UBRARY
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

 

 

5:3;P8105' 04

" Lblvg

 

{ECI’IB‘Jl 5233M

 

 

 

 

 

 

 

 

 

 

 

 

6/01 c:/CIRC/DaIeDue.p65-p. 15

THE EFFICACY OF COMMUNITY-BASED INTERVENTION FOR SERIOUS
JUVENILE OFFENDERS

By

Sean Patrick Varano

A DISSERTATION

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

DOCTOR OF PHILOSOPHY
School of Criminal Justice

2002

ABSTRACT

By

Sean Patrick Varano

The control of j uvenile delinquency emerged as an important social concern
around the late nineteenth century, and since then, substantial levels of time and money
have been expended in an attempt to “fix” delinquency by turning wayward youth into
law abiding members of society. Social development theory is put forth as the most
appropriate theoretical fiamework for both understanding the causes of delinquent
behavior and for conceptualizing how best to structure intervention efforts for youthful
offenders. This research compares the effectiveness of community-based intervention
strategies in reducing delinquency compared to traditional probation for adjudicated
youthful offenders.

The focus of this research was to determine not only if certain types of
intervention models are more effective in reducing delinquency, but also to explain why
some treatment programs are more effective. More specifically, it is argued reductions in
recidivism will be contingent on the extent of initial risk for delinquency (e. g., at court
adjudication), the extent of risk reduction attributable to treatment efforts, and the amount
and type of services received while in treatment. “Delinquency” was operationalized
both as “official recidivism” (any new court referral 18 months after adjudication) and
self-reported delinquency (measured 18 months after adjudication).

Findings were contingent on which dependent variable was used. In terms of

official recidivism, findings generally failed to support the arguments that risk, change in

risk, or level of treatment services received matter in reducing delinquency. The
directions of the relationships suggest the opposite is the case. When controlling for
these three core set of variables, little matters in terms of reducing involvement in
officially recorded crime expect for treatment modality. Specifically, those exposed to
the community-based treatment were most likely to recidivate. Initial risk, especially
negative school behavior, was predictive of later levels of self-reported delinquency.
There was additional evidence that reductions in risk over the study period, especially for
drug use, were related to significant reductions in self-reported delinquency. Treatment
modality (community-based or non-secure residential placement) had no effect on levels
of self-reported delinquency. Finally, there was evidence that treatment services actually
increased instead of reducing official delinquency. Implications for social development

theory and policy are discussed.

COPyright by
Sean Patrick Varano
2002

This work is dedicated to my wife, Tracy A. (O’Connell) Varano, whose unending love
and encouragement were crucial to the completion of this work. Words cannot begin to
express how grateful I am for your support.

ACKNOWLEDGEMENTS

It would take countless pages of space to thank all of the people whose support
and encouragement were important to the completion of this work. However, I would
like to acknowledge and thank a few who have been particularly instrumental. First, I am
deeply grateful to my dissertation committee, Dr. Tim Bynum, Dr. Christina DeJong, Dr.
Bill Davidson, Dr. Ed McGarrell, and the late Dr. Chris Vanderpool, for their countless
hours of work during this process. Their guidance and encouragement kept me on tract
and focused, and I am very proud to consider them mentors, colleagues, and friends.

I would also like to acknowledge a few of the tremendous friendships I have made
throughout the past years at Michigan State University. Beth Huebner and John
McCluskey have proven to be two of the best colleagues a person could have. I am proud
to have been one of the founding members of Team Bynum. I am also very fortunate to
consider Justin Patchin, Amanda Burgess-Proctor, and Sarneer Hinduja as co-workers
and friends. Karen Sherman has proven to be one of the best friends, neighbors, and
office mates. My dissertation greatly benefited not only from her editorial skills, but her
ability to listen to my banter without going mad. Also counted among those that have
made my experiences fun as well as educational are Michelle Bynum, Cynthia Perez
McCluskey, Chris Maxwell, Sheila Maxwell, Dennis Payne, Brandon Kooi, and Cedrick
Heraux. I have been fortunate to follow in the footsteps of a several predecessors who
were instrumental in demonstrating to me how to complete a Ph.D. program while

maintaining some (small) sense of sanity - Joe “The Hula Dancer” Schafer, Meghan

vi

“Let’s Get Georgio’s” Stroshine, and Jeff “I Don’t Have The Resources” Cancino. You
have been missed and are still responsible for some of my fondest memories here.

I would like to thank my family for being living examples of love and hard work.
The strength and dedication of my mother, Marie J. Varano, is second to none. I can not
begin to express how crucial your strength has been to all of our lives. I know my father,
the late Theodore F. Varano, would also have been very proud of this accomplishment. It
was this thought that often kept me going when it seemed like the end would never be
realized. To the rest of my family, including my grandmother, siblings, nieces, nephews,
and in—laws - - The example you live each and everyday have again demonstrated to me
what it means to strive for success.

Finally, I would like to acknowledge and thank my wife, Tracy (O’Connell)
Varano. It is difficult to describe in a concise way how instrumental you have been to all
of my professional and personal successes. A few examples of your support have
included hanging outwith me at the library, conversations over dinner in Baker Hall, and
your ever present sense of understanding when I worked on nights and weekends. My
degree is as much an indication of your love and support as it is an indication of my hard
work.

These acknowledgements barely touch the surface of identifying those who have
played an important role in the completion of my Ph.D. degree and my life. Thank you

again and I hope to make you all proud.

vii

TABLE OF CONTENTS

 

 

 

 

 

 

LIST OF TABLES - - - x
LIST OF FIGURES - - - xii
INTRODUCTION 1
Crises and Innovation in the Juvenile Justice System .................................................... 4
Early Responses to Delinquency .................................................................................... 7
Delinquency Prevention/Intervention and Scientific Inquiry ......................................... 9
Summary ....................................................................................................................... 19
RISK FACTORS FOR DELINQUENCY AND DEVELOPMENTAL
CRIMINOLOGY _ -_ - - - _ - 23
Risk Factors For Delinquency ...................................................................................... 24
Individual Risk Factors ............................................................................................. 24
Social Risk Factors ................................................................................................... 36
Ecological Risk Factors ............................................................................................ 46
The Developmental Perspective ................................................................................... 49
The Need for Theory-Based Interventions ................................................................... 63
Current Research ........................................................................................................... 66
Summary ....................................................................................................................... 75
METHODOLOGY - - ..... - 76
The Data ........................................................................................................................ 78
Sample ........................................................................................................................... 80
Hypotheses .................................................................................................................... 82
Specification Of Variables Of Interest .......................................................................... 83
Dependent Variables ................................................................................................. 83
Independent Variables .............................................................................................. 85
Expected Relationships ............................................................................................. 98
Scaling Procedures ...................................................................................................... 100
Analysis Plan .............................................................................................................. 101
Comparing Measures of Central Tendency ............................................................ 102
Logistic Regression ................................................................................................. 104
Within Subjects Change .......................................................................................... 105
CHAPTER FOUR: FINDINGS.-- — - - .................. -_ 109
Sample Characteristics ................................................................................................ 110
Comparisons of Placement Intake Data ...................................................................... 116
Initial Recidivism Measures ....................................................................................... 124
Comparison of T1 Delinquency Risk Factors ............................................................. 135
Reductions in Risk ...................................................................................................... 139
Programming Services ................................................................................................ 142
Risk Factor Changes and Programming Nexus .......................................................... 144
Final Recidivism Measures ......................................................................................... 147
Self-Reported Delinquency ......................................................................................... 160
Summary ..................................................................................................................... 166

viii

Chapter 5: Conclusion -

 

 

 

 

 

-168

Summary and Discussions .......................................................................................... 172
Impact of Delinquency Risk Factors ....................................................................... 173
Impact of Change in Delinquency Risk Factors ..................................................... 180
Impact of Program Efforts ...................................................................................... 184

Limitations of Current Research and Directions for Future Research ........................ 187

Policy Implications ..................................................................................................... 192

APPENDIX A 197
APPENDIX B 200
APPENDIX C _ 209
APPENDIX D 225
BIBLIOGRAPHY ......... 229

ix

LIST OF TABLES

Table 1. Scale Reliability Coefficients for Dependent Variable .............................. 85

Table 2. Scale Reliability Coefficients for Independent Placement Intake Variables. . ...88

Table 3. Scale Reliability Coefficients for Independent Variables ........................... 95
Table 4. Scale Reliability Coefficients for Neighborhood-Level Independent
Variables ......................................................................................... .1' . .97
Table 5. Exploratory Factor Analysis for Social Disorganization ........................... 98
Table 6. Exploratory Factor Analysis for Socio-economic status ........................... 99
Table 7. Descriptive Information by Study Sample Groups ................................. 110
Table 8. Mean Intake Age by Group Membership ........................................... 111

Table 9. Age At Intake for Instant Offense and Offense Type ............................. 113

Table 10. Age At Intake for Instant Offense and Offense Type by Sample
Group ................................................................................................ 114
Table 11. Comparison of Neighborhood Characteristics .................................... 115
Table 12a. Descriptive Characteristics of Placement Intake Variables .................... 118
Table 12b. Descriptive Characteristics of Placement Intake Variables .................... 119
Table 13a. Correlation Matrix: Placement Variables ........................................ 122

Table 13b. Correlation Matrix: Placement Variables ........................................ 123

Table 14. Recidivism for Study Groups — Binary Coding Scheme ........................ 124
Table 15. Recidivism for Study Groups ....................................................... 125

Table 16. Recidivism by Crime Type .......................................................... 128

Table 17. Multivariate Logistic Regression Model. Official Recidivism as Dependent
Variable ............................................................................................. 129

Table 18. Multivariate Logistic Regression Model of Official Recidivism Controlling
For Experimental Group Membership. Official Recidivism as Dependent Variable. . ..131

Table 19. Multivariate Logistic Regression Model Controlling For Experimental Group

Membership, Days in Detention Included ...................................................... 133
Table 20. Descriptive Statistics for T1 Risk Factors .......................................... 137
Table 21. Mean Change in Delinquency Risk Factors ....................................... 140
Table 22. Comparison of Programming Totals ............................................... 143
Table 23a. Bivariate Correlations Among Risk Factor Change Scores and
Programming ....................................................................................... 145
Table 23b. Bivariate Correlations Between Risk Factor Change Scores and
Programming ....................................................................................... 146
Table 24. Multivariate Logistic Regression: Background Characteristics ................ 150
Table 25a. Multivariate Logistic Regression: Background Characteristics with T1
Delinquency Risk Factors ........................................................................ 152
Table 25b. Multivariate Logistic Regression: Background Characteristics with T1
Delinquency Risk Factors ........................................................................ 153
Table 26a. Multivariate Logistic Regression: Index of Change Scores .................... 155
Table 26b. Multivariate Logistic Regression: Drug Use Change Score ................... 156
Table 27. Multivariate Logistic Regression: Background Characteristics with Measure of
Total Services ....................................................................................... 158
Table 28. T3 Self-Reported Delinquency by Group Membership .......................... 162
Table 29. OLS Regression. Dependent variable T3 self-reported delinquency ......... 163

xi

LIST OF FIGURES

Figure 1. Reciprocal model of delinquency ..................................................... 60

Figure 2. Conceptual Model ....................................................................... 77

xii

CHAPTER ONE:
INTRODUCTION

It is widely agreed that juvenile delinquency has increased considerably in the
United States, most notably since the 19603. This conclusion is evidenced from dramatic
increases in juvenile arrests for serious and violent crime since the early twentieth
century (Cook & Laub, 1998) and from more recent problems associated with the rash of
school shootings and other incidences of school-related violence (Kaufman et al., 2000).
Problems associated with juvenile delinquency have commanded extraordinary levels of
public attention, criminal justice resources, and theoretical inquiry for at least the past
one hundred years. From the advent of the “Progressive Reformers” of the early
twentieth century to the present vast efforts have been harnessed to devise effective
measures to ameliorate the conditions that produce delinquency.

