism but can also demonstrate the increase or
decrease in the likelihood of such recidivism, after they have undergone some change
(Hanson, 1998). Dynamic risk factors can be further sub divided into stable and acute.
Stable factors have the potential of changing but typically endure for long periods of time
(e.g., deviant sexual preferences or alcoholism) and are helpful in predicting the
likelihood of recidivism (Hanson, 1998). Acute risk factors, in contrast, are rapidly
changing states (e.g., sexual arousal or drunkenness) that immediately precede sexual
offenses, and help in determining the time span in which such recidivism is likely to
occur (Hanson, 1998). Thus, a complete risk assessment can only be made after analyzing
both static and dynamic risk factors.

Clinicians have devised chieﬂy three approaches to conducting risk assessments:

guided clinical, pure actuarial and adjusted actuarial (Hanson, 1998). The guided clinical

17

approach does not explicitly determine the method of interpreting the recognized risk
factors, relaying predominantly on the expert evaluator’s judgment (Boer, Wilson,
Gauthier and Hart, 1997). Conversely, the actuarial approach evaluates the offender on a
set of identiﬁed predictors and then combines these variables using a predetermined,
numerical weighting system (Hanson, 1998). The adjusted actuarial approach begins with
an actuarial prediction, but expert evaluators can then adjust the actuarial prediction after
considering potentially important factors that were initially not included in the actuarial
measure (Quinsey, Lalumiere, Rice and Harris, 1995 and Hanson, 1998). Studies have
shown actuarial measures to be the most accurate as compared to the other two
approaches (Hanson, 1998). Now that the basic criteria of the risk assessment process
have been established, literature that empirically tests and evaluated the predictive ability
of some archetypal risk assessment tests shall be reviewed.

In 1991, Hare, introduced the Psychopathy Checklist-Revised (PCL-R), which
identiﬁed deviant sexual preferences, preference for children on plenthysmography,
antisocial personality disorder and prior sexual offenses as the strongest predictors of
future sexual deviance. The test showed reliability of prediction of violence and
recidivism when tested on maximum-security prison inmates and forensic psychiatric
patients (Palermo and Farkas, 2001), albeit generalizing the results of such a study would
be wrong. Also, generalized conduction of this test is not feasible, as collateral and ﬁle
information of the subject is required in order to make a rating, which makes the entire
process cumbersome and time consuming (Palermo and Farkas, 2001). In 1997, Hanson
developed an actuarial scale using four main factors (prior sexual offense, age less that

25, extra-familial victims, and male victims) for assessing risk of recidivism among sex

18

offenders. A study found that Hanson’s four-item Rapid Risk Assessment for Sex
Offense Recidivism scale correlated only .27 with sexual recidivism (Prentky and
Burgess, 2000) showing a very moderate rate of accuracy (Palermo and Farkas, 2001).
Perhaps the most widely used and reported actuarial risk assessment scale in the
empirical literature is the Violence Risk Appraisal Guide (Harris, Rice and Quinsey,
1993; Webster, Harris, Rice, Cormier and Quinsey, 1994; Rice and Harris, 1997;
Quinsey, Harris, Rice and Cormier, 1998 and Harris, Rice, Chaplin and Quinsey, 1999).
The VRAG was developed initially to predict violent recidivism but later was also used
to evaluate the risk of sexual recidivism among known sex offenders (Rice and Harris,
1995). The VRAG consists of analysis of 12 variables including Psychopathy Checklist,
alcohol abuse, elementary school maladjustment, criminal history, female victims and
diagnosis of schizophrenia (Prentky and Burgess, 2000). Empirical studies that have
tested the predictive ability of VRAG have found relatively higher correlation (.46 and
.47) between VRAG and violent recidivism (Harris et. al 1993 and Rice and Harris,
1997). In univariate correlations, Psychopathy Checklist (.34) and elementary school
maladjustment (.31) were illustrated as the best individual predictors out of the twelve
(Harris et. al. 1993). The studies also revealed that although the VRAG is relatively
reliable in predicting generic violence, its ability to predict sexual recidivism is weak
(Rice and Harris, 1997). The correlation coefﬁcient of the relationship between sexual
recidivism and the VRAG stood only at .20 (Rice and Harris). Also, in all the studies the
sample was selected from a maximum-security psychiatric facility, which is by no means
representative of the general violent or sex offender population (Harris et. al. 1993 and

Rice and Harris, 1997). Quinsey et. al. (1998) devised the Sex Offender Risk Appraisal

19

Guide in 1998, by taking ten items from VRAG and adding four of their own, namely,
criminal history for violent offenses, previous convictions for sex offenses, sex offense
history against girls under 14 and Phallometric test results. The ability of SORAG in
projecting sex offender recidivism improves on VRAG but only slightly (Quinsey et. al.
1998). Quinsey et. al. attribute this disability to predict sex offender recidivism to two
factors, 1) higher measurement error associated with sex offenders (as many sex offenses
fare low on the violent offender continuum) and 2) narrow range of risk in the studies on
sex offenders (very few low risk sex offenders are included in the sample). Serin,
Barbaree, Seto, Malcolm and Peacock (1997) devised a comprehensive two-tier risk
assessment model. Under this model (Serin et. al. 1997) tier one consists of central risk
factors i.e. criminality (8 variables) and sexual deviance (12 variables) and tier two
includes moderating risk factors i.e. social competence (7 variables), substance abuse (10
variables) and treatment readiness (4 variables). This model was tested on 466 sex
offenders at the Warkworth Prison and displayed a very modest correlation (.10 to .36)
with the prediction of future sex offenses (Serin et. al., 1997 and Prentky and Burgess,
2000). The California Actuarial Risk Assessment Tables deliver two sets of predictors
one for rapists and one for child molesters (Prentky and Burgess, 2000). The CRAT
however, has not yet passed through any empirical testing and its predictive ability is
hitherto not known. Prentky, Harris, Frizzell and Righthand, (2000) devised the Juvenile
Sex Offender Assessment Protocol for risk assessment of juvenile sex offenders. This test
consists of 4 rationally derived factors: two historical (sexual derive and impulsive
antisocial behavior) and two dynamic factors (intervention and community stability).

This test however is relatively new, and having been empirically tested only once it has

20

shown better reliability and validity than its predecessors; however it may be too early to
make a deﬁnitive statement as to its reliability and validity (Prentky et al. 2000). Like J -
SOAP Prentky also developed Adult Sex Offender Assessment Protocol in order to assess
risk among adult sex offenders (Prentky and Burgess, 2000). The A-SOAP is also in the
process of being tested and there is no knowledge as to the validity and reliability of this
test (Prentky and Burgess, 2000).

Some clinicians have also contended that sexual deviance can be foretold with the
help of personality tests (Palermo and Farkas, 2001). The Minnasota Multiphasic
Personality Inventory (MMPI-I and MMPI-II), commonly used in the general practice of
psychology, has also been used to predict sexual recidivism among adult and juvenile sex
offenders (Marshall and Hall, 1995). The MMPI, apart from the regular factors that
measure dynamics of a personality, also has factors like pedophilic hostility, overt
hostility, over controlled hostility, Welsh anxiety, repression and ego strength, which are
purported to measure sexual deviance (Hanson, Steffy and Gauthier, 1993). Other
personality tests that are also assumed to measure sexual deviance are Eysenck
Personality Inventory, Lykken Anxiety Scales, Fenz Anxiety Scales and Intemal-Extemal
Locus of Control Scale (Hanson et. al. 1993). Hanson et. a1. (1993) conducted a long term
recidivism study on male child molesters with a 10 to 31 year follow up period. This
study tested the extrapolative ability of all the aforementioned tests including the MMPI,
and found that none of these tests to be signiﬁcantly related with predicting sexual
recidivism (Hanson et. al. 1993). Unlike the MJVIPI, the Multiphasic Sex Inventory (MSI)
has been speciﬁcally designed not only to detect sexual offender but also sex offender

subtypes (Schlank, 1995). Schlank (1995) tested 164 incarcerated adult sex offenders and

21

found that three key factors (social sexual desirability, sexual obsession and paraphilia)
were pivotal in predicting sexual recidivism. He also found that the MSI had better
reliability and validity as compared to the MMPI. Schlank conducted his study in the
form of a one shot case study, in which he selected the sample out of a prison population,
divided them in seven different groups depending on the nature of the offense and
administered the MSI on each of them (Schlank, 1995). Such studies however, are
considered very inaccurate as far as external and internal validity is concerned (Singleton
and Straits 1999). Not only are the results from such study affected by history (events in
the environment other than the independent variable that might change the outcome),
maturation (any change that takes place in the subjects during treatment) and attrition
(loss of subjects during treatment), they can hardly be generalized, on account of their
low external validity (Singleton and Straits 1999). For example in Shalak’s study the
sample was chosen from a prison population, which is not representative of the entire
population of sex offenders. After selection they were all categorized into different
subgroups with respect to their offense types, but in actuality, the fact that a person was
convicted of a particular kind of sex offense does not mean that he specialized in that
offense prior to conviction, to the extent that he should be classiﬁed into that particular
offense’s subgroup (Singleton and Straits 1999). This in fact creates a bias in favor of
specialization by introducing it as one of the assumptions made in the study (Singleton
and Straits 1999).

Another category of sex offender risk assessment tests involves the measurement
of the subject’s response to audio and visual stimuli. Card and Dibble (1995) Studied the

predictive value of the Card/Famall Penile Plethysmography Audio-Visual Stimulus

22

Materials and found that even without the use of pornographic, illegal and offensive
stimuli, the response patterns of pedophiles were reliably distinguished (Card and Dibble,
1995 and Palermo and Farkas, 2001). Albeit, the Card and Dibble (1995) study found the
C/FPA-VSM to be reliable, it suffered from the same shortcoming as almost all the
studies conducted by clinicians i.e. generalizing results obtained by a pool of chronic
offenders to all offenders. The subjects in this study (pedophiles) were recruited from a
maximum-security psychiatric facility and were not at all representative of all pedophiles
let alone the entire sex offender population. The Card/Farrall Auditory Stimuli (CAS) is
projective rather than descriptive (Palermo and Farkas, 2001). It works on the assumption
that an exact description of the act is not required to the subjects and that deviant
sexuality generates a triggered response to a mere self-reminder of the deviant act
(Palermo and Farkas, 2001). This test however has not yet been subjected to empirical
testing as far as its predictive value is concerned.

Perhaps the most controversial in this line of tests is the Phallometric measure
(Palermo and Frakas, 2001). This test has been criticized for the use of pornographic
slides with out the informed consent of the subject and there is also an argument that the
entire process is degrading and anti-therapeutic for the sex offender (Marshall and Hall,
1995 and Cumming and Buell, 1998). From an empirical perspective this test has gone
through moderate testing as to its predictive value and the results are mixed. In a meta-
analysis Lalumiere and Quisley (1994) examined 16 studies that used the Phallometric
measure on rapists, to verify if rapists’ response was any different from other offenders to
audio and visual stimuli. The study found that rapists responded differently to such

stimuli as compared to other offenders and that such response got more pronounced

23

against more graphic and descriptive material (Lalumiere and Quisley 1994). While the
result of this study may be considered very comprehensive in the clinical psychology
ﬁeld, from a criminal justice viewpoint they are hardly reliable. First, almost all the
studies recruited their subject from maximum-security psychiatric facilities; hence
generalizing these ﬁndings would be empirically incorrect. And second, the fact that
rapists may have different response to the same stimuli as compared to other offenders,
by no means proves that such a propensity will induce them to recidivate into another sex
offense. Thus under criminal justice, only a recidivism or follow up study can establish
the empirical value of this test. Marnie, Quinsey and Harris (1991) conducted a follow up
study on 136 extra familial child molesters in a maximum-security psychiatric institution,
on whom the Phallometric test had been administered, for an average period of 6.3 years
after their release. The study found that only thirty one percent of the sample was
convicted of another sex offense in the follow up period (Marnie, 1991). The point to be
considered here, is that the sample understudy was extracted from a maximum-security
psychiatric institution, therefore its would be logical to assume that the thirty one percent
rate of recidivism in a 6.3 year period, represents the worst case scenario of sex offender
recidivism. The actual rate of recidivism for the entire population of sex offenders is
bound to be much less.