The recognition of delinquency as an important social priority has co-occurred
with movements to “fix” the problem by transforming delinquents into healthy, law-
abiding members of society. A variety of different treatment approaches have been
implemented that have vacillated between micro-level approaches focusing on
individual-level defects, such as mental illness, to macro-level strategies that have
attempted large scale social change (Davidson, Redner, Amdur, & Mitchell, 1990).
Delinquency treatment strategies in the United States during the past 100 years are
directly related to the prevailing knowledge of the causes of delinquency. The
understanding that individuals are malleable and subject to change combined with
scientific evidence on the causes and correlates of delinquency have influenced local,

state, and national juvenile delinquency policy initiatives that were intended to curb

delinquency by treating the root causes of the behavior. The prevailing notions of
causality have varied over time resulting in various and changing types of interventions.

The purpose of this research is to analyze the effectiveness of a contemporary
treatment approach in the juvenile justice system that has been identified‘as a promising
approach to intervening in delinquent behavior, the community-based day treatment
model. In doing so, several arguments are made. First, since public policy is most often
a reactionary response to apparent crises, the argument is made that much of what has
historically driven “innovation” in the juvenile justice system is the idea that juvenile
delinquency is at unprecedented levels. It is also the case that responses to delinquency
at a given point in time are driven by contemporary understandings of human behavior.
The movement to “treat” juveniles is reasonable only after children are understood to be
malleable and subject to change, a notion that has not always existed. Finally, this
section details the intimate relationship between the evolution of treatment modalities and
cumulative scientific evidence as to the causes and correlates of delinquent activity.

The analysis in the present study do not focus on the etiology of delinquent
behavior but instead on strategies for intervening in delinquent behavior once it has
reached a level sufficiently serious to warrant official intervention by the juvenile justice
system. However, in order to structure appropriate intervention strategies that recognize
the type of intervention services that are most likely to produce the intended outcomes, it
is necessary to identify an appropriate conceptual models that recognize the origins and
complexity of delinquent behavior. Intervention strategies would only be expected to be
effective if programming is directed at the etiology of the problem behavior. This study

proposes that the causes of delinquency are best conceptualized by interactional theories

such as social development theory, a theoretical framework that integrates individual,
social, and ecological causes. Although some have concluded that treatment has limited
efficacy (Lipton, Martinson, & Wilks, 1975) and that most criminal offenders are
generally beyond rehabilitation (Bennett, DiIulio, & Walters, 1996), there is a growing
body of evidence pinpointing effective intervention strategies. Research suggests
appropriate intervention strategies must be developmentally specific (Loeber, 1996),
directed toward a limited population, involve sufficient intensity (Lipsey & Wilson,
1998), and conceptualize the underlying causes of delinquency behavior (Catalano &
Hawkins, 1996; Lipsey & Wilson, 1998; Loeber, 1990, 1996; Thomberry, 1997b).

The current study will analyze the direct and indirect relationships between the
“quantity” of program treatment, delinquency risk factors the treatment programs are
intended to affect, and outcome measures such as self-reported and official delinquency.
In addition, the current study extends previous research by questioning the extent to
which comprehensive and intensive community-based intervention programs actually
provide “intensive” efforts. The current research is important because the design and
implementation features of programs are related to their efficacy (Lipsey & Wilson,
1998). Further examination of these issues may help to clarify the contradictory frndings
in the literature on the effectiveness of intervention programs. This study will further the
discussion about effectiveness of interventions by disentangling the extent to which
program outcomes are related to issues with program design and implementation.
Although there is growing evidence surrounding the risk factors associated with
delinquency, there is a need for research regarding effective intervention strategies that

can affect these factors. Disentangling characteristics of treatment programs that may be

related to outcomes has the potential for creating a better understanding of program
performance and has implications for the design of more effective interventions.
Crises and Innovation in the Juvenile Justice System

Bernard (1992: 4) argued the juvenile justice system operates in cycles driven by
three coexisting perceptions: “that juvenile crime is at an exceptionally high level, that
present juvenile justice policies make the problem worse, and that changing those
policies will reduce juvenile crime.” The perception that juvenile delinquency is at
dangerously high levels functions historically as a crisis prompting change or innovation
in the system and moves the “response” pendulum between one of two extremes: harsh
punishments or no response at all. Although harsh punishments make intuitive sense
from the perspective of punishing undesirable behavior, reasons for a lack of response are
less apparent. The option to do nothing is a result of the lack of effective and reasonable
responses, especially for minor offenders.l In between these two extremes is the option
of providing treatment, an option considered more lenient than that of harsh punishment.
Delinquency a_s_afSo¢al Problem

The social problem of delinquency is a phenomenon that can be traced to the early
nineteenth century. Prior to that period, the control of children was generally considered
a duty of individual families and not a societal responsibility (Bernard, 1992; Mennel,
1973). The movement of delinquency into the arena of public policy can be traced to the
breakdown of traditional control mechanisms but also to important changes in economic
relationships that occurred during the late nineteenth century. The explosion of the

industrial revolution resulted in dramatic changes to the face of urban centers as large

 

’ Offenders that do not require incarceration are particularly difficult if “lenient treatments” (e. g.,
rehabilitative treatments) are not available because policymakers often believe harsh punishments will
make minor offenders worse by exposing them to criminogenic influences (e.g., prison) (Bernard, 1992).

4

numbers of impoverished immigrants and people previously employed in the agricultural
sector migrated to large cities (Krisberg, 1978). The absolute number of highly
impoverished city dwellers increased as many found it difficult to find quality
employment. The new urban population found it increasingly more difficult to control
their children because many were now subject to negative influences from which they
were once shielded. Visible signs of juvenile delinquency such as roaming streets,
stealing, public fighting, and general disorderliness grew as traditional social control
mechanisms provided by strong family units and steady employment dissolved (Thrasher,
1936).

Juvenile delinquency quickly developed into a crisis as local charities and city
governments were presented with increased numbers of delinquent and at-risk youth.
Official statistics indicated a distinct change in the nature of juvenile arrests starting
around 1950 (Jensen, 1992: 10-11).2 Prior to 1950, the arrest rates for individuals below
18 rested somewhere near 200 per 100,000. However, exponential increases in the
juvenile arrest rate are evident starting around the mid-to—late 19503. Although this
sudden in crease in juvenile arrests was at least partially an artifact of changes in data
collection, it provided the “proof” for many to conclude juvenile crime was increasing at

an unprecedented rate. The FBI, for example, disregarded its own warning and later

 

2 It was difficult to assess the exact extent of delinquency at either the local or national level prior to the
1950s because there was little data collected. The development of national data collection procedures such
as the Federal Bureau of Investigation’s (FBI) Uniform Crime Report (UCR) served as a major
advancement in measuring trends in delinquency. Data collection procedures during the early years of the
UCR program were not standardized, a problem that makes comparison of juvenile arrest rates prior to
1950 difficult. Prior to the early 19508, juvenile arrest statistics were obtained from fingerprint records, the
only records that showed the age of a person arrested. Since it was common practice for many jurisdictions
to not fingerprint arrested juveniles, these early data are argued to be inadequate (Glueck, 1959: 8). The
FBI recognized this problem and initiated a remedy in 1952 to collect the basic demographics of every
person arrested in the United States. The potential impact of the change in data collection procedures was
so dramatic the FBI warned against the comparison of pre and post 1952 trends (Federal Bureau of
Investigations, 1954: 11 1).

interpreted this increase as an apparent juvenile crime wave (Federal Bureau of
Investigations, 1961: 1).

Taken as a whole, the information detailed above indicates there have been
changes in juvenile delinquency over the past 100 years that are reflected partly
explained by the breakdown of traditional social control mechanisms but also partly
explained by changes in data collection procedures. Although there has been a decrease
in juvenile arrests for violent crime since 1994, the general perception is that even today
juvenile crime rates remain high. At nearly every point in time over the past 100 years,
there has been a perception that juvenile crime is on the increase and that the current
generation of youth is somehow much more violent than in the past. It is this perception
that drives policy in the juvenile justice system (Bernard, 1992).

Bernard (1992) argued innovation in the juvenile justice system is induced by the
perception of public officials that juvenile crime is at an unacceptably high level and that
a new strategy is needed for correcting the behavior. Irmovation, however, must exist
within the prevailing ideology of what is an “acceptable” likely alternative to be
explored. Similar to Thomas Kuhn’s (1970) argument about incremental advancements
in scientific thought, “normal” innovation in criminal justice policy is highly contingent
on two important considerations: the strategies currently being used and popular
conceptualization of the underlying causes of the behavior. These two factors direct
responses to the delinquency problem along the continuum from very punitive to very

lenient.

Early Responses to Delinquency

The notion of implementing strategies to “fix” delinquent behavior developed into
meaningfirl options only after children were conceptualized as a legal class
developmentally distinct from adults and malleable (Aries, 1962). The House of Refuge
and Child Savers movements were two early initiatives organized by religious reformers
faced with growing levels of delinquency during the nineteenth century. Early reformers
were members of the privileged class that became involved in delinquency work out of a
sense of charity.3

The House of Refuge (1 825-1 860) was one of the first major movements to direct
resources to delinquent and wayward youth (Krisberg, 1978). The movement was started
by a group of wealthy Christian philanthropists who viewed themselves as “God’s elect”
and responsible for providing moral leadership. The philanthropists attributed
delinquency to inadequate family lives and the pervasiveness of social vices, and thought
reformation could be achieved through the development of regimented work schedules
and daily routines in institutional settings. These types of institutional reformatories
stressing moral righteousness and a strong work ethic were considered favorable for
children to adult jails or workhouses because housing children with adult offenders was
believed to increase exposure to delinquent lifestyles. Moreover, creating youth-specific
institutions was also believed to decrease the likelihood that judges would acquit young
offenders to avoid sending them to adult jails (Mennel, 1973).

Similar to the House of Refugee movement, the Child Savers movement (1850-

1890) identified growing levels of juvenile delinquency as a major social problem. There

 

3 Although most accounts describe the motives of early reformers as “doing Gods work,” those critical of
the movements suggest their motivations were based more in a desire to maintain existing social relation
and to quell rising dissention among the urban poor (Platt, 1991).

7

was, however, a fundamental difference in their approach in that these reformers
identified urban poverty as one of the primary factors underlying delinquency.
Reforrners in this movement were more optimistic about the possibilities of reforming
delinquent youth and believed firmly that institutionalization was not effective
reformation. Instead they believed the family environment was intuitively better to
institutional care and devised methods to either strengthen the families of delinquents or
to align wayward youth with families through a process of “placing out.” The idea of
placing out children was based on the notion of the ideal American family and served as a
way of connecting children in need with what were thought to be “good” families. The
strategy was very similar to the traditional practice in agricultural communities of
“binding out” delinquent children to other families so that they could learn a trade
(Bernard, 1992).

There were several important implications to the House of Refuge and Child
Saving movements. First, both groups clearly articulated the notion that children should
be treated differently than adults. For the House of Refuge movement this entailed
housing them in separate institutions, but the shift marked an important point of
departure. Second, the movements also stand as important transition points in
understanding the “causes” of delinquency. For the first time, delinquency was
beginning to be understood as related to prevailing economic and social conditions. In
addition, the Child Savers movement was particularly important because it was reform
ideas fi'om this period that were responsible for the development of the first juvenile court

system in Chicago, Illinois in 1899.

Delinquency Prevention/Intervention and Scientific Inquiry

Attention to juvenile delinquency as a separate form of deviance resulted from
both a conceptualization of “childhood” as a developmental phase distinct fi'om
adulthood and by a grth in the prevalence of delinquents and other wayward youth in
urban centers. Additionally, principles of scientific inquiry began to be applied to the
study of criminal behavior that included both implicit and explicit treatment implications.
Cesare Lombroso’s book The Criminal Man (1 8 76) used the scientific method to study
the physical attributes of institutionalized criminal offenders concluded that criminals are
biological throwbacks that are less highly evolved.4 Although methodologically crude,
the technique added a welcomed sense of formality to the question of delinquency. The
use of the scientific method eventually gave way to applying many different paradigms of
thought to the study of delinquency including Freud’s psychoanalysis (V old, Bernard, &
Snipes, 1998).

William Healy’s The individual delinquent: A textbook of diagnosis and
prognosis for all concerned in understanding oflenders (1915) moved scientific inquiry
to the forefront of delinquency research and offered social, medical, and psychological
viewpoints on the causes of delinquency. The institutionalization of science in early
treatment models provided an analytical framework and added legitimacy for a group of
“experts” and “professionals” to devise future courses of action. Healy was influential in
the early development of the juvenile court and its related treatment model. Influenced
heavily by Freud’s advances in psychology, Healy believed the key to delinquency lay

deep in the personalities of offenders. Although he and his colleagues were not able to

 

‘ The implication of such an argument is that criminals are inherently different and more dangerous than
law-abiding citizens. As such, a reasonable “solution” would be to ensure they would not be able to
commit additional crimes by incarcerating them for long periods of time.

9

isolate a limited set of causes for delinquency, they concluded that delinquency was a
result of mental dissatisfactions, adolescent mental instabilities and irnpulsions, and
mental peculiarities (Healy, 1915: 34). Healy did, however, also realize that delinquency
was a product of unjust social and economic relationships.

After his work in Chicago, Healy moved to Boston and became an important
figure in the spread of child treatment centers throughout the United States. Krisberg and
Austin (1978) describe Healy as a “proselytizer” for the institutionalization of over 230
child guidance clinics by 1931. The child guidance clinics have been described as a
“branch of mental hygiene” (Glueck & Glueck, 1934: 34) and were concerned with
childhood personality and conduct disorders. The clinics recognized the complex
interaction of children with their environments and wanted to help needy children gain a
“workable orientation” to their world. The child guidance clinics collected a wide range
of social and psychological information on children referred to clinics and their families.

Clifford Shaw, a sociologist at the University of Chicago, extended the inquiry
into the causes of delinquency by identifying the environmental factors associated with
crime. Influenced by the work of Robert Park, Shaw and his colleague Henry McKay
applied the concept of contextual influences to delinquency. They questioned ‘Vvhy. ..
relatively large numbers of boys from the inner urban areas appear in court with. ..
striking regularity” (Shaw & McKay, 1969: 140). The work of Shaw and McKay was
unique in that it abandoned the grand macro level analyses initiated by Emile Durkheim
in favor of research that focused on the effects of smaller geographical areas such as
neighborhoods. More importantly, their research shifted the focus from the individual to

contextual influences.

10

Shaw and McKay hypothesized that the invasion of businesses and immigrants
resulted in disorganized communities that lack effective social control mechanisms.
Social control mechanisms are important features of communities because they intervene
in the transmission of delinquency through cultural norms. Social disorganization theory
posits that areas characterized by economic deprivation experience higher rates of
population turnover (movement in and out of an area) and population heterogeneity, two
features that reduce the formulation of social control (Bursik Jr., 1988). These findings
led Shaw and McKay to conclude “the basic solution of [juvenile delinquency] and other
problems of urban life lies in a program of the physical rehabilitation of slum areas and
the development of community organization” (Shaw & McKay, 1942: xi). During the
early 1930’s, Shaw instituted the Chicago Area Project (CAP), a program that intended to
create long-term change in the life course of juveniles by ameliorating contextual
symptoms. Based on their understanding of the primary causes of delinquent behavior,
the authors to suggest the form and function of intervention efforts should be directed
toward community level attributes.

Shaw perceived community organization and mobilization as a promising strategy
to increase levels of formal and informal social control. Program staff developed social
programs that mobilized local neighborhoods, provided participants the authority to be
active decision-makers, and decentralized program management responsibilities to local
community members. Chicago’s CAP was based on the assumptions that individuals are
involved in a web of regular relationships, individuals will only participate in programs if
they have a meaningful role, and that there are people in most communities that have the

capacity to organize and run programs if given adequate training (Krisberg, 1978).

ll

Within the first few years, twelve independent and self-governing community committees
were developed with seed money from grants.

CAP projects ranged from improvements to schools and sanitation to traffic
patterns, resources directed to levels of law enforcement, and community recreational
renewal projects (Schlossman & Sedlak, 1983). There were also projects intended to
work directly with delinquent youth including “visitation programs for incarcerated
children, with delinquent gangs, and volunteer assistance in parole and probation”
(Krisberg, 1978: 33). The concept of “detached workers,” agency workers that were
removed from their offices and assigned to local communities to work directly with gang
members, was also initiated under the CAP program (Howell, 1998). Woodson (1981)
argued the residual effects of CAP programs were largely responsible for large decreases
in gang deaths between 1973 and 1976. Ultimately, CAP was a major initiative that was
one the first to move away from institutional treatment approaches and psychological
explanations for delinquent behavior.

During the 19608, scientific theories continued to influence policy developments
in the juvenile justice system. Cloward and Ohlin’s (1960) strain theory, for example,
had a tremendous affect on President Kennedy’s Commission on Juvenile Delinquency
and Youth Crime and President Johnson’s “War on Poverty.” Strain is particularly likely
to exist in high poverty areas. Individuals living in high poverty areas experience a
greater disparity between socially defined goals of success (e. g., wealth and status) along
with few legitimate opportunities to achieve socially defined goals of success. Strain was
hypothesized to be positively related to the development of delinquent behavior.

Individuals experiencing strain do not lack the motivation to conform their behavior but

12

instead “the desire to meet social expectations itself becomes the source of delinquent
behavior” (Cloward & Ohlin, 1960: 44).

Strain theory had a profound impact on public policy. The War on Poverty, for
example, was a large-scale program that developed federal antipoverty measures such as
compensatory job training/schooling and income redistribution policies. The purpose of
the War on Poverty was to give the disadvantaged the income and skills they needed to
function in the free market. Mobilization for Youth was one such effort directed
primarily at minority youth in Manhattan’s lower east side. This program was a
comprehensive effort that provided resources in the areas of educational assistance, job-
training, subsidized work, and vocational-guidance programs for unemployed or out-of-
school youth (Jensen, 1992).

The Juvenile Delinquency and Youth Offenses Acts (1961, 1968) and other
related initiatives5 were devised to increase the legitimate opportunities for youth to
succeed in conventional society. They were particularly important because they were the
first major pieces of federal legislation aimed at controlling and preventing delinquency
and were part of larger initiatives that provided services to youth and families such as
Head Start. Federal attention continued to be given to juvenile justice issues, and in 1968
the Juvenile Justice and Delinquency Prevention Act was enacted with responsibilities for
the prevention and control of juvenile delinquency. In addition to these pieces of
legislation, additional measures were enacted that gave the federal government
increasingly more involvement in fimding efforts directed at curbing juvenile

delinquency.

 

5 Similar programs were implemented by the Department of Health, Education, and Welfare, the Equal
Employment Opportunity Act (1964), and Manpower Development and Training Act (1962).

13

The labeling theory perspective is another important framework that affected
public policy around the same period. Advocates of the labeling theory posit delinquents
are essentially no different fiom nondelinquents and that process of applying the
' delinquent label marginalizes certain individuals from non-labeled youth. Marginalized
youth subsequently associate with delinquent youth, thereby increasingly their own
involvement in delinquency (Eddy & Gribskov, 1998; Liska, 1987). Strain and labeling
theories were popular with both politicians and researchers during the 19603 and 19703.
Those advocating the strain perspective argued increasing opportunities for youth would
reduce the likelihood of delinquency. From a labeling perspective, the affects of the
“delinquent label” would be mitigated by decriminalizing status offenses and by reducing
the frequency and extent to which juveniles were incarcerated.