An analysis of the above literature has principally expounded three facts. First,
that all the risk assessment tests are essentially based on the unfounded assumption that
specialization exists among sex offenders. Secondly, that the results of almost all the
studies conducted to establish the empirical value of the predictive ability of the risk

assessment instruments, can not be generalized. Lastly, independent of the ﬁrst two facts,

24

no study has actually presented any over whelming evidence that any of the risk
assessment tests predict sex offender recidivism with reasonable accuracy.
Recidivism Studies on Sex Offenders

Another branch of research that partially deals with the sex offender
specialization issue are the sex offender recidivism studies. As stated in the ﬁrst chapter,
“Specialization refers to the tendency among criminal offenders to repeat the same type
of crime across their criminal careers” (Britt, 1996); recidivism studies on the other hand,
concentrate on the offense immediately subsequent to the beginning point in the study,
instead of the whole criminal career. There is a reasonable body of empirical literature
available on sex offender recidivism; however the results presented by these studies are
contrasting and do not lead to any deﬁnitive conclusions. Criminologists attribute this
variance in the results of different studies to four key factors, 1) the deﬁnition of
recidivism, 2) inclusion criterion for the subjects, 3) the source of the data used and 3) the
length of the follow up period (Doren, 1998). A stringent operationalization of the word
recidivism in sex offenses (e. g. reconviction) as compared to a more ﬂexible one (e. g.
arrest) can actually cause an artiﬁcial shrinkage in the resulting recidivism rates against
specialization that might not exist in reality, and vice versa (Doren, 1998). Similarly,
enforcing a broad based inclusion criterion for sex offenders (e. g. arrest) as compared to a
more moderate criterion (e. g. previous conviction) can generate a sample excessively
representative of the actual population, creating an over projection of the recidivism rates
in favor of specialization, and vice versa. Also, according to criminologists ofﬁcial data
tend to generate samples under representative of the population of sex offenders in

comparison to self-reported data (Doren, 1998). Russell (1982) stated that only about ten

25

percent of the rapes against women were ofﬁcially reported. In opposition, ofﬁcial data
have greater reliability as compared to self-reported data, which renders it as the
preferred source in conducting empirical research (Doren, 1998). Lastly some
criminologists also state that the true recidivism rates of the sex offenders are revealed in
studies that involve very long term follow up periods and that studies involving short
term follow up periods have a tendency of under estimating such recidivism rates
(Prentky, Lee, Knight and Cerce, 1998 and Doren, 1998). Some studies have shown “ﬁrst
time” sex offender recidivism 20 to 28 years after initial incarceration (Hanson, Steffy
and Gauthier, 1993 and Prentky et. al. 1998). To date, only two such recidivism studies
have been conducted that had a follow up period of 25 years or more, and have failed to
produce any irrefutable evidence in favor of sex offender recidivism (Hanson et. al. 1993
and Prentky et. al. 1998).

Sex Offender recidivism studies rarely deal with the phenomenon of sex offenders
in its entirety. Typically, sex offenders are bifurcated into child molesters and rapists and
then each category is independently dealt with (Doren, 1998). However, such a
distinction is unwarranted on two accounts; a) ﬁrst it has to be proved whether sex
offenders specialize in general before determining what sub category of sex offenders is
more specialization prone and b) the policies and laws that deal speciﬁcally with sex
offenders hardly ever make such distinction. Even so, this study shall separately review
the recidivism literature on child molesters and rapists in the parts following.

Child molester recidivism studies that have had a follow up period of 5 years or
less have essentially shown very low recidivism rates among this class (Doren, 1998).

Radzinowicz (1957) conducted the pioneering research in this regard by studying the

26

recidivism rates of convicted child molesters over a ﬁxed period of 4 years and found
sexual offense recidivism to be 11.3 percent. As the criterion of recidivism was
reconviction, the resulting ﬁgure of 11.3 percent might have been less than the actual sex
offense recidivism rate, due to the stringent operationalization of recidivism. Hanson and
Bussiere (1996), performed a meta analysis of 61 different sex offender data bases and
found only a 12.7 percent sex offense recidivism rate for 9,603 child molesters, for a 4 to
5 year follow up period. Abel, Mittelman, Becker, Rathner and Rouleau (1988), collected
self report data on a non-incarcerated sample of child molesters and found a 12.2 percent
sex offense recidivism rate in a one year follow up period. They attribute this high
recidivism rate to the fact that self-report data are more representative of the real
recidivism rates (Abel et. al. 1988). However as stated above, self-reported data fares low
on the authenticity scale and in the case of this study the use of a non-incarcerated sample
may also have falsely inﬂated the recidivism rates. Broarhurst and Maller (1992)
conducted an ll-year follow up study, over a sample dominant by child molesters and
observed a 34 percent recidivism rate. This ﬁgure however is not statistically
representative of the actual rate of recidivism of child molester, as deﬁnition of
recidivism taken in this study includes not only sex offenses but also other non-sexual
violent offenses (Broarhurst and Maller, 1992). Now, focusing on studies involving long-
term follow up periods, Hanson, Scott and Steffy (1995) investigated the sex offense
recidivism rates of a sample of 191 child molesters over a period of 31 years and found it
to be 35.1 percent. Doren (1998) criticized these results as an “inaccurate estimate” of the
true recidivism rate, on account of the narrow interpretation levied on recidivism in the

study i.e. reconviction. Prentky et. al. (1998) using a recidivism deﬁnition of “new sex

27

offense charge”, examined the recidivism rates of 115 extra-familial child molesters.
Subjects had been previously incarcerated and the follow up period was 25 years after
such incarceration (Prentky et. al. 1998). This study revealed a 52 percent sex offense
recidivism rate among the sample of child molesters (Prentky et. al. 1998). The results of
this research have been quoted in many studies and are probably considered the closest to
the actual sex offender recidivism rates (Doren, 1998). However this study bears a major
ﬂaw from a specialization perspective. The study only includes the new sex offense
charges with in the 25 years follow up period, which means that any other non sexual
charges brought on the subjects before or after the incidence of the “new sex offense
charge” would not be recorded. Hence, Prentky et. al. (1998) studied sex offense
recidivism among child molesters after removing every possibility of proving
generalization from the data set, rendering the empirical value of this study unreliable.
The patterns of recidivism found in child molester studies are pretty much similar
to those found in the rape studies. Studies that analyze sex offense recidivism among
rapists for a follow up period of ﬁve years or less reveal a variety of results. Radzinowicz
(1957) found the rate of recidivism among rapists to be 10 percent in a follow up period
of four years. However this study may have under represented the real rate of sex offense
recidivism among rapists, by using reconviction as criterion of recidivism. Frisbie and
Donidas (1965) found a 35.6 percent recidivism rate among, what they called “sexual
aggressors”. However, these results are not reliable on account of the deﬁnition taken in
this study of the term sexual aggressor, which does not include child molesters. Sturgeon
and Taylor (1980) took two samples of rapists for a ﬁve-year follow up study; the ﬁrst

sample consisted of rapists that had been diagnosed as mentally disordered showed a 19

28

percent recidivism rate and the second sample, consistent of rapists not diagnosed as
mentally disordered revealed a recidivism rate of 28 percent over the same follow up
period. This study puts into contention the entire body of work reviewed in the ﬁrst
section of this chapter relating to the risk assessment instruments. Albeit this research
does not entirely nullify the empirical value of all the psychological tests devised by
clinicians, it does however prove the point made earlier in this chapter that clinicians
have devised all these tests that measure specialization among sex offenders without ever
proving the existence of such specialization. In 1990, Rice, Harris and Quinsey, found a
28 percent rate of recidivism over a period of four years. In a more recent study,
Marques, Day, Nelson and West (1994) found only an average 17.5 percent rate of
recidivism among rapists for a minimum period of 5 years.

Similar to the child molester recidivism studies, as the follow up period of rapist
recidivism studies is increased the rate of recidivism also marginally goes up. Van der
Werf (1991), found a recidivism rate of 17 percent among a sample of 119 rapists over a
period of 6 years. Grunfeld and Noreik (1986) studied 83 rapists over a period ranging
from 9 to 14 years, and found an average of 21.7 percent recidivism among the sample.
Soothhill and Gibbens (1978) have to date, conducted the recidivism study with the
longest follow up period, on a sample of rapists i.e. 22 years; and found an average sex
offense recidivism rate of only 22 percent.

As stated in the beginning of the chapter that for the purpose of the present study,
recidivism literature is useful to look at, yet it fails to address the question of
specialization in its entirety. The review of the recidivism literature however, has brought

a few facts to light, which shall be now separately elaborated upon. 1) That most of the

29

recidivism studies bifurcate the sex offender population into child molesters and rapists
and deal with each segment separately, which mars their empirical value from a
specialization standpoint. 2) Recidivism rates are subject to manipulation through the
operationalization of three factors i.e. criterion of recidivism, source of the data set and
adequate length of follow up period. So far, criminologists have failed to set a universal
standard for the operationalization of these factors, and without the existence of such a
standard, a comparison between different recidivism studies is not possible. Finally,
barring the two factors stated above, with the exception of the Prentky et. al. study
(1998), recidivism studies have failed to demonstrate that the sex offense recidivism rates
are signiﬁcant enough to prove specialization among either the entire sex offender
population or any subgroup of it.
Empirical Literature on Specialization among Sex Offenders

Specialization among sex offenders is an area that has been particularly neglected
in the ﬁeld of criminal justice. Thus far, there has been only one study that has actually
compared the criminal histories of convicted sex offenders and convicted non-sex
offenders to measure the incidence of sex offender specialization (Simon, 1997 & 2000).
However there have been other studies by criminologists that have claimed to measure
sex offender specialization but have failed to understand and establish a proper
specialization criterion to begin with. One such recent research that studied the
recidivism patterns of convicted offenders in a three year follow up period, revealed that
convicted rapists were 3.2 times more likely to commit another rape than convicted non-
rapists, hence proving a degree of specialization (Lagan and Levin, 2002). Lagan and

Levin (2002) also declared, that the odds for specialization in rape were greater that the

30

 

odds for specialization in robbery (1.7) or violent offenses in general (1.3). However,
specialization criterion used in this study states that “ specialists are prisoners who, after
being released, commit the same crime they were just in prison for” (Lagan and Levin,
2002), which brings it more in queue with recidivism than with specialization. Moreover,
like other recidivism studies this study only looked at rapes and failed to regard sex
offenses in their entirety.

AS has been mentioned above, Simon (1997 and 2000) has conducted the only
research that looks at sex offenders from a specialization standpoint. Using two data
sources simultaneously, one for child molesters and the other for violent offenders and
rapists and ending up with a pool of 493 subjects (142 child molesters, 290 violent
offenders and 51 rapists); she hypothesized that child molesters, rapists and violent
offenders have comparable versatility rates and do not specialize (Simon, 1997). To
measure specialization a versatility scale was created reﬂecting the sum of different types
of adult crimes that each offender committed. Using multiple regression analysis to
determine signiﬁcant differences in versatility, it was found that child molesters
demonstrated signiﬁcantly less versatility that rapists and violent offenders (Simon,
1997). The versatility among rapists and violent offenders on the other hand was
comparable (Simon, 2000). Simon (2000) attributes these recorded differences in
versatility, between child molesters and the other two categories, to an explanation
outside of the data. She states that due to the current publicity that surrounds heinous
sexual crimes involving children, child molester cases may deliberately being singled out

by public prosecutors for prosecution (Simon, 2000).