The federal government has been increasingly involved in the prevention and
control of juvenile delinquency for some period of time. Although early interventions
tended to be initiated and funded by social reformers, by the 19603 the federal
government evolved into the primary power broker directing research and treatment
agendas. The federal govemment’s power to direct research and treatment agendas
comes from the infusion of large amounts of research dollars. In 1974, for example,
federal appropriations for the Office of Juvenile Justice and Delinquency Prevention were
$75 million, a figure that was to increase to $125 million in 1975 and $150 million in
1976 (Olson-Raymer, 1984: 37). Much of the funding was disseminated a3 grants for
research and treatment initiatives. In general, juvenile justice policy has remained in
concert with popular theories of delinquency. However, the federal govemment’s

involvement in delinquency prevention was a piecemeal approach to legislation and

14

lacked clarity about big picture objectives, goals, and responsibilities of the various
agencies. Olson-Raymer (1984: 33) concluded:

“such policies were built upon traditional fragmented philosophical and
methodological foundations — little agreement about children and youth’s
needs; no clear differentiations between delinquent, neglected, abuse, or
exploited youth; and no consensual body of professional knowledge
pointing to delinquent causation factors or efficient treatment methods.”
One of the most significant advancements in the treatment of juvenile

delinquency during the early to mid-19003 was the impact of research on the
understanding of the causes of delinquency. The direction of prevention and
intervention efforts was tightly connected to prevailing ideology around the
causes of delinquency. Ultimately, there was an important shift from prevention
efforts that focused on the individual offender during the early 19003 to holistic
strategies that sought to prevent delinquency by creating opportunity.

Yet increases in federal funding for research and advances in
criminological theory did not “solve” the problem of crime. Instead, the United
States experienced one of the largest increases in crime rates during the 19603 and
19703. Macro efforts failed to create the desired effects of reducing delinquency
and the “cycle” of juvenile justice policy (Bernard, 1992) refocused attention to
the traits of individual offenders. This shift was influenced by the
groundbreaking work by Marvin Wolfgang and his colleagues who suggested that
important characteristics such as age of delinquency onset and extent of

specialization demarcate criminal offenders. More importantly the authors’

finding that crime is disproportionately concentrated among a small population of

15

offenders created a fervor around the notion of identifying serious and violent
offenders early in the their careers by the 19803.
Birth Cohort Studies: Characteristics of Criminal Careers

Wolfgang, Figlio, and Sellin’s (1972) Delinquency in a Birth Cohort is an
important contribution to delinquency research because it details the
developmental patterns of delinquent careers. In many ways an extension of the
work of Eleanor and Sheldon Glueck, Delinquency in a Birth Cohort detailed
delinquent careers of a cohort of all males born in the City of Philadelphia in 1945
and traced their official contacts with police during their juvenile court age from 7
to 18. The cohort study indicated juvenile offenders could be differentiated based
on the onset of delinquency and the extent of persistence or desistence from
delinquent activity.

Of the nearly 3,500 juveniles born in 1945 with records of official police
contacts, 46 percent were one-time offenders while the remainder had at least one
additional recorded contact after their initial contact. Recidivists could be
differentiated from nondelinquents and one-time delinquents based on the extent
of their transience between various homes and schools, IQ, and socioeconomic
status. Lower socioeconomic status boys, for example, were much more likely to
be delinquents, recidivists, and chronic offenders than those fi'om higher socio-
economic statuses.

Wolfgang et al.’s (1972) data also revealed important findings regarding
the age of delinquency onset. For example, the mean age of delinquency onset

was 14.4 years (Wolfgang et al., 1972: 130-135) and has been consistently

l6

confirmed by other research (Gottfredson & Hirschi, 1990; Steffensmeier, Allan,
Harer, & Streifel, 1989). Age of onset is an important characteristic to consider
because individuals that begin their delinquent careers before age 13 will commit
more offenses through age 17 than those that began later even when controlling
for time at risk (Benda & Toombs, 2002). Wolfgang et al.’s data indicated age at
first contact was positively associated with proportion of offenses classified as
index or serious crimes (murder, rape, robbery, burglary, motor vehicle theft, and
arson) (142). Moreover, the probability of desisting from crime decreased
substantially after each police contact. For example, although slightly less than
half of the delinquents desisted from delinquency after the first arrest, only 35
percent desisted after the second arrest and 25 percent after the fifth arrest
(Wolfgang et al., 1972: 163). Thus, the probability of desistance decreases with
each subsequent arrest.

There are several implications to the findings of Wolfgang and his
colleagues. First, a small proportion of all offenders (6 percent) are responsible
for a large percentage (52 percent) of all crime committed by individuals in the
birth cohort.6 Second, not all juvenile offenders are the same with regard to the
probability that they will recidivate. The data indicate, for example, that a large
percentage (54 percent) of first time juvenile offenders are likely to desist from
future offending with little to no direct intervention by the juvenile justice system.

Also implicit in these findings is that there is a potential to offset substantial

 

6 A different way to assess the prevalence of these chronic offenders is to determine what percentage they
represent not of the entire birth cohort in general, but the birth cohort that experienced at least one arrest.
In this case, the data indicate 18 percent of all individuals arrested are considered to be chronic offenders
(Blumstein, Farrington, & Moitra, 1985).

17

levels of crime in the future if intervention efforts addressed the needs of high-
risk offenders early in their offending careers. Clearly, then, juvenile delinquents
are not a uniform population of offenders that demonstrate constant likelihood of
recidivism. Instead, there are discernible differences between offenders that are
manifest in the prevalence of offending, age of onset, duration of criminal career,
escalation and de-escalation of behavior, and desistance from criminal behavior.
Other cohort studies have demonstrated there is regularity to offending
patterns. Criminal career research in Philadelphia, London, Marion County,
Oregon, and Racine, Wisconsin reveal similar findings. Namely, a high level of
involvement in official delinquency that ranges from 25 percent of the birth
cohort in Philadelphia to 70 percent in Racine, a rapidly increasing probability of
reoffending through the first few arrests, stability in recidivism rates through
approximately the sixth arrest, and increasingly high and stable recidivism rates
for subsequent involvement (Blumstein et al., 1985).7 Although there is
disagreement as to the extent of regularity in criminal careers (see Blumstein,
Cohen, & F arrington, 1988a, 1988b; Gottfredson & Hirschi, 1986) the findings
can be interpreted as providing useful ways of classifying offenders based on
characteristics of their offending patterns. Early cohort studies, however, were
lacking in their ability to explain why the apparent difference between subgroup
of delinquents existed. For example, the research failed to address why some
juveniles are more likely to partake in delinquency earlier than others. Moreover,

the research failed to answer why age of onset should substantively matter in the

 

7 Blumstein, Farrington, and Moitra ( 1985) noted substantial variation in many of these attributes between
sites yet interpreted the findings as being consistent with findings fiom the Philadelphia Cohort Study.

18

trajectories of criminal careers. Detailing the etiology of delinquency, especially
the developmental sequencing of serious and chronic offending, has the ability to
serve as the next step to intervening in the transition from nondelinquents to
delinquents, occasional delinquent to persister, and more importantly, how
desisters are differentiated from persisters.

Summary

The juvenile justice system has undergone continual change throughout
much of the nineteenth and twentieth centuries caused primarily by perceived
increases in delinquency. The perception that a “crisis of delinquency” exists,
however real or imaginary, has served as one of the main factors driving
innovation. In the face of crises, bureaucracies are driven to institute responses
that give the appearance that the necessary steps are being taken to increase order
and reduce problem behavior (Haberrnas, 1988).

“Progressive” reformers of the early 19003 were one of the first reform
movements to devise systematic responses to delinquency (Platt, 1977). Although
early efforts identified institutional reformatories as the most effective mechanism
for “correcting” wayward and delinquent youth, prevention and intervention
efforts were soon moved beyond institutions to community settings. Responses to
delinquency are not historical accidents, but were and continue to be intimately
connected to the prevailing ideology about the causes of delinquent behavior.

The scientific understanding of the causes and correlates of delinquency
have had a tremendous impact on prevention and intervention efforts. In the early

19003, Dr. William Healy identified the primary causes of delinquency as mental

l9

defects. Prevention efforts were then geared toward correcting this behavior
through intensive counseling and often, the removal of children from their
families. Later, research by Shaw and McKay identified ecological correlates of
delinquent behavior, findings that directed prevention efforts toward community-
level changes that gave ftmding to local organizations to create recreational and
economic opportunities.

During the past several decades, major research efforts have been initiated
aimed at detailing a picture of the causes of delinquent behavior. Research
initiatives operate in certain paradigms that are generally accepted as presenting
the most accurate representation of the relationships of interest during a particular
time period. In the 19303 and 19403, the prevailing theory argued delinquency is
a product of ecological factors that increase exposure to criminogenic influences.
According to the Chicago School, delinquency has its roots in detachment from
conventional groups caused by residential and employment segregation caused by
the “natur ” ecology of urban areas. At risk populations in urban areas become
increasingly concentrated in locations that are characterized by physical decay
and criminogenic influences and limited their exposure and attachment to groups
with normative value systems thereby increasing delinquency (Shaw & McKay,
1931, 1942). As would be expected, conceptual and empirical models posited
causal relationships between features of urban ecology such as patterns in
migration, land use, urban decay, and delinquency rates.

Until the 19603 criminological inquiry concentrated primarily on the

relationships between micro influences and delinquency. The work of Merton,

20

and later Cloward and Ohlin, expanded the discussion on the relationships
between macro influences and delinquency by suggesting from unequal access to
the means to achieve socially defined success. In this case, individuals experience
“strain” between socially acceptable means of achieving wealth and success and
instead resort to anti-social methods such as delinquency. Based on this notion,
researchers concentrated analytic methods on the conceptual relationships
between socioeconomic characteristics, the extent of blocked opportunities, and
involvement in crime.

The above are two examples of the relationships between prevailing
paradigms and how they influence the direction of delinquency research.
Throughout much of the twentieth century the paradigms compelling research
have fluctuated between individual, social, and ecological explanations. Implicit
to most of the above theories is also a set of interventions that should diminish the
impact of risk factors if the theoretical fi'amework is sound. For example,
policymakers implemented several national policy initiatves such as Head Start
and other efforts during the 19603 and 19703 that were aimed at increasing access
to legitimate opportunities, policies that were based on Cloward and Ohlin’s
version of strain theory.

While successful in advancing theoretical relationships among elements of
the causes of delinquency, these conceptual fi'ameworks have been overly
restrictive in their ability to make the necessary connections between the
theoretical frameworks. Social development theory has been proposed as a

theoretical framework for integrating many of the existing theories into a unified

21

theory of delinquency. Thomberry and others propose that social development
theory holds the potential for explaining the complex interactions of individuals
and their environments in the production of delinquency. More importantly, the
theory has the potential for conceptualizing the relationships between these
factors and developmental maturation. Social development theory also has the
potential for suggesting appropriate intervention strategies, especially for more
serious juvenile offenders. If delinquency truly is a byproduct of individual,
social, and ecological forces, intervention strategies must recognize these risk
factors in treatment models.

Developmental theory holds a tremendous promise for integrating multiple
risk factors into models that present a comprehensive assessment of
characteristics of delinquency. Interactional theories such as social
developmental theory have influenced but have also been influenced by “birth
cohort” research and “pathways to delinquency” models. The combination of all
of these initiatves were instrumental in the implementation of the US.
Department of J ustice’s Office of Juvenile Justice and Delinquency Prevention

(OJJDP) “Comprehensive Strategies” (Howell, 1995).

22

CHAPTER 2

RISK FACTORS FOR DELINQUENCY AND DEVELOPMENTAL
CRIMINOLOGY

The juvenile justice system has been faced with an unprecedented increase in the
number of delinquents adjudicated for serious and violent crimes (Snyder & Sickrnund,
1999) and a subsequent need to develop and implement effective intervention programs
for these offenders. Since juvenile offenders have a right to rehabilitation (Rotman,
1986) effective intervention should be recognized as a crucial part of the dispositions and
sanctions applied by the juvenile justice system. The first step in structuring an effective
program for juvenile offenders is to identify the existence and interactions of the causes
of delinquency.

This section will identify the risk factors traditionally associated with
delinquency. It will focus on individual, social, and ecological influences that
differentiate delinquents fiom nondelinquents, and also delineate characteristics of
delinquent careers. I will make the argument that although delinquency explanations
under traditional theoretical frameworks of strain theory, social control theory,
differential association theory, and learning theory have been supported by empirical
evidence, they have been overly restrictive in limiting models to a narrow set of causal
variables. Integrative theories such as developmental and interactional models present a
more representative picture of the onset, trajectory, persistence, and diversification of
criminal careers. A better understanding of the dynamic characteristics of delinquency
provides a framework for structuring effective interventions to offset continued

delinquency.

23

Risk Factors For Delinquency

Cumulative evidence from the past several decades of delinquency research have
clearly documented five categories of causes and correlates of delinquency: (1) individual
risk factors such as demographics, attitudes, self-esteem, and substance abuse; (2) family
influences such as family history of substance abuse and criminality; (3) school
influences such as school bonding and low academic achievement; (4) peer influences
such as involvement in delinquency, experimentation with illegal substances, and
attitudes toward delinquency; and (5) neighborhood influences such as crime rates, low
neighborhood attachment and levels of disorganization (Wilson & Howell, 1995). The
concept of readily identifiable predictors of delinquency lends itself to the design of

intervention efforts directed at particular target populations.

Individual Risk Factors

During the past 100 years researchers have identified precursors of delinquent
behavior. Primary among these “risk factors” has are individual level traits such as age,
gender, and ethnicity. In addition, individual risk factors important to the production of
delinquency include self-esteem, alcohol and drug abuse, and attitudes toward
delinquency. While neither sufficient nor necessary correlates of delinquency, research
continues to demonstrate their importance in the prediction of characteristics of

delinquency careers.

Age
Age is one of the most important predictors of onset and levels of involvement in
delinquency. Age-specific arrest rates increase steadily from early adolescence (10-13)

through the early to late teens. This pattern remains relatively high as individuals mature

24

to their mid-twenties and then decreases at a similarly dramatic rate (F arrington, 1985;
Greenberg, 1985; Shavit & Rattner, 1988). The trend is not only apparent for males but
also remains reasonably consistent for female offenders. This “law of nature” (Goring,
1913) is largely invariant across communities, cultures and historical time periods
(Gottfredson & Hirschi, 1990).

There are several age-related features of adolescence that increase the likelihood
of involvement in delinquency. Since the family serves as the most important socializing
agent during early childhood (pre-adolescence), children spend most of their time with
their parents and similar caregivers during this period who are responsible for
childrearing. Parents and other caregivers are vital in establishing core value systems,
attitudes, ambition, and structuring behavior of children (Furstenberg Jr. et al., 1999).

Parental influences during early childhood have been linked to important aspects
of childhood development such as self-control, a factor argued to be largely responsible
for variations in delinquent behavior (Gottfredson & Hirschi, 1990). “In order to teach
the child self-control, someone must (1) monitor the child’s behavior; (2) recognize
deviant behavior when it occurs, and (3) punish such behavior” (Gottfredson & Hirschi,
1990: 97). In this situation, children are generally exposed to a set of attitudes and
behaviors that are, under ideal settings, consistent with mainstream pro-social values.
Families are the central socializing agent to children during much of adolescence and
during these early years children are more likely to model their attitudes and behaviors
(Furstenberg, Cook, Eccles, Elder J r., & Sameroff, 1999). In the case of healthy family
environments, young children would be expected to model pro-social behavior that is

consistent with parental examples.

25

Children begin to spend more time with their peers and less time with their
parents as they make the transition to adolescence. It is during this transition that they
are generally given freedom to spend more time out of the watchful eyes of their parents
or guardians. Their free time is spent with peers and socializing in peer networks and
they are more likely to look to these groups (as opposed to parents/ families) for important
cues of behaviors and attitudes to model (Elliott & Menard, 1996). For the “normal”
child, the influence of peers and negative peer networks is also age-specific and gradually
begins to diminish as children mature through the teenage years and into early adulthood.
Peer networks eventually lose importance in favor of other stage salient issues such as
employment, intimate relationships, and children thereby decreasing their involvement in
delinquency (Gottfredson & Hirschi, 1990).

There are several useful theoretical frameworks for conceptualizing the
relationship between age, peer networks, and delinquency. Social control theory and
social learning theories are two common frameworks for understanding the age-
delinquency relationship. From the perspective of social control theory (Hirschi, 1969),
delinquency is likely to increase for juveniles with poorly developed attachments to
traditional control mechanisms such as parents and other family members. For young
juveniles, control is primarily exercised through interactions with parents and family
members. Children internalize parental value systems that serve as a foundation for
judgments of the acceptability of future behavior. Young children with strong pro-social
attachments to their parents are less likely to get into situations in which delinquency is

possible because they spend most of their free time in their presence, but also because of

26

value systems instilled in children through regular interactions with parents and other
caregivers.

During the maturation process, juveniles tend to stray away from the strict control
of their parents. The extent of their movement beyond acceptable behavioral limits is
influenced both by features of local environments and the extent to which attachment to
parents is maintained, but for some delinquency gains a favorable status. Temporally,
involvement in delinquency precedes extensive association with delinquent peer
networks (Elliott & Menard, 1996; Thomberry, Lizotte, Krohn, F amworth, & J ang,
1994). Delinquents seek to establish such networks as a movement to surround
themselves with individuals with similar value systems. As juveniles with weak family
attachments or delinquent internalized value systems begin to experiment with
delinquency they increase their association with peers who share similar delinquent
values.

In contrast, social learning theories such as differential association assume
delinquency is learned through interactions with other delinquents. Association with
delinquent peers is proposed to precede delinquency. Involvement in delinquency
increases through a “learning process” where individuals learn the techniques of
delinquency as well as the cognitive justifications that diminish the moral culpability for
participating in such activities. Increased contact with peers with attitudes favorable to
delinquency but who are also involved in delinquency accelerates the learning process.
Juveniles are expected to continue their involvement in delinquency as they maintain
contact with other delinquents. Though not exhaustive explanations, control and

differential association theories are useful for understanding the age-delinquency

27

relationship. Issues of temporal ordering aside, there are important transitions that occur
between early childhood, adolescence, early adulthood, and adulthood that alter the
nature of social interactions. Transitions between various stages of life change the nature
of family dynamics, peer relationships, employment, and intimate relationships. It is
these features that are important mediators in the production of delinquency (Gottfredson

& Hirschi, 1990).

Gender

Gender is also one of the strongest predictors of delinquency in criminological
research. Males are more likely to demonstrate an earlier age of onset of delinquency
(Loeber & Dishion, 1983; Wolfgang et al., 1972), involvement in serious and violent
offending (I-Iuizinga & J akob-Chien, 1998; Loeber, F arrington, & Washbush, 1998), and
persistent offending (Blumstein et al., 1985; Wolfgang, Thomberry, & F iglio, 1987). In
self-reported surveys, Hindelang, Hirschi, and Weis (1981: 140) reported that males were
responsible for 3.6 times more thefts ($50 or more) and assaults than females, 3.4 times
more auto thefts, and almost 3 times more robberies. In a sample of adolescents from the
National Youth Survey, Mears, Ploeger, and Warr (1998) reported males to be
significantly more likely to participate in minor delinquency such as cheating on tests and
stealing items under $5 to serious delinquency such as burglary and theft over $50. In
addition, males were significantly more likely to report drunkenness and alcohol and
marijuana use.

The relationship between gender and delinquency is traditionally explained by
differences in opportunities to commit crime. Opportunity theorists do not attempt to

explain the motivation for crime because they assume there will always be people

28

motivated to commit crime. Instead it is suggested that it is more important to explain
the situations and circumstances in which motivated offenders find suitable victims
(McCarthy & Hagan, 2001; Vold et al., 1998). Situational theories such as opportunity
theory propose delinquency is a function of the co-occurrence of a motivated offender, a
potential victim (including vulnerable property), and a suitable location for the event
(Birkbeck & LaFree, 1993). For a crime to occur, these three things must co-occur in the
same time and space.

One of the traditional explanations for the disparities in the male/female rates of
involvement in delinquency is that females simply have fewer opportunities to commit
crime because they are under the watchful eyes of capable guardians (Felson &
Gottfredson, 1984). Assuming most people to be likely motivated offenders, females are
differentiated fiom males primarily in levels of supervision. This assumption serves as
the basis for the argument that women would develop offending patterns consistent to
men as they entered the work force in greater number (N ettler, 1984), a reality which has
yet to be realized. There is evidence, however, to support the argument that juvenile
males are given more freedom by their parents. Data fiom the National Youth Survey
indicate males spend more of their free time with peers during weekday afternoons and
weekday evenings than females (Mears et al., 1998).