31

 

In précis, a specialization study that focuses on sex Offenders, similar to any other
specialization study, has to establish three basic criterions: a typology of crimes (in this
case sex offenses), the methodology employed for data collection and the statistical
technique used to measure specialization (Bursik, 1980). The typology question in this
case is to some extent prearranged. It has already been established that the policies, which
speciﬁcally deal with sex offenders, rarely make any distinctions among sex offenders
(Palermo et. al., 2001). Thus, an empirical evaluation of such policies should also not
make any such distinctions either and ought to regard sex offenders from a generic
perspective. The next issue that pertains to methodology of data collection has also been
principally predetermined. Although, much of the delinquency remains hidden from
ofﬁcial data, which the self-report data might unravel, the fact that a relatively accurate
sequence of such delinquency can only be established through the ofﬁcial data, make
them a requisite for conducting specialization studies (Bursik, 1980). The question of
statistical technique on the other hand still remains an object of contention. For example,
the Simon study (1997 and 2000) follows through with the ﬁrst two criterions, however
the statistical method used in that study is by no means universal. The statistical
technique to be used in a specialization study depends mainly on the nature of the data
understudy and the scope of the study, thus rendering it impossible to make a
comparative analysis as to which technique is the over all best. Typically, criminologists
have used the Markov chain (Bursik, 1980), the transition matrix (Benson and Moore,
1992 and Piquero, 1999), Logistic Regression (Schwaner, 2000) and two-way successive
crime type tables (Britt, 1996) to measure specialization. For the proposed study

however, the best statistical technique would be the survival analysis, primarily because

32

this type of analysis regards the sample in uniformity over time rather that reporting mere
percentages of recidivism from the beginning point to the ending point in the study
(Prentky et. al., 1998). There is also the added advantage that the survival analysis is
compatible with prospective studies, as this study will also follow a quasi-prospective
model. An extensive description of the data set and the statistical methodology to be used
in the study shall be explicated in the next chapter.
RESEARCH OBJECTIVE

After reviewing the theoretical precepts upon which specialization, sex offender
specialization and the generalist perspective are buttressed upon and the empirical
literature available on them, it is eminent that the research object of the proposed study be
established. The primary objective of this study as mentioned in the introduction is
whether sex offenders specialize or not as compared to other violent offenders. In this
case robbery and assault shall signify other violent offenses. Under the present scenario
sex offender specialization can only be established by answering three empirical
questions: a) whether all violent offenders specialize (i.e. sex offenders, assaulters and
robbers specializing within their violent sub categories), b) whether sex offenders and
assaulters belong to the same category of violence and specialize within it while robbers
do not (this hypothesis shall be discussed extensively in the methods section) and c)
whether only sex offenders specialize while robbers and assaulters do not.

METHODS SECTION

This section shall instate the empirical aspect of the proposed study by laying

forth a description of the methodology to be employed for conducting this study. The

methodology of any study is chieﬂy reliant upon on three concepts: the acquired or

33

 

collected data set, the operationalization of the data set in context of the research
hypothesis, and the statistical techniques employed to empirically verify the research
hypothesis. Accordingly this section shall be divided into the aforementioned three
sections.
Data

The original data set comprised of the arrest dates, the charges levied, the names,
addresses and some other demographic characteristics of all the people arrested in the
mid-Size Michigan City in between the years 1989 and 1999. These data were collected
by the Police Department of that city. For ethical and security purposes, unique
identifying numbers were used to replace the names and addresses of the arrestees, before
this study was commenced. The total number of arrests contained in the original data set
was 138,833. Using SPSS the data set was restructured, so that each row chronologically
laid out the arrests for each person (or unique identifying number) in the 10 years these
data were collected. Initially, all the female offenders were ﬁltered out of the data set, as
studies have shown that sex offenses are chieﬂy a male phenomenon while robbery and
assault are not, keeping females in the sample would thus have introduced a bias in the
favor of non specialization of sex offenders. Then, the people that were arrested at least
once for a violent charge in their criminal careers within the 10 years understudy were
selected. A violent charge was deﬁned to include felonious assault, robbery or criminal
sexual conduct. As the main purpose of this study is to empirically evaluate specialization
among sex offenders in comparison to other violent offenders, the incidence of at least
one violent arrest in each case was necessary. Later, out of the people with at least one

violent arrest, only those arrestees were selected who had commenced their criminal

34

careers as juveniles (18 years or less) within the 10 years understudy, either violently or
non violently, resulting in a total number of 800 cases. As it has been the contention of
criminologists that age brings about a decline in crime versatility exhibited by the
offender (Peterson et. al., 1980; Gottfredson and Hirschi, 1983; Gottfredson and Hirschi,
1990 and Piquero et. al., 1999) this study aims to capture the very time in the criminal
careers of the arrestees where they have started off with a violent or non violent offense
as juveniles and have thus stepped into adulthood, to see if there is an inclination towards
specialization later in the career. The remaining 800 cases left in the data set were then
merged with two other data sets, the Michigan Department of Corrections felony
conviction database and the Michigan Department of Corrections Juvenile conviction
database, to see if these offenders were convicted as juveniles either for a violent or a
non-violent offense. The convictions incurred by each subject as an adult along with the
charges with in the ten years understudy were also merged into the dataset. However, it is
important to highlight at this point that the adult and juvenile conviction records so
obtained were extracted from data sets that represented the entire State of Michigan not
just the mid-size Michigan city. Hence, these convictions might have been the result of
offenses committed out side of the mid-size Michigan city but within the State of
Michigan. Resultantly, for empirical reasons this study does not claim compatibility
between the arrest and conviction records at any level. The conviction variables were
treated as a separate set of dependent variables completely independent of the ﬁrst set of
dependent variable (the arrest records) but underwent the same statistical procedures in
order to see if the results produced by both sets were congruent. Albeit it has been seen in

the examination of the recidivism and specialization literature that a recidivism criterion

35

based on conviction may cause artiﬁcial shrinkage in the acquired data set and the
statistical signiﬁcance of the effect so produced compared to a recidivism criterion based
on arrest (Russell, 1982; Simon, 1997 & 2000; Doren, 1998 and Prentky et. al. 1998),
however the direction of such results should be the same. Also, the increased geographic
circumference of the conviction records (Michigan instead of mid-size Michigan City)
may mollify the above discussed shrinkage effect to some extent.

Now that the source and nature of the data employed for conducting this study
and the subsequent ﬁltration processes ran on it have been explained, its suitability for
conducting the proposed study shall be discussed. Originating from an ofﬁcial source, the
data was later merged with two other ofﬁcial data sets illustrating it to be reliable. Also,
as stated before in the previous chapter, much of the delinquency that remain hidden from
ofﬁcial data, might be unraveled by the self reported data, albeit a relatively accurate
chronological sequence of such delinquency can only be established through the ofﬁcial
data (Bursik, 1980). The data prevail for a time span of 10 years, rendering it adequate
for conducting a longitudinal specialization study. However, some criminologists opine
that a reliable sex offender’s specialization study should have a follow up period of at
least 20 years (Hanson et. al., 1995 and Prentky et. al., 1998). The sample of 800 cases to
be studied represents the entire population of the people who were arrested as juveniles
and had at least one violent arrest within the duration of the study period in the mid-size
Michigan city. Hence, the results extracted out of such a sample would more
generalizable as compared to the clinical psychology studies evaluated in the previous
chapter that employed samples of convicted sex offenders who had served time in

maximum-security prisons (Marnie et. al. 1991; Hanson et. al., 1993 and Schlank, 1995).

36

The inclusion criterion is based on arrest as compared to conviction for three
reasons. Firstly, a conviction-based criterion dramatically reduces the sample size,
making it less representative of crime in actuality as compared to an arrest-based
criterion. Secondly, an arrest-based criterion takes out the effect produced by plea-
bargaining, which a conviction-based criterion cannot. Thirdly, on a general note, not all
guilty people, once injected into the legal system necessarily get convicted. Conversely,
the ﬂip side of the same argument would be that not all the people arrested are actually
guilty. However, this study also aims to look at the conviction records as a measure of
cross checking the results obtained by the analysis of the arrest variables. Furthermore,
violence was operationalized to include only arrest made for felonious assault, robbery,
or a criminal sexual conduct. Violent misdemeanors were not included, thus narrowing
the scope of the study to only specialization and not escalation. Lastly, another limitation
in the data set was introduced by the enactment of the Sex Offender Registration Act
(1994) in the state of Michigan (Freeman-Longo, 2000 and Palermo et. al., 2001). There
is to date no empirical study that has measured the effect produced by these laws if any,
on the sex offender recidivism rates.

Variables

This section will include a generic description of the independent and dependent
variables employed in the study.

Independent Variables.

The independent variable “1‘“t Violent Offense” was employed to classify the 800
subjects into three distinct categories of violence, on the basis of their ﬁrst violent arrest.

This was a categorical variable which classiﬁed the ﬁrst violent arrest into a C.S.C.,

37

Robbery or Assault (C.S.C. = 1, Robbery = 2, and Assault = 3). The second independent
variable was continuous and measured the age of the arrestee at the time of ﬁrst offense
(“Agez 1St Violent Arrest”). Similarly for the conviction analysis the continuous variable
“Age at 1st Conviction” was calculated, which demarcated the beginning of the criminal
career for each offender from a conviction perspective. “Race” was the third independent
variable, which was divided into 4 categories (White = 1, Black = 2, Hispanic = 3, Other
= 4). The fourth independent variable counted the total number of non-violent arrests for
each subject prior to their ﬁrst violent arrest (“Total Prior Arrests”). A dummy variable
for “Juvenile Incarceration” was also computed which had a ‘0’ for all the subjects who
did not have a juvenile incarceration and a ‘1’ for all those who were incarcerated as
juveniles.

Dependent Variables.

The ﬁrst dependent variable (“Total Post Violence Arrests”), counted the total
number of arrests for each arrestee after their ﬁrst violent arrest. The next three
dependent variables were derivatives of the “Total Post Violence Arrest” variable, by
dividing it into three distinct categories based on the different charges levied in each post
violent arrest ("Total Post Violence CSC”, “Total Post Violence Robbery” and “Total
Post Violence Assault”). The “Assault Conviction”, “CSC Conviction” and “Robbery
Conviction” variables computed the number of times each subject was convicted for an

assault, CSC or Robbery respectively.

The dependent variable “C. Career Length (General)” was calculated by
subtracting the date of the ﬁrst arrest from the last day that each subject was observed, so

as to indicate the maximum length of each arrestee’s criminal career within the data set.

38

Likewise, “C. Career Length (Violent)” variable denoted the maximum length of each
offender’s violent criminal career within the data set. The dependent variable “At Risk
Period”, computed the at risk time for each offender of getting arrested after their ﬁrst
violent arrest by subtracting the last day that every one was observed with the date of
arrest subsequent to the ﬁrst violent arrest. Lastly, the variables “Time to Failure for
CSC”, “Time to Failure for Robbery” and ”Time to Failure for Assault” calculated the
time between arrest date for a CSC, robbery or assault respectively, subsequent to the
ﬁrst violent arrest or if no such arrest was present within the data set then the last day that
each subject was observed.
Statistical Analysis

This study proposes to apply a bi-faceted approach to measure specialization. This
approach works on the premise that theoretically, specialization is founded on two focal
concepts, a) the time to next event (arrest, in this case) and b) the total count of the
occurrence of the event (or arrest). To operationalize these concepts this study aims to
employ the aforementioned longitudinal data and the variables thus extracted from the
data set in a bimodal approach.

Model I

The ﬁrst model aimd to measure the time to the subsequent arrest for each subject
starting from the time that they committed their ﬁrst violent arrest, by using the survival
mode]. The survival model is a special kind of probability distribution (London, 1988),
which commences after the “occurrence of an event” (ﬁrst violent arrest) and determines
either the “time to failure” (the date of the subsequent arrest) or the “length of future life

time” (the last date a subject was observed, for those subjects that did not recidivate after

39

 

their ﬁrst violent arrest). This model is employed in prospective studies involving
longitudinal data (London, 1988). To meet these requirements the present study was
performed quasi- prospectively i.e. the data albeit not collected prospectively was treated
so in the statistical analysis. However, using the survival model on the mid-size Michigan
City arrest data introduced one limitation to the study. As the study lasted for only ten
years all those subjects who did not recidivate after their ﬁrst violent offense were
considered to have “not failed” in the statistical analysis although this may not have been
the case in reality (Lee, 1992). This limitation, termed as “right censoring”, introduces an
error in the favor of the null hypotheses (sex Offenders do not specialize). Also, some
subject may have migrated out of Michigan to other states during the period the data was
being collected, were counted in the data set as non recidivists, which may have not been
true in actuality.
Model [I

This model measured specialization by computing Lambda i.e. the
frequency/count of arrests for each subject and dividing the resulting variable by the
number of days observed while “at risk” of arrest. This approach has been previously
utilized in research pertaining to the criminal career paradigm by Blumstein et. al. (1986).
The count began from the CSCS, robberies or assaults respectively subsequent to the ﬁrst
violent offense, and the ‘at risk’ period was initiated from the date of ﬁrst violent arrest to
the last day that each subject was observed.