McCarthy and Hagan (2001) argued the study of “criminogenic situations” has
been unjustly disregarded during recent years in favor of developmental perspectives.
The “General Theory of Crime” discredits “opportunity" as the discriminating difference
between male and female offending patterns and instead suggests that girls have more

self-control restraining them from crime than boys because parents are more stringent in

29

their recognition and punishment of early non-conforming behavior (Gottfiedson and
Hirschi, 1990). Males are also socialized to be more aggressive than females and tend to
respond to stressfirl situations with extemalized behavior (aggression) whereas females
respond with internalized behavior (depression) (Leadbeater, Blatt, & Quinlan, 1995). It
is more likely that there is an interaction between gender and childrearing practices.
Although males are exposed to delinquent peers (e.g., differential association theory) at a
rate greater than females, females report less delinquency than males when exposed to
similar levels of delinquent friends (Mears et al., 1998). This suggests while “the number
of delinquent peers an adolescent has is the strongest known predictor of delinquent
behavior,” Mears et a1. (1998) speculate “the moral judgments of females are apparently

sufficient to reduce and even eliminate the impact of delinquent peers.”

Race/Ethnicity

At nearly every aspect of the criminal justice system, minority youth and adults
are over-represented among criminal offenders. The black-to-white ratio for officially
recorded delinquency show that minority youth are arrested at a ratio nearly 2 times more
than those of non-minority youth (Wolfgang et al., 1972), have almost twice as many
police contacts (Hirschi, 1969), are 3-6 times more likely to have an arrest record, and
experience a homicide commission rate over 10 times that of non-minorities (Cook &
Laub, 1998). Moreover, African American and Latino youth are also more likely to
experience risk factors associated with delinquency including school dropout (National
Research Council, 2001), a factor positively associated with later delinquency (Thomas

& Bishop, 1984).

30

Racial differences in offending patterns have been primarily explained with a
subcultural perspective. The subculture of violence thesis suggests the deprivation of
African Americans leads to the development of value systems that condones the use of
violence. Criminality is not a rational response to “need” but a manifestation of the
internalization of delinquent norms. For example, Anderson’s (1999) qualitative analysis
of urban “ghetto” life in Philadelphia depicts a community so devastated and plagued by
poverty that street values such as “respect” and “manhood” circumvent many mainstream
values. Cohen (1955) argued that in the face of continued exposure to delinquent norms,
values, and behaviors, people become desensitized to the negative consequences of
delinquent behavior. Comments detailed from an interview reported by Thrasher (1936:
28) offer an interesting account of this mentality:

“We did all kinds of dirty tricks for fun. We’d see a sign, ‘Please keep the

streets clean,’ but we’d tear it down and say, ‘We don’t fee like keeping it

clean.’ That would make us laugh and feel good, to have so many jokes”.

There is also evidence that not all ethnic groups respond the same when faced
with similar stressors. In a test of the applicability of traditional strain theory to Latinos,
Afiican Americans, and Whites, McCluskey (2002) reported an important interaction
between socio-economic status and ethnicity in the production of delinquent behavior.
Regardless of race/ethnicity, individuals who are economically disadvantaged (receiving
public aid) experience blocked opportunity and strain similarly. However, for those not
receiving public aid, both Latinos and African Americans reported higher levels of strain
than their White counterparts. One of McCluskey’s (2002) most important findings is
that Latino males are less likely than African Americans and Whites to respond to strain

with delinquency, a relationship hypothesized to be mediated by extensive family

31

networks and tighter family structures that help Latino youth deal more positively with

economic disadvantage.

Attitudes Toward Delinquency

Some researchers have suggested participation in delinquency is contingent on
attitudes favorable toward delinquency. In his extensive analysis of drug dealing in New
York “barrios,” Bourgois (1995) provided extensive evidence of a subculture of drug
dealing and violence where criminal activity was accepted as a norm of everyday life. In
one account, Bourgois (1995) articulates his inability to reconcile the almost blasé
attitudes of individuals he befiiended during his research as they recounted the story of a
gang rape they participated in with a sense of glee. Similarly, Elijah Anderson (1999)
explains how attitudes favorable toward delinquency have been internalized in some
urban communities to the point where the search for “respect” encourages the use of
violence. Anderson (1999: 91-92) explains:

“For many young men, the operating assumption is that a man, especially

a ‘real’ man, knows what other men know -— the code of the street. And if

one is not a real man, one is diminished as a person. Moreover, the code is

seen as possessing a certain justice, since everyone supposedly has the

opportunity to learn it, and thus can be held responsible for being familiar

with it. If a victim of a mugging, for example, does not know the code

and thus responds ‘wrong,’ the perpetrator may feel justified in killing him

and may not experience or show remorse. He may think, ‘Too bad, but

it’s his fault. He should have known better.’”

Elliot (1994) reported a positive relationship between deviant attitudes at age 11-
17 and self-reported violence during adolescence and adulthood. In a sample of 732
males and females, violent attitudes at age 14 was a significant predictor of self-reported

violence at age 18 (Maguin et al., 1995). Using a slightly different measure such as

“hostility to police” at age 14-16, Farrington (1989) found a significant positive

32

relationship with self-reported violence at age 16-18. In this case, individuals hostile to
police early in life were 2.5 times more likely to self-report violence later.

There are reasons to believe attitudes toward delinquency may have both direct
and indirect effects on delinquency. Through the conceptual framework of differential
association theory, delinquent peer networks serve as one of the primary influences on
delinquency. Delinquent peer networks would precede delinquent behavior and increase
delinquency through an environment of learning where the “tricks of the trade” are taught
through interaction with individuals with more expertise and skill. However, differential
association theory also hypothesizes delinquent peer networks affect delinquency through
the reinforcement of delinquent values and beliefs. Delinquents not only learn the trade
of delinquency, but also internalize peers’ delinquent value systems (Sutherland &
Cressey, 197 8). The causal relationship between attitudes and delinquency remains
unclear. It is possible that delinquents express attitudes favorable to delinquency as a
neutralization technique that enables them to engage in behavior they believe is wrong in

most situations (Costello, 2000; Sykes & Matza, 1957).

Self-Esteem

Self-esteem is a complex concept that includes multiple aspects of personal
perception including self-acceptance, self-respect, and feelings of self-worth. It is
comprised of four aspects including reflected appraisals, social comparison, self-
attribution, and psychological centrality, and is an important concept in personal
development because it indicates how individuals navigate their external world.
Reflected approaches are those implied by Cooley’s (1912) “looking glass self.” From

this perspective, peoples’ self image is based heavily on what others think of them,

33

especially those that play an important role in our lives. Social comparisons imply
people develop their self-identify by drawing comparisons between themselves and
others. That is, they judge their relative strengths, weakness, and relative “worth” by
comparing themselves to others. Self-attribution suggests individuals judge themselves
independent of others by observing and evaluating their own behavior.

In most cases, self-esteem is expected to have a simple direct negative effect on
delinquency (Kaplan & Damphousse, 1997). Others suggest more complex relationships
where self-esteem and delinquency interact or where delinquent peers mediates the
relationship between self-esteem and delinquency (J ang, 1998). At the point of initial
delinquency, a negative relationship would be expected where low self-esteem leads to
increased levels of delinquency. However, involvement in delinquency is then expected
to increase the level of self-esteem and reinforce the value of delinquent behavior. This
feedback loop subsequently increases the likelihood of additional deviant behavior (J ang,
1998). Empirical findings about the relationship between self-esteem and delinquency
remain unclear. While Ellickson and McGuigan (2000) reported a statistically significant
relationship between low self-esteem at the 7th grade and both relational and predatory
violence at age 18, J ang and Thomberry (1998) found little support for relationships
between low-self esteem and delinquent associations or delinquent behavior.

Self-esteem is not a unidimensional construct but can be differentiated based on a
sense of global versus specific self-esteem (Rosenberg, Schooler, & Schoenbach, 1989),
and part of the reason self-esteem has failed as a consistent predictor of delinquent peers

or delinquent behavior is due to poor theoretical and methodological specifications of the

34

construct. Global self-esteem is more relevant to psychological well-being while specific

self-esteem is a better predictor of behavior (e.g., delinquency and school performance).

Alcohol and Drug Abuse

Delinquency and the use of both drugs and alcohol are some of the most persistent
problems facing many societies today. Like much that has already been discussed, it is
difficult to determine the exact causal relationships between substance abuse and
delinquency. One of the first challenging aspects in the relationship is that in many cases
the two concepts are highly correlated because they are often used as measures of one
another. The confounding issue in the debate is alcohol and drug abuse can be conceived
of as examples of delinquency and not necessarily distinct concepts. If considered
conceptually distinct, there is a general consensus that delinquency causes substance
abuse and not vice versa (Brooks, Cohen, Whiteman, & Gordon, 1992; Huizinga, Loeber,
& Thomberry, 1994).8 More specifically, early childhood antisocial behavior is a strong
predictor of adolescent substance abuse. While delinquency often precedes substance
abuse, delinquency is neither a necessary nor sufficient predictor of problem substance
abuse.

Dembo and colleagues (1987) reported 33 percent of juveniles entering a
detention facility in Florida tested positive for the use of a single drug with marijuana (26
percent) the most common. Moreover, 6 percent tested positive for the use of two drugs
with the combination of marijuana/hashish the most common (3.5 percent). Interviews
with a sample of 125 inner-city males aged 16-25 identified drug and alcohol use as the

most common type of social/recreational activity for young males (Fagan & Wilkinson,

 

8 See also Stice, Myers, and Brown (1998).

35

1998). Lipsey and Derzon’s (1998) meta-analysis of predictors of violent and serious
delinquency showed consistently strong relationships between early substance use and
delinquency. Additional research also demonstrates that juveniles who experiment with
illegal substances between ages 6-11 are over 8 times more likely to be characterized as
serious or violent delinquents by age 15-25. The significance of substance use as a
predictor of involvement in delinquency is so substantial, Lipsey and Derzon (1998)

reported it second only to influences of early criminality and gender.

Social Risk Factors

Risk factors for delinquency are not limited to individual characteristics such as
age, gender, and race but also include influences outside the direct control of individuals
themselves. Delinquency research has identified several features of social settings that
affect the onset and trajectory of delinquency: peer groups, school settings, family

environments, and stressful life events.