The results obtained from Model I and II were sufﬁcient to determine whether
offenders who initiated their violent careers with a CSC, assaults or robbery specialize in

that type of violence later on in their criminal careers or not, with reference to time to

40

next arrest and the total frequency within the observed period. However, certain pivotal
questions still remained unanswered. The generalization of a scientiﬁc theory once
properly hypothesized and processed through various forms of empirical testing depends
on three things: Regularity, whether the direction of effect produced by each test is the
same or not; Size, what is the magnitude of the effect produced in each test; and
Statistical Signiﬁcance, whether the results so produced are statistically signiﬁcant or not.
Applying the same criterion to the proposed study, in order to measure regularity, the
direction of the coefﬁcients so produced after the application of model I and H, were
checked. To measure size, the resulting coefﬁcients in each model were compared to
other coefﬁcients of the same model to see if the effect produced by sex offenders was
greater than that of other offenders.

Finally, the most pivotal question in regard to this study still remains unanswered:
in the light of the results what is the criterion of specialization? Tables 1A, 13 and 1C
roughly represent the result format after the negative binomial has been run on model 11
and the 2 test has been run on model I. The rows represent the resulting coefﬁcients of the
two dependent variable i.e. Robbery and Assault. Here it may be observed; that CSC is
excluded as in both the regression models (survival model and negative binomial) CSC
was the excluded group. The columns are representative of the independent variables
“Type of 1St Violent Arrest” categorized as CSC, Robbery and Assault. The hypothetical
results shown in Tables 1A, 1B and 1C are representatives of the three instances of
perfect specialization among sex offenders. The results of table 1A show that the sex
offenders have refrained from committing any robberies and assaults (represented by the

two negative relationship), robbers have only committed robberies further in their

41

criminal careers and no assaults (represented by a positive relationship and no
relationship respectively) while assaulters have perpetrated no robberies and only assaults
within their criminal careers (denoted by no relationship and a positive relationship
respectively). Hence the results show in table 1A symbolize perfect specialization for all
violent offenders within their particular violent subgroups (CSC, Robbery and Assault).
The results of table 1B Show that sex offenders refrain from committing robberies yet
transverse into assaults (signiﬁed by a negative relationship and a positive relationship
correspondingly). Robbers on the other hand commit both robberies and assaults further
in their careers (denoted by two positive relationships). Finally, assaulters perpetrate only
perpetrate assaults within their criminal careers (represented by no relationship and a
positive relationship). These results prove specialization on the premise that robbery
although a violent offense is essentially different from a sex offense and an assault. In sex
offenses or assaults, violence is the motive for the offense while in robbery, Violence is
used as an instrument to further materialistic the motivations of the offenders. Hence the
results of table 1B show that sex offenders and assaulters specialize while robbers do not.
Lastly, the results of table 1C Show that sex offenders refrain from perpetrating either
robberies and assaults (denoted by two negative relationships) while both robbers and
assaulters transverse into other categories of violence (represented by two positive
relationships in both columns respectively) further within their criminal careers. Thus, the
results of table 1C symbolize that only sex offenders specialize within their violent
subgroup. The above—mentioned outcomes represent three scenarios of perfect

specialization among sex offenders. Albeit, in real life such perfect relationships are

42

rarely to be found, however these exemplars have expounded the opposite poles of a line

along which the actual value of specialization may lie.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Table 1A

CSC Robbeq Assault
Robbery - + 0
Assault - 0 +
Table 1B

CSC Robbery Assault
Robbery - + 0
Assault + + +
Table 1C

CSC Robbery Assault
Robbery - + 4-
Assault - + +

RESULTS

Univariate Results

Beginning with the independent variables, Table 2 shows that within the measure
“1” Violent Arrest”, the majority of the people who commenced their violent careers with
an assault (70.4 %), while only 20.3% and 9.3% started off their violent careers with a
robbery or a CSC respectively. The mean of 16.92 years and a median and mode of 17.10
years and 14.54 years respectively for the measure “Age: 15t Violent Arrest” establish the
fact that most of the offenders were juveniles when they were arrested for their ﬁrst
violent offense. The mean, median and mode of the measure “Age: 1‘5t Conviction” shows
that a majority of subjects were convicted for the ﬁrst time as adults (19.03, 18.65 and
18.11). As the “Age: 1St Conviction” measure contains only adult conviction records for

each subject, its mean, median and mode are representative of the fact that on average

43

 

 

 

offenders within the data set got convicted very early after they reached adulthood. Only
8% of the 800 offenders had a juvenile conviction either for a violent or a non-violent
offense. The racial distribution within the data set as shown by the “Race” measure
signiﬁed that Blacks represented the majority (57.3%) within the data set, while Whites
(30.8), Hispanics (11.4) and Others (0.6%) stood in minority. The variable ”Total Prior
Arrests” shows that each subject on average was arrested once for a non-violent offense
(Mean = 1.10) before his ﬁrst violent arrest. However, after looking at the range (13.0)
standard deviation (1.77), median and mode (Both zero) it is evident that the mean
misrepresents the dispersion of the arrest prior to the ﬁrst violent arrest and that for most
of the subjects their ﬁrst violent arrest was their ﬁrst arrest. Thus contrary to the concept
of escalation this study shows that instead of commencing with a non-violent minor
offense and gradually escalating to violent offenses, most of the violent offenders started
off with a violent offense.

As regards the dependent variables, albeit the mean and median (1.72 and 1.0) of
“Total Post Violence Arrests” measure show that on average each arrestee incurred one
arrest after their ﬁrst violent arrest, the range (16.0), standard deviation (2.29) and mode
(0) reveal that the such arrests are irregularly distributed and that a considerable chunk of
the population did not get arrested after their ﬁrst violent arrest. The mean (0.03),
standard deviation (0.17) and range (2.0) for the “Total Post Violence CSC” measure
illustrate that a very small segment of the population committed a CSC after their ﬁrst
violent arrest. This trend is conﬁrmed when one looks at the univariate statistics produced
for the “CSC Conviction” variable (Mean = 0.02, SD. = 0.16 and Range = 2). AS

compare to the “Total Post Violence CSC” variables the “Total Post Violence Robbery”

and “Total Post Violence Assault” variables show that robbery arrests were almost three
times more (Mean = 0.09, SD. = 0.31 and Range = 2.0) and assault arrests were almost
14 times (Mean = 0.44, SD. = 0.9 and range = 7.0) more than the CSC arrests within the
data set. The trend of the “Total Post Violence Robbery” variable corresponds with that
of the “Robbery Conviction” variable (Mean = 0.08, SD. = 0.3 and Range = 3.0),
however there is a marked difference between the descriptive statistics of “Total Post
Violence Assault” and the “Assault Conviction” variable (Mean = 0.18, SD. = 0.47 and
Range = 3.0). Although, the reason for this difference is unknown, it has been previously
Stated that this study does not claim compatibility between the arrest and conviction
records.

The average general exposure period (time between the ﬁrst arrest and the last day that
last day that each subject was observed) for all the offenders captured within the data set
was almost 5 years (Mean = 1851.45 days). Average violence exposure period (time
between the ﬁrst violent arrest and the last day that last day that each subject was
observed) was around 4 years (1476.88 days). However, the Standard Deviations (989.9
and 890.0 days respectively) for both these measures show that the lengths of these
durations vary a great deal from case to case. Each subject was on average at risk of
failing (getting arrested for a violent offense) for almost 2 years (Mean = 733.58 days).
The mean time to failure for CSC (Mean = 1425.50) was marginally more than the time
to failure for robbery (Mean 2 1393.79). However there is almost a year’s difference
between the mean time to failure for CSC and the mean time to failure for assault
(1016.50 days), indicating that out of the three violent offenses on average the subjects

failed by committing assaults much earlier in comparison to robberies and CSCs.

45

 

 

 

 

889x

 

 

emote. .nom cog—22>
3o 2. o.o sod ca 5.893. 8835 eon E a 38528 use See
amok—4
Omu 886$
: .o o.o 9o 86 am 985. 80 8535 eon E a 88580 :8 Rae
manta.
38.5w .mm< 8:205
86 o.o od 38 cs :33. 88.05 :8 E a Boséeoo :3 Bob
Emcee
A: od o.o S a 3: e95 .8 a 88528 BE Boa
=OEQ_>COU
N: :2 SE 8.2 S: was» .6 a 88558 a a?
“mug “co—OQV
43 an: 2.: No.2 85 e8» e a 88528 a Home.
d1 282 :33: :82 sues. «5x6 aluls 35 azure,
V0 9.x .05 .>=—. H ﬂ €23.89:on
.05 .>:—. O: H O Eotowouau 22.52:.

m Ed 850 n a

a 4.: SENSE u m

me. 0% seem n N
9% men 323 u a .8585 8am

8m 4.2 :23? n m
we now cannon n m :22
E one 08 us 38036 “835 .2
.355: Z ounce wemgo rewcuev 83a€a>

 

memes 35 newcwmceﬁ eécmtoneQ
N 2.1:.

46

 

ﬁmmw

waw

meow

0wa

Odom

adwo

omd

Ed

Edd

mmd

odmm

odmm

0.5mm

od

0.0mm

odoE

od

od

o.o

ed

00::

922

mémi

m.NOm

WmVE

025

ad

od

o.o

o.—

0.82

wdmﬂ

v.33

9mm

0.05:

vgmwﬁ

mod

Nod

2.0

Nb.“

o.m~wm

o.m~wm

o.:wm

0.5mm

o.:wm

cdowm

o.m

o.m

Qm

92

“383‘ Eon—Oomnnm use
“more H555 .L 5953 DEC.

ﬂopped Heoegmnem Ea
58.2 5205 .L 5033 DEF

0m0 055035 98

58.2. 05.05 i 5953 DEF
52.14 0:205 .L 5% 52.3

at Sea I :oumtomno .5 >5 3.3

38.2 5205
“L can I 593.355 we >5 33

Hebe. “L
we can I 5%?530 no .95 $3

55350 E wet—30m Dom «o t
55350 E wen—30m 0m0 be t

55350 E mes—Boa 0m0 we #

Sacha cop—22> “mom Eek

3m

wDODCUr—OU

952550

wDOSCUﬁOU

£52550

£52550

£52550

wSOSGUr—OU

mSOSH—SCOU

505350

mDOSCUH—OU

"2
0333..
50 828m
9 0:5.
33%
Be 22E
9 DEC.
Omu

Le ease
OH. 055.

55m HE to.
3223
55m
ocsmaxm
€559
5Com
8:59am
55350
conga
55350
0m0
55350
5894
Smog
coco—05

“mom .38.

 

#3303 N 29d.

47

Bivariate Results

The results of the bivariate association are show in tables 3, 4 and 5. Table 3 shows a
cross tabulation analysis between the classiﬁcation variable “Type of lSt Violent Arrest”
variable and the other two categorical independent variables namely race and juvenile
incarceration. The results reported are as follows, out of the 246 subjects who were White
12.2 percent started off their violent criminal careers with a CSC, 14.2 percent with a
robbery and 73.6 percent with an assault. The dataset also contained 458 Black offenders
out of which 7.6 percent began their violent career with a CSC, 25.2 percent with a
robbery and 67.2 percent with an assault. Out of the 91 Hispanic subjects 9.9 percent
commenced their violent career with a CSC, while 13.2 percent and 76.9 percent with a
robbery and assault respectively. The entire population of subjects belonging to races
other than White, Black or Hispanic (5 offenders in total) began their violent careers with
an assault. Only 9.4 percent out of the total of 64 offenders who had prior juvenile
records initiated their violent careers with 21 CSC, while 31.2 percent and 59.4 percent of
the subjects out of this population began with either a robbery or assault respectively. The
results explained above indicate that although Blacks are in majority within the dataset,
percentages wise represent the majority only for people who began their violent careers

Table 3
Cross tabulation between Race and Type of 1“" Violent Arrest

 

Race J uv.Inc.

 

White Black His anic Other

 

Total 246 458 91 5 64
Type of lSt Violent Arrest
CSC (%age) 12.2 7.60 9.90 0.0 9.40
Rob. (%age) 14.2 25.2 13.2 0.0 31.2
ASS. (%age) 73.6 67.2 76.9 100 59.4

48

with a robbery. With a difference of a mere 3 percent between the two, Whites and
Hispanics represent the majority for both the people who commenced with a CSC and the
people who began with an assault. Most interestingly, in the context of this study Whites
were seen most likely to begin their violent careers with a CSC as compared to the other
races. Also out of the three categories, the people whose ﬁrst violent arrest was a CSC
were least likely to have been incarcerated as juveniles.

Table 4 shows the comparison of means between the classiﬁcation variable (Type
of lSt Violent Arrest) and the two continuous independent measures (Age at 1St Arrest and
Total Prior Arrests). The results indicate that the mean age of all the people who initiated
their criminal careers with a CSC was the highest as compared to robbery and assault
while the mean number of prior arrests was the lowest. Further analysis of these statistics
revealed that only 22.97 percent of the people who began their violent careers with a CSC
had a prior non-violent criminal record in comparison to 54.32 percent and 45.92 percent
for the people who commenced their violent careers with a robber or assault respectively.

Table 4
Comparison of Means between Type of 1" Violent Arrest and Age at I” arrest

Type of First Violent Arrest
CSC Robbery Assault

Age at lSt Arrest 16.80 16.43 15.91
Total Prior

Arrests 0.34 1.42 1.1 1

Table 5 shows the comparison of means between the categorizing independent
measure and the three dependent arrest measures (Total Post Violence CSC, Total Post
Violence Robbery and Total Post Violence Assault) along with the three dependent

conviction measures (Assault Conviction, CSC Conviction and Robbery Conviction).

49

The results illustrate that regarding arrest, people who initiated their violent careers with
a CSC had more post violence CSC arrests (0.08) than robbery and assault (0.01 and
0.03). This effect is even more pronounced in the bivariate analysis between the
independent CSC category and the “CSC Conviction” measure (0.15 for CSC, 0.00 for
Robbery and 0.01 for Assault). Similarly, for the people who commenced their violent
career with a robbery the mean number of arrests and convictions for robberies later in
their career (0.18 and 0.28) was higher compared to the people who initiated with a CSC
or assault. For the people who started off with an assault although the mean number of
assault arrests later in their careers is by far the greatest (0.51), the mean number of
assault convictions is not (0.17). The bivariate analysis of the “Assault Conviction”
measure illustrates that the people who started their violent career with a robbery had the
highest mean number of assault convictions later in their careers.

Table 5
Comparison of Means between Type of 1‘" Violent Arrest and the Arrest and Conviction

measures
Type of First Violent Arrest

 

QC Robbery Assault
Total Post Violence CSC Arrests 0.08 0.01 0.03
Total Post Violence Rob. Arrests 0.04 0.14 0.08
Total Post Violence Ass. Arrests 0.15 0.33 0.51
CSC Conviction 0.15 0.00 0.01
Robbery Conviction 0.00 0.23 0.04
Assault Conviction 0.01 0.28 0.17

 

Multivariate Results

It has been previously explained in the methods section that the multivariate
analysis shall encompass a bimodal approach (the Cox Regression and the Negative

Binomial Test) to measuring specialization.

50

Model I

In implementing model 1, the Survival Model was executed on the three arrest
measures only (as there was not enough accurate data to run it on the conviction
measures) and the key independent measures thus obtaining their respective hazard rates.
The cumulative betas and the corresponding standard errors are reported in table 6. One
thing that has to be born in mind while looking at the results from all the regression
models is that the sub category ‘CSC’ out of the categorizing measure “Type of 1“
Violent Offense” and the sub category ‘White’ out of the “Race” measure have been
treated as comparison categories. Meaning that the results shown for the rest of the sub
categories of the two above mentioned variables are in fact their coefﬁcients in relation to
their respective comparison categories. Going back to the results reported in table 6, the
“Total Post Violence CSC Arrests” were observed to have a negative (-2.0 and -1.32)
signiﬁcant relationship (p < 0.01) with those subjects within the data set who initiated
their violent career with either a robbery or an assault, as compared to people who began
with a CSC. Thus, according to the results obtained by this model people who initiated
their violent careers with either a robbery or assault were signiﬁcantly less likely to
commit a sex offence later in their careers as compared to those offenders who started off
with a CSC. These results are in accord with the bivariate test results and record a
signiﬁcant degree of specialization among sex offenders. In addition, signiﬁcantly more
CSC arrests were born by Hispanics as compared to Whites. Evaluating this observation
against the racial constitution of the Type of 1St Violent Arrest measure reveals that
although Whites are more likely to commence their violent careers with a sex offense,

once the violent career has started Hispanics are signiﬁcantly more likely to get arrested

51

for committing sex offenses compared to Whites. “Total Post Violence Robbery Arrests”
is shown to have positive signiﬁcant relationships with Blacks and a negatively
signiﬁcant relationship with “Age at 1St Arrests”. Corresponding with the results of table
5 Blacks were observed not only more likely to commence their violent careers with a
robbery but were also signiﬁcantly likely to get arrested for committing robberies later
on. In relation to “Total Post Violence Assault Arrests” there is present a strong
signiﬁcant relationship (p < 0.005) between offenders who commenced their careers with
an assault as compared to the subject who began with a CSC. Furthermore, there is a
signiﬁcant relationship (p < 0.05) between the number of assault arrests incuned and

Table 6
Survival Modelfor the Arrest measures with the Type of 1“ Violent Arrest and Race

Total Post Violence Total Post Violence Total Post Violence

 

CSC Arrests Rob. Arrests Ass. Arrests
_b S. E. b E b S. E.
Type of 1St
Violent Arrest ** ***
Robbery -2.0 O.84** 0.79 0.63 0.60 0.40
Assault -1.31 0.50** 0.45 0.60 1.09 0.36***
Race
Black 0.66 0.59 0.99 0.37** 0.35 017*
Hispanic 1.36 065* 0.50 0.53 0.27 0.24
Other -10.1 893.8 -9.18 388.9 0.67 1.01
Total Prior
Arrests 0.02 0.16 -0.07 0.08 0.01 0.43
Age at 1st Arrest -0.00 0.00 -0.01 0.00* -0.00 0.00

 

*p < .05; **p < .01; ***p < .005

52

being Black, as compared to being White. Hence showing that although Whites were
more likely to commence their violent career with an assault Blacks were signiﬁcantly
more likely to be rearrested for committing assaults later in their careers, as compared to
Whites.

Model 11

Certain changes were made in the categorical independent measures in order to
facilitate the execution of the negative binomial test. The “Type of 1St Arrest” measure
was converted into three dummy measures (CSC, Robbery and Assault) representing
each violent sub category. The same operation was performed on the “Race” measure.
The CSC (dummy) measure and the White (dummy) measure were treated as comparison
categories. The results of the negative binomial test as applied upon the two sets of
dependent measures are reported in tables 7 and 8.

The “Total Post Violence CSC Arrest” variable (as shown in table 7) has a
negative signiﬁcant relationship (p < 0.05) with the subject who commenced their violent
careers with a robbery or assault as compared to the people who stated off with a CSC. .
The negative relationship to assault and robbery indicates that people whose violent
careers began with a robbery or an assault had signiﬁcantly less CSC arrests later in their
criminal careers as compared to the people who commenced with a CSC. Hence
corroborating the evidence found in the survival Model regarding sex offender
specialization. Also, being Hispanic had a positive signiﬁcant relationship (p < 0.05) with
getting arrested for committing CSCs subsequent to the ﬁrst violent arrest. This fact is
also validated by the results shown in table 6. “Total Post Violence Robbery Arrest”

variable shows a positive signiﬁcant relationship (p < 0.005) between Black arrestees and

53

Table 7
Negative Binomial for the Arrest measures with the Type of 1" Violent Arrest and Race

Total Post Violence Total Post Violence Total Post Violence

 

CSC Arrests Rob. Arrests Ass. Arrests
Constant -2.67 1.58 -1.22 1.07 -1 .05 0.62
Robbery -l.89 0.84* 0.93 0.64 0.65 0.39
Assault -1.12 0.51* 0.52 0.62 1.18 0.35***
Black 0.75 0.59 1.07 0.37*** 0.53 0.18***
Other -25.7 144.3 -26.4 131.1 0.20 0.93
Hispanic 1.47 0.67* 0.58 0.54 0.48 025*
Total Prior
Arrests -0.09 0.16 -0.1 1 0.08 -0.06 0.04
Age at 1“t
Arrest —0.03 0.08 -0.02 0.05** -0.07 0.03*
Aljha 1.04 2.13 0.86 0.76 1.73 0.33***

 

*p < .05; **p < .01; ***p < .005

a negative signiﬁcant relationship (p < 0.01) with “Age at lSt Arrest”. Hence showing that
a) Blacks are more likely to get arrested for committing robberies subsequent to their ﬁrst
violent arrest and b) the lower the age at ﬁrst arrest the higher the likelihood is of getting
arrested for robberies, once the violent career has commenced. The Survival Model again
conﬁrms these results. “Total Post Violence Assault Arrest” measure shows a positive
signiﬁcant (p < 0.01) relationship with arrestees who began their violent careers with an
assault as compared to subject who began with a CSC. Furthermore, this dependent

variable has a positive signiﬁcant (p < 0.01) relationship with the Black subjects in the

54

data set and a negative signiﬁcant (p < 0.05) relationship with age at ﬁrst arrest. The the
Survival Model, again conﬁrms these results with the exception of the fact that the
relationship between age at ﬁrst arrest and total assault arrests after ﬁrst violent arrest is
not shown to be signiﬁcant in table 6.

Regarding the conviction measures it was not possible to run the negative
binomial test on the “CSC Conviction” measure because the estimated variance matrix of

Table 8
Poisson Regression and Negative Binomial for the Conviction measures with the Type of

1'" Violent Arrest and Race

CSC Conviction Robbery Conviction Assault Conviction

Coef. S. E. Coef. S. E. Coef. S. E.
Constant -2.40 1.05 * -33.6 25.3 —6.04 1.24***
Robbery 4.05 l.02*** 31.7 25.2 2.99 l.02**
Assault -5.56 0.32*** 30.03 25.2 2.42 1.01 **
Black -0.32 0.30 -0.27 0.51 0.10 0.23
Other -26.8 1461657 ~30.2 97.1 -28.1 137.3
Hispanic -1.25 0.74 -0.87 1.03 0.51 0.31
Total Prior
Arrests 0.33 0.13 0.02 0.10 0.10 005*
Age at 1St
Arrest 0.00 0.00 -0.04 0.10 0.09 0.04*
Alpha 0.47 3.46 0.94 0.44*

 

*p < .05; **p < .01; ***p < .005

Note: Poisson regression was used to analyze the CSC Conviction measure instead of the
Negative Binomial model because the estimated variance matrix of the estimates was found to be
singular for the CSC Conviction measure.

55

the estimates was found to be singular. Thus the Poisson regression was used to analyze
the CSC Conviction measure’s relationship to the key independent measures instead of
the Negative Binomial model. However for the assault and robbery conviction measures
the Negative Binomial method was used to test existence of specialization across
chronological conviction records for each offender. The Poisson regression results for the
CSC Conviction measure show a very signiﬁcant (p< 0.005) negative relationship
between the commencement of a violent career with a robbery or assault and number of
CSC convictions incurred later in the violent career, as compared to beginning a violent
career with a CSC. Preliminary evidence of specialization among sex offenders as
regards their conviction records mentioned in the bivariate analysis is buttressed bythe
results shown in table 8. Hence, the results of both the multivariate models have revealed
that such specializations exist on the arrest as well as the conviction level. Surprisingly,
the “Robbery Conviction” measure does not show any signiﬁcant association with any of
the independent measures. Reverting back to the results of the bivariate analysis shown in
table 5 it was observed that the people who commenced their violent career with a
robbery were most likely to get convicted for robberies later in their careers; however this
trend was not replicated by the results of the Negative Binomial test shown in table 8.
The numbers of assault convictions had a signiﬁcant positive relationship (p < 0.01) with
the people who commenced their violent careers with an assault or robbery as compared
to the people who began with a CSC. The age at ﬁrst arrest was also observed to be
positively associated (p < 0.05) to the number of assault convictions during the course of
the violent criminal career for each offender. Lastly, for the ﬁrst instance in any of the

multivariate tests, the results relayed in table 8 showed that there was a marginally

56

positive (0.09) signiﬁcant association (p < 0.05) between the total prior non-violent
arrests and the number of assault convictions incurred after the ﬁrst violent arrest.