Peer Influences

Delinquency is predominantly a group phenomenon9 (Zirnring, 1981, 1998)
among both boys and girls (Erickson & Jensen, 1977). In a sense, the term “gang”
symbolizes the group dynamic of youth crime and the extent to which peer networks
influence delinquency. Classic works on youth gangs have clearly documented the
connections between peer networks and participation in gang activity (Spergel, 1964;

Spergel & Curry, 1990; Sullivan, 1989; Thrasher, 1936). The group dynamic is,

 

9 Elliot and Menard (1996: 31) suggest this argument is overstated. The group dynamics of delinquency
also vary by age, race, and gender (Reiss, 1986).

36

however, stronger for certain crimes such as alcohol use and vandalism and less so for
other crimes such as theft and shoplifting (Erickson & Jensen, 1977).

As indicated earlier, males are involved in more delinquency than females.
Although suggested this may be partly explained by increased opportunities for
adolescent boys caused by greater freedom, it may also be because males are more likely
than females to have delinquent peers. In an analysis of data from Wave III of the
National Youth Survey”), males reported significantly larger percentages of friends
involved in serious property crime (18.5 percent reported having friends involved in
burglary compared to 7.3 percent for females), violence (54.5 percent of males reported
fiiends involved in assaults compared to 32 percent for females), and the sale of hard
drugs (12.2 percent compared to 8.6 percent for females). Males were also slightly more
likely to report friends that use alcohol (males 77.7 percent; females 72.7 percent) yet
both reported statistically similar percentages of friends that use marijuana (males 58.5
percent; females 54.7 percent, n.s.) (Mears et al., 1998: 258). Maguin, Hawkins,
Catalano, Hill, Abbot, and Herrenkol (1995) reported that individuals with negative peer
associations at ages 14 and 16 are significantly more likely to self-report violence at age
18.

Peer networks can be differentiated into five distinct groups (Elliott & Menard,
1996). Saints are individuals that report having at least two fiiends but none involved in
delinquency. Those in the prosocial group report very low exposure to delinquency.
Respondents in mixed groups report exposure to delinquency where several friends are
involved in more than one type of delinquency. Individuals in the delinquent peer group

report extensive exposure to delinquency. In this case, most fiiends were involved in

 

'° Collected in 1978, N=1,626.

37

more than one type of delinquency. In contrast to the above classifications, isolates
report low levels of exposure to delinquency via peers, yet this is often a function of
minimal ties to peers.

As also discussed in the section on age-related factors that influence delinquency,
the temporal relationship between delinquency and exposure to delinquent peers is
unclear. Elliott and Menard (1996) tested the relative efficacy of both control and
learning theories in explaining the relationship between delinquency and peer networks.
They specifically tested the merits of both control and learning theories to determine if
delinquency precedes exposure to delinquent peers (control theory) or if exposure to
delinquent peers precedes onset of delinquency (social learning theory). Their analysis
provides initial evidence for social learning theory as the authors generally conclude
exposure to delinquent peers precedes involvement in delinquency for both younger and
older offenders. Among individuals involved in index offending (burglary, theft of over
$50, motor vehicle theft, strong-arm robbery, sexual assault, aggravated assault, and gang
fighting), 91 percent of individuals between 1 l-17 reported exposure to delinquent peers
before involvement in delinquency (Elliott & Menard, 1996: 41). Only 2 percent of
respondents reported delinquency prior to exposure to delinquent peers. This relationship
remains true, albeit to a lesser extent, for individuals reporting involvement in minor
nondrug offending and general offending. For those reporting involvement in minor
nondrug offending (excludes index offenses but includes status offenses such as runaway
and sexual intercourse, larceny less than $50, receiving stolen goods, prostitution, selling
marijuana or hard drugs, simple assault, joyriding, and disorderly conduct), 52 percent

reported exposure prior to delinquency and 17 percent delinquency before exposure. This

38

was nearly identical for individuals reporting involvement in general offending (minor

delinquency plus index offending plus marijuana and hard drug use).

School Influences

Schools have been linked to delinquency through several different features of
personal achievement including academic failure (Denno, 1990; Farrington, 1989;
Maguin etal., 1995), low school bonding (Gottfredson, 2001; Maguin et al., 1995), and
truancy/school dropout (Farrington, 1989; Thomberry, Moore, & Christenson, 1991).
Both males and females with poor academic achievement at ages 7 and 13-14 were
experienced more arrests for violent crime between ages 10-17 (Denno, 1990). Males
that experienced academic failure at age 11 also reported significantly more self-reported
violence at ages 16-18 and age 32 (F arrington, 1989).

Students with low levels of school bonding are report involvement in more
violence later on in life; nonetheless the affects of school bonding appear to be more
important at different ages (Ellickson & McGuigan, 2000). Students experiencing early
(grade 7) problems with school bonding, something evidenced by poor grades and
transitions between schools, report significantly more involvement in violent crime at age
18. Students with poor grades at the 7th grade reported 32 percent more relational
violence and 49 percent more predatory violence at age 18 (Ibid). Although low school
bonding is an important predictor of self-reported violence at age 18, the strength of
association is strongest for students age compared to age 10 and 16 (Maguin et al., 1995).
Thus, school bonding appears to be particularly important during certain developmental
phases of adolescence. Research indicates males with truancy records between ages 12-

14 were over twice as likely to self-report violence between ages 16-18 and at age 32 and

39

almost 4 times more likely to have a conviction for a violent crime between ages 10-32.
Similarly, males that drop out of school before age 15 were 3.5 times more likely to self-
report violence between 16-18 (F arrington, 1989).

Labeling and strain theories are the two explanations for how school factors
influence delinquency. The relationship between school failure and delinquency was first
identified by Tannenbaum (193 8) in his classical delinquency research. The early stages
of delinquency are related to how parents, teachers, and other adult authorities react to
children who for some good reason do not like school (Tannenbaum, 1938). The child
that does not like school generally do not do well, then are further punished by continued
forced attendance. Those that continue to function poorly are then caught in a cycle
where the their continued lack of success discourages them from attending school, a
dynamic that increases their involvement in truancy (Gottfredson & Hirschi, 1990).
During the early 19003, mandatory education was implemented as one of the many
initiatives proposed by early delinquency reformers based on the “progressive” argument
that all children, not just wealthy aristocratic youth, were entitled to education (Platt,
1977)

Not all children experience equal levels of success in the school environment. For
a variety of reasons some children simply do not like school and do not “succeed” as
measured by traditional evaluation criteria (e.g., good grades). School processes such as
the creation of academic tracks and continual evaluation procedures can actually function
to accentuate differences in the academic abilities of students (Kelly, 1982). Built into
such a system is the notion that some students will do well and other students will fail, in

a sense “creating failure” and perpetuating the poor academic attainment and sense of

40

failure among the most at risk students. Children “labeled” as failures operate in a way
that reinforces the negative label already applied by important authority figures. The
inescapable and central conclusion to labeling theory is:

“The attribution of stigrnatizing labels, particularly when that attribution

process involves formal agents of social control, initiates a social process

that results in altered self-conceptions, a reduction in the availability of
conventional opportunities, a restructuring of interpersonal relationships,
and an elevated likelihood of involvement in the real or imagined conduct

which stimulated initial intervention efforts” (Thomas & Bishop, 1984:

1226)

Cohen (1955) proposed strain theory as a explanation for the relationship
between school-related achievement/bonding and delinquency. Cohen argued the
perpetuation of the class system associated with the industrial revolution created
distinct problems for the lower class. The class system in the United States
presents a situation where the relative competency of individuals is not compared
against an objective standard of “good/bad” but instead middle-class, values and
qualities that represent a tempered version of the Protestant work ethic. Success
in conventional social institutions such as schools is contingent on middle-class
values such as “ambition, individual responsibility, self-denial, rationality, delay
of gratification, industry, manners, control of aggression, wholesome recreation,
and respect for property” (Gottfredson & Hirschi, 1990: 161), values proposed to
be in short supply among the lower class. Strain is experienced when youth are
presented with a situation of considerable disjuncture between definitions of
success (e.g., good grades and good behavior) and available means (e. g.,

appropriate middle-class values). In the face of a disjuncture, criminal acts are

comnritted as a means of attaining “status” goals.

41

Family Influences

The Child-Saving Movement identified the family environment as the primary
breeding ground for delinquent activity and relegated most delinquents to
institutionalized settings to separate them from delinquent influences. Exposure to
criminal influences is an important predictor of delinquency. Children that live in
households with who have histories of criminality, for example, experience 2 to 3 times
more delinquency than those with noncriminal parents (Glueck & Glueck, 1950), and
fathers’ criminality significantly increases violence by age 18-23 (Baker & Mednick,
1984). Fanington (1989) similarly found male children whose parents experience an
arrest before age 10 are more than twice as likely to self-report violence between ages 16-
18 and 3 times as likely to have an arrest for a violent crime before age 32.

More recently, Rowe and Farrington (1997) used both path analysis and
regression techniques to test the relationships between family criminality (convictions)
and child criminality (convictions). Testing the extent to which aspects of family
environment (family size, parental supervision, and parental childrearing style) mediated
the hypothesized direct positive relationship, the results supported the conclusion that
parental criminality has a direct and positive effect on child criminality even when
controlling for dynamics of the family environment. The findings also suggest
“transmission” lines may be gender specific whereby the fathers’ criminality is a better
predictor of the criminality of male children and mothers’ for female children.

In addition to parental criminality, family management style is also an important
predictor of delinquency (Farrington, 1989; Wells & Rankin, 1988). Children that

experience poor childrearing practices at age 8 (males) are more than twice as likely to

42

report violent crime victimizations at age 32 (Farrington, 1989). Moreover, children
growing up in a household with high levels of parental strictness and parental
punitiveness are significantly more likely to participate in violence throughout
adolescence (Wells & Rankin, 1988). Data from the Chicago Youth Development Study
suggests early family neglect is correlated with later covert (e.g., property crime) and
overt (e.g., aggression and violent crime) delinquent behavior, and early family conflict
consistently demonstrates a positive correlation with authority conflict (authority
avoidance) and both covert and overt delinquency (Gorman-Smith, Tolan, Loeber, &
Henry, 1998).

In their analysis of dynamics of urban families from several areas of Philadelphia,
F urstenberg et a1. (1999) attribute much of the relative success of youth to characteristics
of the family management style. Although the authors argue families exist in a larger
community context that have different degrees of adult control, cohesiveness, opportunity
structures, dangers, and internal networks, aspects of the family history, family culture,
and family organization have direct and substantial impacts on the long term success of
children. Family processes such as positive family climate, discipline effectiveness, and
support for autonomy were significantly associated with academic competence,
psychological adjustment, self-competence, and negatively associated with problem
behavior. Similar relationships were evidenced for family management strategies
including institutional connections and social networks.