In order to establish a measure for further cross checking the results of the two
regression models the General Linear Model was also run on the two sets of dependent
variables. The results shown in table 9 represent the out come of the GLM as it was run
on the arrest and the key independent measures, and coincide completely with the results
Table 9
General Linear Model for the Arrest Measures with the Type of 1‘" Violent Arrest and

Race

Total Post Violence Total Post Violence Total Post Violence

 

CSC Arrests Rob. Arrests Ass. Arrests
9 Si 9 .S-_E- 9 ﬂ
Intercept 0.08 0.05 0.23 0.08** 0.48 1.25*
Type of 1St Violent
Arrest
Robbery -0.07 0.02** 0.07 0.04 0.13 0.13
Assault -0.06 0.02** 0.03 0.04 0.34 0.11**
CSC 0.00 0.00 0.00
Race
Black 0.02 0.01 0.07 0.03** 0.20 0.07**
Hispanic 0.05 002* 0.03 0.04 0.19 0.11
Other -0.01 0.08 -0.02 0.14 0.05 0.40
White 0.00 0.00 0.00
Total Prior Arrests 0.00 0.00 -0.01 0.01 . -0.02 0.02
Age at lSt Arrest -0.00 0.00 -0.01 0.00** -0.01 0.01*

 

*p < .05; **p < .01; ***p < .005

57

of the survival model and the negative binomial model with only one exception. The

number of post violent criminal career initiation assault arrests, were observed in the

Negative Binorninal to have positive signiﬁcant (p < 0.05) association with being

Hispanic as compared to being black, which the GLM failed to capture. However, this

association was not present in the Survival Model either.

The results obtained for the CSC conviction variable (shown in table 10) also

evidence the existence of specialization among sex offenders, shown in the Negative

 

 

gzzhialloLinear Model for the Conviction Measures with the Type of 1" Violent Arrest
and Race
CSC Conviction Robbery Conviction Assault Conviction
h E D ﬂi- 9 ﬂ
Intercept 0.31 0.07*** -0.05 0.08 -0.25 0.12*
Type of 1St Violent
Arrest
Robbery -0.40 0.04*** 0.24 0.04*** 0.25 0.06***
Assault -0.38 0.03*** 0.05 0.04 0.15 0.06**
CSC 0.00 0.00 0.00
Race
Black 003 0.02 -0.02 0.02 0.01 0.04
Hispanic 006 0.03* -0.05 0.04 0.09 0.05
Other -0.06 0.12 -0.07 0.13 -0.16 0.20
White 0.00 0.00 0.00
Total Prior Arrests 0.00 0.01 0.00 0.01 0.02 0.01
Age at 1St Arrest 0.01 000* 0.00 0.00 0.01 0.00*

 

*p < .05; **p < .01; ***p < .005

58

Binomial test results. However the GLM also exhibited a negative signiﬁcant association
between the number of CSC convictions after the first violent arrest and being Hispanic
as compared to being White, and a positive signiﬁcant association between age at ﬁrst
arrest and the number of CSC convictions after the initiation of the violent career. Such
associations were not demonstrated by the results of the negative binomial tests. The
Negative Binomial tests did not show any association the robbery conviction measure and
any of the independent variable however the GLM showed that this measure was
positively associated (p < 0.005) with initiation of the violent career with a robbery. The
only difference between results of the Negative Binomial and the GLM as regards the
Assault Conviction measure was that the GLM failed to exhibit a positively signiﬁcant
relationship which was captured by the Negative Binomial test results between the
dependent measure and the number of prior non-violent arrests before the ﬁrst violent.
DISCUSSION

This chapter shall analyze both the individual and collective essence and effect of
the observations made in the previous chapter in the context of this study. Beginning with
the structure and attributes of the classifying variable “lSt Violent Arrest” it can be
observed that sex offenses are the least committed within the three violent categories.
However the bivariate analysis has established a degree of uniqueness among sex
offenders. For example sex offenders were observed to have been least likely to be
arrested for a prior non-violent offense and to be incarcerated as juveniles. Furthermore,
race was seen to have a unique affect upon sex offenders as compared to the other two
violent categories, as Whites were more likely to commence their violent careers with a

CSC while Blacks and Hispanics dominated robbery and assault respectively.

59

Interestingly, Blacks who clearly represented the majority in the dataset were least likely
to have initiated their violent careers with a CSC. Initial evidence of generic
specialization within all three violence subcategories can also be noted in table 5, with
the exception of the “Assault Conviction” variable where people who initiated their
violent careers with robbery were more likely to be convicted for assaults later in their
violent careers as compared to people who began with an assault.

The results produced by the three multivariate tests were generally consistent for
both arrest and conviction, with some minor exceptions. In all three multivariate tests for
both arrest and conviction, people who began their criminal careers with an assault or
robbery were signiﬁcantly less likely to be arrested or convicted for committing a CSC
further in their violent careers as compared to the people who initiated with a CSC.
Hence, contrary to the ﬁndings of prior literature (Simon, 1997 and 2000) these results
render evidence of specialization among sex offenders. Regarding the subjects who
commenced their violent careers with a robbery, although they were not signiﬁcantly
likely to get arrested for either robbery or assault, their likelihood of getting convicted for
both was found to be signiﬁcant, hence falling short in reciprocating the degree of
specialization demonstrated by sex offenders. These results are inconsistent with the
Schwaner (2000) study, which speciﬁcally found robbers to be specialists. The subjects
who initiated their violent careers with an assault were found in the three regression
models signiﬁcantly likely to get both arrested and convicted for committing assaults
later in their violent careers. However robbers were found to be even more likely to

render assault convictions.

60

Now that the evidence regarding specialization produced by the present study has
been relayed their compatibility with the three research hypotheses shall be assessed. The
ﬁrst hypothesis, which stated that, “all violent offenders specialize within their violent
subcategories” can be rejected on account of the fact that this study has evidenced no
specialization among robbers and only partial specialization among assaulters. The
second hypothesis that advocated a joint category of violence for sex offenders and
assaulters and claimed a specialization within it has also been disproved by the negative
relationship between people who initiated their violent careers with a CSC and those who
began with an assault as regards CSC arrests and convictions later in their careers. The
third hypothesis that professed, “Specialization existed only among sex offenders” is
probably closest to the truth, barring the fact evidence of partial specialization was found
among assaulters as well.

Albeit the results produced by this study lean towards the veracity of an exclusive
specialization among sex offenders as compared to robbers and assaulters the actual truth
lies not within the black and white line drawn by the research hypothesis but with in the
gray area surrounding that line. As stated above there was signiﬁcant evidence found
regarding existence of specialization among assaulters also. Moreover, the specialization
among sex offenders found in all the three regression models hinges upon the deﬁnition
of specialization taken in the study. If specialization is taken to mean committing only
one kind of offense through out the career (a deﬁnition that was not taken in this study),
in the light of the results obtained sex offenders can hardly be called specialists. Although
not signiﬁcantly, but the people who began their violent careers with a CSC did commit

more assault and robberies as compared to the CSCs committed by the people who

61

commenced with an assault or robbery. Hence, although robbers and assaulters rarely
crossed over to sex offenses later in their violent careers, sex offenders often traversed
into robbery and assault. Further analysis showed that the majority of the people who
initiated their violent careers with a sex offense simply did not recidivate into another
violent offense (at least within the period that they were understudy).

Now that the results generated from this study have established sex offenders’
specialization to a signiﬁcant degree, their distinctive characteristics as compared to the
other violent categories shall be looked at through their interaction with the other
independent variables. Initially it has been stated in this chapter that the results in the
bivariate analysis have shown that sex offenders are least likely to have a prior offense or
prior juvenile incarceration. Further analysis showed that out of the ﬁrst time sex
offenders who had prior non—violent arrests, very little had property arrests and almost
none had any drug arrests. For the initial CSC arrest, Whites dominated the other races
but the regression models have shown that later in their careers Hispanics were
signiﬁcantly more likely to get arrested for committing CSCs as compared to Whites and
Blacks. Contrary to the results obtained from the arrest measures, the relationship
between Hispanics and CSC convictions was observed to be negatively signiﬁcant. It has
also been seen in the bivariate analysis that the mean age of getting arrested for the ﬁrst
CSC was greater that the mean age of getting arrested for the ﬁrst assault or robbery,
hence sex offenders began their violent careers later that the other violent offenders. It
has also been observed that compared to the other violent offenders, majority of the

people who initiated with a sex offense did not recidivate at all later in their careers.

62

Interestingly, after reviewing the abovementloned facts, the profile of a sex
offender that takes shape within one’s mind is almost inexplicable by any of the theories
that laid the foundation of this study. Under the generalist perspective (Gottfredson and
Hirschi, 1990, Mofﬁtt, 1993 and Piquero et. a1. 1999) it was stated that onset age is
inversely related to variety of offenses, however the results of this study have shown that
sex offenders are more likely to commit all the three types of violent offenses while their
mean onset age is greater, as compared to robbers and assaulters. Hence, making the
observed behavior of sex offenders in this study, unexplainable by the “Generalist
Perspective” (Gottfredson and Hirschi, 1990) and the “Mofﬁtt’s Taxonomy” (Mofﬁtt,
1993). Similarly, the “specialization perspective” (Loeber and LeBlanc, 1989, Nagin and
Paternoster, 1994 and Tittle, 1995) states that specialized criminal behavior blooms later
in the criminal career but the results of this study have shown that specialization sets in
right at the time of the initiation of a violent career when it is commenced with a sex
offense. Lastly, the construct of escalation is buttressed upon the assumption that
offenders begin their criminal careers with minor offenses and as the criminal career
matures with time the intensity of the offenses perpetrated also increases. However, the
result of this study have demonstrated that for most of the offenders who initiated their
violent careers with a sex offense, which would rank very high on the severity scale, their
initial sex offence most likely their ﬁrst offense.

One method of looking at the results so produced can be to arrange the three
violent subcategories onto a severity scale. On the severity scale sex offenses shall stand
out as the most heinous of the three, hence are punished more severely. In accordance

with this perspective those offenders who begin their violent careers with the most

63

heinous of crimes are more likely to repeat those heinous crimes along with other lesser
violent crimes. On the other hand those offenders who commenced with a violent crime
of a lesser degree of severity may commit other violent crimes of the same or marginally
greater degree of severity later in their careers, but are less likely to traverse into the
crimes of the severest category. One reason for such inhibition could be the increasing
level of moral or legal reprimand levied to such heinous crimes. Whereas those who
initiate with the severest offense will have already experienced this reprimand and will
either completely cease criminal activity or will continue violent criminal activity at the
next available opportunity without giving any consideration to the reprimand thus
involved. Although this method renders an apposite explanation to the behavior of violent
offenders in the light of the results produced by this study, its theoretical soundness and
empirical practicability is contingent upon future research ﬁndings.

The question still remains that if sex offenders belong to a select class of violent
offenders that specialize within their violent subcategory then what are their identifying
indicators and whether such indicators are best tackled by a preventive or a curative
approach. The static, dynamic and acute factor system developed by clinical
psychologists (Hanson, 1998) could be resourced to here. Exploratory research regarding
the validity of tests like the Psychopathy Checklist (identifying prior sex offenses as an
indicator), Violence Risk Appraisal Guide (which treats violence in general and school
maladjustment as the strongest predictors) and the different personality tests (as have
been discussed in the literature review) could be conducted to isolate and treat such
predictive indicators (Harris, Rice and Quinsey, 1993; Webster, Harris, Rice, Cormier

and Quinsey, 1994; Marshall and Hall, 1995; Rice and Harris, 1997; Quinsey, Harris,

64

Rice and Cormier, 1998 and Harris, Rice, Chaplin and Quinsey, 1999, Palermo and
Farkas, 2001). However, as it has been already pointed out in the literature review section
that such studies need to focus not only on sexual predators but on the general population
of sex offenders in order to render more meaningful results.

Although after regarding the results of this study the proposed research in the
ﬁeld of clinical psychology is not entirely unwarranted, but the fact still remains that this
has been a singular study of its nature with many limitations and its results need
validation by other empirical studies. To begin with the dimensions of the canvas used to
conduct this study i.e. the mid-size Michigan city for arrest and the state of Michigan for
conviction, was too small to command universal empirical value. Albeit, generally
records spanning over the entire state of Michigan would have been considered adequate,
however in this case it has to be borne in mind that these records were of conviction
(which is know to artiﬁcially shrink the effect recidivism) and only of those subjects that
were ﬁrst arrested in the mid-size Michigan city, whilst pertaining to a certain age group.
This shortcoming has also disabled the capacity of this study to capture the effect of
migration of sex offenders to a different county or state consequent to the enforcement of
the sex offender registration and notiﬁcation laws (Freeman-Longo, 2000 and Palermo et.
al., 2001). Some solace may be taken in the fact that if such migration was on a county
level then the conviction records, which spanned over the entire state, might have
captured some part of their subsequent criminal career. A state level migration is not very
likely in the wake of the present circumstances under which all states have enacted this
law. However the varying levels of severity in the enforcement of these laws across states

may come into play. Also, this study did not have any record of the period that each

65

offender spent under incarceration. The longitudinal data used in this study spanned over
ten years and on average captured only four years of the subjects’ violent criminal careers
rendering it less adequate for conducting a specialization study. A review of previous
empirical literature has shown that specialization in sex offenses blooms in the later part
of a sex offender’s criminal career (Hanson, Steffy and Gauthier; 1993; Prentky et. al.
1998 and Simon 1997 and 2000), which the present study clearly failed to capture.
Future studies that aim to test the specialization hypothesis for sex offenders are
recommended to use arrest records encompassing an entire state or at least a major city.
As the arena within which the study will work gets bigger the migration effect is also
likely to diminish albeit removing it entirely would be impossible. The longitudinal
dataset should provide on average at least a ten years follow up period for each subject
after his or her ﬁrst violent offense. As this study has shown that the on average offenders
commenced their violent careers around the age of seventeen, a ten year follow up period
from that point onward would capture their violent career after they age past the mid 20s
point; hence allowing enough time for the latent specialization tendencies earlier in the
violent career of an offender to surface (Prentky, Lee, Knight and Cerce, 1998 and Simon
1997 and 2000). Furthermore, this study aimed to capture the beginning of the violent
careers for subjects and then follow their careers from that point onward, in order to have
a complete record of their violent crime history with in the data set. Hence, only a select
group of violent offenders who had commenced their criminal careers as juveniles and
had at least one violent offense later in their careers were concentrated upon, however as
stated above some studies have surmised that sex offender specialization reaches its peak

much later in an offender’s career (Hanson, Steffy and Gauthier, 1993 and Prentky et. al.

66

 

1998). Hence, a study that targets known sex offenders in their mid twenties and follows
them up for the next ten or twelve years could shed light on the veracity of this statement.
If the results of the present study are conﬁrmed (i.e. that sex offenders specialize) other
studies on specialization among sex offenders might be conducted that bifurcate sex
offenses into rape and child molestation and monitor the isolated effect of each category.
The results of this study have shown that specialization among sex offenders sets in from
the beginning or from the ﬁrst sex offense, which might lead to the conclusion that if this
fact were true one might replace specialization studies with recidivism studies on sex
offenders. This conclusion would however be erroneous, for the present study has also
revealed a propensity among sex offenders to commit other lesser violent offenses and
this tendency can only be further conﬁrmed and explained by looking at the entire violent
career rather than the mere next offense. Lastly, a criterion of specialization in general
should be developed which measures the varying degrees of specialization across
criminal careers instead of establishing the existence or non-existence of it in the absolute
sense.

Finally, the results so generated in this research have established specialization
among sex offenders and their uniqueness as a distinct class of offenders in comparison
to other violent offenders. However, while making a comment upon the policy
implications of this study it would be premature to conclude at this stage that these results
justify the existence of all the federal and state laws that speciﬁcally target sex offenders.
First of all, sex offender recidivism was observed to be the lowest among the three
violent offense categories, which raises issues (barring any moral concerns) as to the

legislative and economical prudence of a law that aims to target such a small segment of

67

the offender population. There is of course an existing possibility that the observed levels
of low recidivism among sex offenders are the product of the Sex Offenders Registration
and Notiﬁcation laws. The abovementioned laws speciﬁcally targeted sex offender
recidivism and came into force within the state of Michigan in 1994, which is the mid
point of the longitudinal data set used in this study. Assuming that such laws had a
preventive effect on sex offender recidivism (a fact that has not yet been shown by any
empirical study), such an effect within the context of the present study might have
molliﬁed effect of sex offender specialization existing prior to the enactment of the laws.
Albeit, it has been noted that chronologically all the cases of recidivating into a sex
offense by a previously violent offender, within the dataset used, have occurred after
1994. This trend could either be consequent to the enactment of the aforementioned laws
or the fact that the cohort of violent offenders selected out of the entire data set belonged
to an age group that was more likely to recidivate in the later part of the dataset.
Secondly, as stated previously the present study has been exposed to numerous
limitations arising out of the extent and comprehensiveness of the dataset, marring the
empirical value of the results so produced to a certain extent. Thirdly, the scope of this
study only focuses upon a few aspects out of many that capture the research hypothesis at
hand in its entirety. However, as an exploratory effort this study has partially achieved its
objective by identifying and studying a lacuna present between the available scientiﬁc
evidence regarding sex offender specialization and the policy making which assumes the
existence of such specialization and has pin pointed the directions for further exploratory

and explanatory empirical work upon the subject.

68

 

BIBLIOGRAPHY

Abel, G., Becker, J ., Murphy, W., & Flanagan, B. (1975). Identifying dangerous
child molesters. In Stuart, R. (ed.), Violent behavior: Social leamingapproaches. New
York: Brunner Maze].

Abel, G., Mittelman, M., Becker, J .B., Rathner, J., and Rouleau, J .L. (1988).
Predicting child molesters’ response to treatment. In 3 RA. Prentky & V.L. Quinsy
(eds), Human sexual aggression: Current perspectives (pp. 223 - 234). New York:
Academy of Science.

American Psychiatric Association (APA) (1994). @gnosticimd Statistical
Manual-Rev. (DSM-IVL Washington, DC: Author.

Bandura, A. (1977). Social learning theogy. Englewood Cliffs, NJ: Prentice- Hall.

Becker, H. (1957). Race and career problems of the Chicago public school
teachers (1980 edition). New York: Arno Press.

Benson, M.L., & Moore, E. (1992). Are white-collar and common offenders the
same? An empirical and theoretical critique of the recently proposed general theory of
crime. Journal of Research in Crime and Delinquency, 29(3), 251 — 272.

Berlin, F.S., & Meinecke, CF. (1981). Treatment of sex offenders with anti-
androgenic medication: Conceptualization, review of treatment modalities, and
preliminary ﬁndings. Amemn Journal of Psychiatry. 138. 601 — 607.

Bentham, J. (1998). Introduction to the principles of morals and legislation. A
Williams, F.P. & McShane M.D. (eds), Criminology Theorv: Selected Classic Readings,
(pp. 23 — 39). Cincinnati: Anderson Publishing Company.

Blumstein, A., Cohen, J ., & Visher, GA. (1986). Criminal careers and career
criminals (vols. 1 & 2). Washington, DC: National Academy Press.

Blumstein, A., Cohen, J ., & Farrington DP. (1988). Criminal career research: Its
value for criminology. Criminology, 26, 615 — 425.

Boer, D. P., Wilson, R. J ., Gauthier, C. M., & Hart, S. D. (1997). Assessing risk
for sexual violence: Guidelines for clinical practice. In C. D. Webster & M. A. Jackson
(Eds), Impulsivitv: Theory, assessment and treatment (pp. 326—342). New York:
Guilford.

Booth, A. & Osgood, D.W. (1993). The Inﬂuence of testosterone on deviance in
adulthood: Assessing and explaining the relationship. Criminology, 31(1), 93 - 117.

69

 

Britt, CL. (1996). The measurement of specialization and escalation in the
criminal career: An alternative modeling strategy. Journal of Quantitative Criminology,
L2, 193- 222.

Broadhurst, R.G., & Maller, RA. (1992). The recidivism of sex offenders in
Western Australian prison population. British Journal of Criminology. 32(1). 54 - 80.

Brooks, AD. (1996). Megan’s law: continuality and policy. Criminal Justice
Ethics, 15, 56 — 66.

Brownmiller, S. (1975). The psychology of women: Against our will. New York:
Simon & Schuster.

Bursik, R.J., Jr. (1980). The dynamics of specialization in the juvenile offenses.
Social Forces, 58, 851 — 864.

Card, R.D., & Dibble, A. (1995). Predictive value of the Card/Farral stimuli in
discriminating between gynephilic and pedophilic sexual offenders. Sexual Abuse: A
Journal of Research and Treatment, 7(2). 129 — 141.

Cames, P. (1983). Out of the shadows: Understanding sexual addiction.
Minneapolis: CompCare Publishers.

Clarke, R., & Cornish, D. (1987). Understanding crime displacement: An
application of rational choice theory. Criminology, 25, 933 — 947.

Cleckley, HM. (1941). The ma_sk of sanity: An attempt to clarify some issues
about the soglled psychopathic personali_ty. St. Louis, MO: Mosby.

Cloward, R.A., & Ohlin, L. (1960). Delinquency and Opportunity. New York.
Free Press.

Cullen, F.T., & Agnew, R. (1999). Criminological Theory: Past to Present. Los
Angeles, CA: Roxbury Publishing Company.

Cumming, G. & Buell, M. (1997). Supervision of the sex offender. Brandon, VT:
Safer Society Press.

Currie, E. (1985). Confronting Crim_e: An American Challenge. New York:
Pantheon.

Doren, D.M. (1998). Recidivism base rates, predictions of sex offender
recidivism, and the “Sexual Predator” commitment laws. Behavioral Sciences and the
Law 16, 97 — 114.

 

70

 

Ellis. L. (1989). Theories of rape: Inquiries into the causes of sexual aggr_ession.
New York: Hempisphere.

Erikson, EH. (1963). Childhood and society (2ud ed.). New York: Norton.

Evans, T.D., Cullen, F.T., Burton, V.S., Dunaway, G., & Benson, ML. (1997).
The social consequences of self-control: Testing the General Theory of Crime.
Criminology, 35, 475 — 504.

Freeman-Longo, RE (1982). Sexual learning and experience among adolescent
sexual offenders. International Journal of Offender Therapy and Comparative

Criminology, 26(2), 235 — 241.

Freeman-Longo, RE. (2000). Revisiting Megan’s Law and sex offender
registration: Prevention of problem. Sexual Abuse in America: Epidemic of 21St Centug.
Brandon: Vermont: Safer Society Press.

Freud, S. (1954). Letters to Wilhelm Fliess, Drafts and notes: 1887 - 1902. In
Ponparte, M., Freud, A., & Kris, E. (eds), The origina of psychoanalysis (Mosbacher, E.,
& Strachey, J .). New York: Basic Books.

Freud, S. (1965). Normality and pathology in chilglhood: Assessment of
development. New York: International University Press.

Frisbie, L.V., & Dondis, EH. (1965). Recidivism amorgtreated sex offenders:
Mental health research monoggaph (No. 5). California: Department of Mental Health.

Gottfredson, M., & Hirsschi, T. (1983). Age and Explanation of Crime. American
Joumal of Sociology, 89(3). 565 - 588.

Gottfredson, M., & Hirsschi, T. (1988). Science, public policy, and the career
paradigm. Criminology, 26, 427 — 445.

Gottfredson, M., & Hirsschi, T. (1989). A propensity event theory of crime. A

Laufer, W.S., & Freda, A. (Eds), Advances in Criminology Theory (pp. 57 - 67). New
Brunswick NJ: Transaction Publishers.

Gottfredson, M. R., & Hirschi, T. (1990). A General Theory of Crime. Stanford:
University of Stanford.

Groth, A.N. (1979). Men who rapp. New York: Plenum Press.

Grunfeld, B., & Noreik, K. (1986). Recidivism among sex offenders: A follow up
study of 541 Norwegian sex offenders. International Journal of Law and Psychiatg, 9, 95
- 102.

71

Hanson, K.R., Steffy, R.A., & Gauthier, R. (1993). Long term recidivism of child
molesters. JoumleConsultinﬂnd Clinical Psychology, 61(4), 646 — 652.

Hanson, K.R., Heaher, S., & Steffy, RA. (1995). A comparison of child molesters
and non sexual criminals: Risk predictors and long-term recidivism. Jorml of Research
in Crime and Delinquency. 32(3), 325 — 337.

Hanson, R.K., Scott, H. & Steffy, RA. (1995). A comparison of child molesters
and nonsexual criminals: Risk predictors of long-term recidivism. Journal of Research in
Crimeapd Delinquency.32, 325 — 337.

Hanson, R.K., & Bussiere, MT (1996). Predictor of sexual offense recidivism: A
meta analysis. Ottawa, Ontario: Ministry of Solicitor General of Canada.

 

Hanson, R.K. (1997). The development 021 brief actuarial a risk scale for sexual
offense recidivism. Public Works and Government Services: Department of Solicitor
General of Canada.

 

Hanson, KR. (1998). What do we know about sex offender risk assessment? [The
science of sex offenders: risk assessment, treatment, and prevention]. Psychology, Public
Policy. and Law. 4(1). 50 - 72.

Hare, RD. (1991). Manual for the Hare Psychopathy Checklist-revised. North
Tonowanda, NY: Multi-Health Systems, Inc.

Harris, G.T., Rice, M.E., & Quinsey, V.L. (1993). Violent recidivism of mentally
disordered offenders. The development of a statistical prediction instrument. Criminal
Justice and Behayior. 29, 315 — 335.

Harris, G.T., Rice, M.E., Chaplin, T.C., & Quinsey, V.L. (1999). Dissimulation in
phallometric testing of rapists’ sexual preferences. Archives of Sexgl Behﬂior. 28(3),
338 — 341.

Healy, W., & Bronner, A. (1926). Delinquents and criminals: Their making and
unmaking. New York: Macmillan (reprinted Montclair, NJ: Patterson Smith, 1969).

Hughes, B. (1952). The sociological study of work: An editorial foreword.
American Joumg of Sociology. 57, 5.

Klein, M.W. (1984). Offence specialization and Versatility among Juveniles.
British Journal of Criminology, 24, 185 - 194.

Knopp, RH. (1984). Retraining adult sex offender: Methods and models.
Syracuse, NY: Safer.

72

Kuhn, TS. (1996). The Structure of Scientiﬁc Revolutiona (3rd ed.). Chicago: The
University of Chicago Press.

Lalumiere, M.L., Quinsy, V.L. (1992). The discriminability of rapists from non-
sex offenders using Phallometric Measures: A Meta Analysis. Criminal Justice and
Behavior. 21(1). 150 - 175.

Lane, S. (1994). The sexual abuse cycle. In Ryan, G. & Lane, S. (Eds), Juvenile
Sexual Offending (2nd ed.), (pp. 77 — 121). San Francisco: Jossey-Bass Publishers.

 

Langan, P.A., & Lavin, DJ. (2002). Recidivism of prisoners relea_sed in 1994.
(Publication No. NCJ 193427). Bureau of Justice Statistics: US. Department of Justice.

Lee ET. (1992). ﬁatistﬁl Methods for Survival Data Analysis (2nd ed.). New
York: John Wiley & Sons, Inc.

Loeber, R., & LeBlanc, M. (1989). Male criminaﬁctivity from childﬂod through
youth: Multilevel and developmental perspectives. New York: Springer-Verlag.

 

London, D. (1988). Survival Models and their estimation (2nd ed.). Connecticut:
ACTEX Publications.

Marnie, R.E., Quinsey, V.L., & Harris, GT. (1991). Sexual recidivism among
child molesters released from a maximum security psychiatric institute. Journal of
Consulting and Clinical Psychology. 59(3). 381 — 386.

Marques, J .K., Day, D.M., Nelson, C., & West, MA. (1994). Effects of
cognitive-behavioral treatment on sex offender recidivism: Preliminary results of a
longitudinal study. Criminal Justice and Behavior. 21, 28 — 54.

Marshall, W.L. & Hall, G.C.N. (1995). The Value of MMPI in deciding forensic
issues in accused sexual offenders. Sexual Abuse: A Joumal_of Research and Treatment
1, 205 — 219.

 

Matson, S., & Lieb, R. (1997). Megan’s Law: A review of state and federal
legislation. The evergreen State College, Washington State Institute for Public Policy.
Available: http:/l www.wa.gov/wsipp.

Mofﬁt, TE. (1993). Adolescence limited and life-course-persistent antisocial
behavior: A developmental taxonomy. Psychological review, 100, 674 — 701.

Nagayama, H., & Gordon, C. (1995). Sexual offender recidivism revisited: A
meta—analysis of recent treatment studies. Journal of Consulting and Clinical Psychology,
6315 ), 802 — 809.

73

Nagin, D.S., & Paternoster, R. (1994). Personal capital and social control: The
deterrence implications of a theory of individual differences in criminal offending.
Criminology, 32, 581 - 606.

Park, R.E., Burgess, E., & McKenzie, R. (1925). The city. Chicago: University of
Chicago Press.

Palermo, G.B., & Frakas MA. (2001). The dilemma of sexpal offendara.
Springﬁeld, Illinois: Charles C. Thomas.

Paterson, R.A., & Braiker, H., & Polich, S. (1980). Doing crime: A survey of
California inmates. Santa Monica, CA: Rand.

Patterson. GR. (1980). Mothers, the unacknowledged victims. Chicago:
University of Chicago Press for the Society for Research in Child Development.

Pavlov, L. (1927). Conditioned Reflexes. Oxford: Clarendon Press.

Piaget, J. (1928 1. Judgment and reasoning in the child. London: Routledge &
Kegan Paul.

Piquero, A., Paternoster, R., Mazerolle, P., Brame, R., & Dean, CW. (1999).
Onset age and offense specialization. Journal of Research in Crime and Delinquency.
36(3), 275 — 299.

Piquero, A. (2000). Frequency, specialization, and violence in offending careers.
Journal of research in crime and delinquency. 37 L4), 392 — 418.

Piquero, A., & Buka, S.L. (2002). Linking juvenile and adult patterns of criminal
activity in the Providence cohort of the National Collaborative Perinatal Project. Journal
of Criminal Justice 30, 259 — 272.

 

Prentky, R.A., Lee, A.F.S., Knight R.A., & Cerce, D. (1998). Sex offender
recidivism. Law and Human Behavior.

Prentky, R.A. & Burgess, A.W. (2000). Forensic Management of Sexua_l
Offenders. New York: Kluwer Academic / Plenum Publishers.

Prentky, R.A., Harris, 3., Frizzell, K., & Righthand, S. (2000). Development and
validation of an actuarial instrument for assessing risk among juvenile sex offenders.

Sexual Abuse: A Journal of Research and Treatment. 12. 71 — 93.

Radzinowicz, L.D. (1957). Sexual Offenses: A report of the Cambridge
Department of Science. London: Macmillan.

74

Rice, M. E., & Harris, G. T. (1995). Violent recidivism: Assessing predictive
ability. Joungl of Consultingand Clinical Psychology. 63, 737 - 748.

Rice, M. E., & Harris, G. T. (1997). Cross-validation and extension of the
Violence Risk Appraisal Guide for child molesters and rapists. Law and Human
Behayior. 21, 231—241.

Quinsy, V.L., Harris, G.T., Rice, M.E., & Cormier, GA. (1998). Violent
offenders. Appraising and managing risk. Washington, DC: American Psychological
Association.

Rice, M.E., Harris, G.T., & Quinsey, V.L. (1990). A follow-up of rapists assessed
in a maximum-security psychiatric facility. Journal of Interpersonal Violence. 5, 435 —
448.

Russell, D.E.H. (1982). The prevalence and incidence of forcible rape and
attempted rape of females. Victimology. 7. 81 — 93.

Ryan, G. (1997). Theories of Etiology. In Ryan, G. & Lane, S. (Eds), Juvenile
Sexual Offending (2‘Id ed.), (pp. 19 — 35). San Francisco: Jossey-Bass Publishers.

Ryan, G. (1997). Phenomenology: A developmental—contextual view. In Ryan, G.
& Lane, S. (Eds), Juvenile Sexual Offendiag (2“d ed.), (pp. 122 - 135). San Francisco:
Jossey-Bass Publishers.

Sex Offenders Registration Act, 295, M.S.A. 4.475(1) (1994).

Schlank, A.M. (1995). The utility of the MMPI and MSI for identifying sexual
offender typology. Sexual Abuse: A Journal of ResearcL and Treatment. 7(3). 185 - 194.

Schrag, C. (1944). Social types in a prison community. Seattle, University of
Washington, M.A. thesis.

Schwaner, S.L. (2000). “Stick ‘em up, buddy.” Robbery, lifestyle, and
specialization within a cohort of parolees. Journal of Criminal Justice 28 371 - 384.

 

Serin, R.C., Barbaree, H.E., Seto, M., Malcolm, B., & Peacock, E. (1997). A
Model for ﬁlinically-informed risk assessment risk assessment strategy for sex
offenders (Research Report No. 56). Ottawa: Correctional Service of Canada.

Shaw, C. (1931). Natural history of a delinquent career. Chicago: University of
Chicago Press.

 

Simon, L.M.J. (1997). The myth of sex offender specialization: An empirical
analysis. New England Journal on Criminal and Civil Conﬁnement. 23, 387- 411.

75

Simon, L.M.J. (2000). An examination of the assumption of specialization, mental
disorder, and dangerousness in sex offenders. Behavioral Sciences anﬁd the Law. 18. 275
— 308.

Singleton, R.A. Jr. & Straits, BC. (1999). Approaches to Social Resear_c_h (3rd
ed.). New York: Oxford University Press.

Soothhill, K.L. & Gibbens, T.C.N. (I978). Recidivism of sexual offenders: a re-
appraisal. British Journal of CriminologY. 18, 267 — 276.

Sturgeon, V.H. & Taylor, J. (1980). Report of a ﬁve-year follow-up study of
mentally disordered sex offenders released from Atascadero State Hospital in 1973.

Criminal Justice Journal, 4131), 31 — 63.

Sutherland, E. (1937). The professio_nal thief. Chicago: University of Chicago
Press. -

Tittle, C. (1995). Control Balance: Toms a General Theory of Deviance.
Boulder, CO: Westview.

Tracy, P.E., & Kempf-Leonard, K. (1996). Continuity and discontinuity in
criminal careers. New York and London: Plenum Press.

Van der Kolk, B. (1986). Psychological Trauma. Washington, DC: American
Psychiatric Press.

Vold, G.B., Bernard T.J., & Snipes B]. (2002). Theoretical Criminology (5th
ed.). New York: Oxford University Press.

Ward, T., Hudson, 5., Marshall, W., & Seigert, R. (1995). Attachment style and
intimacy deﬁcits in sexual offenders. Sexual Abuse: A Journal of Research aad
Treatment, 7(4).

Webster, C.D., Harris, G.T., Rice, M., Cormier, C., & Quinsey, V.L. (1994). m
violence prediction scheme: Assessing dangerousness in high risk men. Toronto:
University of Toronto Center of Criminology.

Wolfgang, M.E., Figlio, R.M., & Sellin, T. (1972). Delinquency in a birth cohort.
Chicago: University of Chicago Press.

Yochelson, S., & Samenow, S. (1976). The criminal mrsonality (Vol. 1).
Northvale, NJ :Aronson.

76

 

S

IIIIIIIIIIIIIIIIIIIIIIIIIIIIII