While Furstenberg et al. (1999) argued the effectiveness of family processes were
mediated by community characteristics that limit institutional connections, other family

processes such as support for autonomy and consistent, clear, and fair discipline remained

43

important regardless of resource issues. Substantial levels of within community variation
indicate families or individuals were not necessarily much better off in communities
characterized by greater social cohesion or better institutional resources, but instead there
is a strong connection between family management practices and community resources
(Chase-Lansdale, Gordon, Brooks-Gunn, & Klebanov, 1997; Furstenberg Jr. et al., 1999;
Furstenberg Jr. & Hughes, 1997; Halpem-Felsher et al., 1997). Acknowledging the
complexity of the problem, Furstenberg et al. (1999: 214) concluded: “To be sure, the
family is a powerful and essential influence in setting children on a successful course, but
parenting practices are only part of the ingredients that provide successful development in

early adolescence.”

Stressful Life Events

Recent emphasis of criminological research has identified the impact of life
transitions and turning points in the life course on delinquency. Life transitions such as
marriage may result in desistance from crime (Farrington, 1995) where other features
such as chronic unemployment increase involvement in delinquent behavior (Crutchfield
& Pitchford, 1997; Fanington, 1995). Developmental criminology presupposes that
stressful events in the lives’ of children increase involvement in delinquency (Elliot,
Huizinga, & Ageton, 1985). Elliot (1994) found no association between stress in the
home during childhood (e. g., serious illness, unemployment, divorce or separation, or
serious accident during the past year) and self-reported violence during adolescence and
adulthood.

The impact of stressful life events is related to the cumulative number of stressful

life events during a certain time period. In a more comprehensive study, Hoffinan and

Cerbone (1999) found a statistically significant relationship between stressful life events
(similar to those just mentioned) and delinquency in a nonrandom sample of 651
adolescents age 11-14. After controlling for the effects of age-related influences,
stressful life events remain associated with significant increases in delinquent behavior.
Furthermore, stressful life events are not mediated by demographic characteristics such as
gender or socio-economic status. Contrary to the hypotheses of the researchers, the
relationship between stressful life events and delinquency was also not mediated by other
personality characteristics such as self-esteem. The relationship is particularly strong
among individuals reporting higher cumulative levels of events (Hoffinan & Cerbone,
1999)

The theoretical relationships between stressful life events and delinquency relate
to Agrrew’s (1992) general strain theory. Adolescence is a time when young people
expand their social worlds, experience more occurrence of poor treatment by others, and
are generally faced with stressful situations. They also experience increased
responsibility over their own worlds at home and school present. These new experiences
are often more challenging in both reality and in perception because juveniles have not
yet developed effective strategies for handling these situations. “Due to developing
cognitive capacities, adolescents’ perceptions of their social world are often self-directed
and introspective. Stresses and strains tend to be magnified by adolescent eyes”
(Hoffman & Cerbone, 1999: 345). In the face of real and magnified challenges faced by
adolescents, they lack the capacity to deal with adversity and the ability to effectively
handle difficult situations (Agnew, 1992). Thus, juveniles who experience multiple

stressful life events are expected to respond with delinquency.

45

Ecological Risk Factors

Shaw and McKay (1942; 1969) propelled the study of the structures of urban
communities into the forefiont of criminological research during the 19403. The study of
delinquency and urban structure was based on several observations, namely that “among
areas in the city there are wide differences in the rates of truants, of delinquents, and of
adult criminals, as well as in disease and mortality rates and other indexes of well-being”
(Shaw & McKay, 1969: 17). The social disorganization perspective supports the
argument that these urban characteristics are not random, but are directly related to
employment, housing, and immigration patterns. More specifically, Shaw and McKay
(1942; 1969) argue the structure and social organization of communities are adversely
affected by structural factors such as poverty, ethnic heterogeneity, and residential
mobility.

Social disorganization theory does not imply direct relationships between macro-
level influences such as mobility and ethnic heterogeneity and rates of delinquency but
instead hypothesize indirect relationships (Bursik Jr., 1988; Cattarello, 2000; Sampson &
Groves, 1989; Veysey & Messner, 1999; Welsh, Greene, & Jenkins, 1999; Welsh,
Stokes, & Greene, 2000). Residential mobility, ethnic heterogeneity, and poverty make it
difficult to establish quality social institutions (Komhauser, 1978) such as schools (Welsh
et al., 2000), but they also reduce the development of informal social control structures
such as watchful neighbors (see discussion of role of “old heads” in urban communities
by Anderson, 1999; Berry & Kasarda, 1977; Sampson & Groves, 1989; Veysey &
Messner, 1999). Some features of communities increase levels of unsupervised youth in

communities and reduce levels of participation in neighborhood groups, and these

46

features of commrmities ultimately increase levels of personal violence victimization
(Sampson & Groves, 1989). Levels of community poverty and community stability, for
example, are among the strongest predictors of both community crime and also school
violence in situations where schools were previously considered unstable (Welsh et al.,
2000).

Social disorganization has implications not only to macro-level characteristics of
communities but also to the study of variation in the behavior of individuals. For
example, nonchronic and chronic violent juvenile offenders from the Rochester Youth
Development Study were significantly more likely to come from neighborhoods with
high crime rates (Thomberry, Huizinga, & Loeber, 1995). Cross-national research has
also found youth that live in poverty experience significantly more court referrals and
significantly more self-reported delinquency. Economic hardships (low socio-economic
status, low family income, family on welfare, or poor housing) between ages 8-10
increase delinquency later in life, and youth living in bad neighborhoods are reported to
experience twice as many court referrals and twice as many delinquent episodes than
their counterparts living in better communities (F arrington, 1998). Gottfiedson, McNeil
and Gottfiedson (1991) reported community disorganization increased self-reported
interpersonal aggression in a sample of urban school students, a relationship that is even
stronger for females.

Defining community disorganization based on respondents’ perceptions of crime
levels, drug selling, presence of gangs, and poor housing, Maguin et a1. (1995) indicate
youth living in disorganized communities between ages 14-16 report a greater variety of

criminal acts at age 18. Similarly, youth report increased levels of delinquency if they

47

live in neighborhoods with high availability of drugs, high exposure to delinquency
(Maguin et al., 1995), and higher levels of adults involved in crime (Maguin et al., 1995;
Thomberry et al., 1995). Thus, it is generally expected that ecological factors such as
poverty, family instability, and residential mobility increase the prevalence of
delinquency, chronic delinquency, and serious violent delinquency.

Research during the past century has identified a comprehensive set of
characteristics that are positively associated with features of delinquent careers that relate
to individual, social, and ecological risk factors. Attention to different subsets of these
features has vacillated depending on the dominant paradigm accepted during a particular
timefrarne. For example, research primarily focused on ecological influences on crime
during the 19403 as Shaw and McKay’s research dominated criminological thought.
Similarly, theoretical and empirical work shifted to aspects of strain theory during parts
of the 19603 and 19703 as academics and policymakers were influenced with the
argument that aspects of unequal socio—economic relationships and commensurate
opportunities increased criminal propensity (see Cloward & Ohlin, 1960). In many cases
researchers operating within certain theoretical paradigms have attempted to discredit the
efficacy of one school of thought in favor of their own, an attempt to establish the relative
merits of one over the others. Theories such as the developmental perspective have
integrated several different paradigms of thought into unified dynamic models that are
advanced models that represent a better picture of the onset and trajectories of delinquent

careers.

48

The Developmental Perspective

Human development has commanded the attention of scholars and philosophers
for centuries and is concerned with the readily identifiable stages in grth that move
individuals through the life cycle from birth to death (Strom, Bernard, & Strom, 1987).
Several methodological techniques have been utilized to study human develop including
biographical studiesl 1, questionnaire studies, cross-sectional studies, longitudinal
investigations, experimental methods, and clinical studies. The study of human
development has waned between constitutional explanations that attribute developmental
characteristics to basic differences between individuals (e. g., heredity), behavioral
conditioning and social interactions (e. g., social learning theory). Behavioral
conditioning explanations propose behavior is coded through processes of responses to
stimulus that reward or punishment certain behavior. Developmental criminology
integrates many of these frameworks into unified conceptual models that elucidate the
interplay between individual constitution, social, and environmental influences in the
sequencing of “normal” versus “abnormal” developmental patterns.

Developmental criminology captures the interaction between delinquency risk
factors and normal developmental sequencing. Childhood is a time of unprecedented
physical, cognitive, psychological, and social development. Each age-related stage in
childhood is important to the development of a healthy and “normal” child, that is, a child
that is likely to be a well-functioning member of society. Developmental perspectives

identify stage-specific developmental sequencing, age-appropriate behavior, how

 

" Pestalozzi (1746-1827) initiated the process of writing detailed biographies of children as a way of
adequately describing problem behavior. This method was not without problems, most pronounced of
which was the ability of the author to maintain objectivity when describing behavior of interest (Strom et
al., 1987).

49

developmental deficiencies or advances affect later development, and how the
interactions of multiple “risk” or “protective” factors at different stages influence current
or future behavior. Among the many normative events12 that are experienced, several can
be distinguished in terms of their significance including marriage, parenthood, and
employment. “Each of these events brings a change in the role status that alters the
general patterns and character of a person’s social relationships, creates a distinctive set
of role expectations, and requires persons to undergo experiences of role socialization”
(Adams, 1997: 328). Behavior can only be defined as “abnormal” when compared to a
standard of normative or normal behavior. The categorization of behavior as normal or
abnormal is not only contingent on the behavior itself, but on a characterization of
particular behavior as age or stage appropriate.

“For example, no one can say whether it is abnormal for a child to cry

when asked to separate from parents unless they know where the

separation is taking place and, more importantly, the child’s age. Most

three-year olds are expected to show some signs of discomfort on being

separated from their parents in a clinic waiting room. A thirteen-year—old

girl, however, would not normally cry and hang on to her parents in these

circumstances” (Yule, 198 1 : 6).

There are “stage-salient” issues at multiple developmental points in early
childhood that are, from the point of view of normal development, important
developmental milestones. As stage-salient issues emerge, they are construed to be
continually important throughout the life course because they remain critical to the
child’s continual adaptation.

“As new [developmental] tasks emerge, old issues may decrease in

relative salience [yet they remain important]. Consequently, each issue

represents a life-span developmental task that requires ongoing
coordination and integration in the individual’s adaptation to the

 

'2 The terms “normal” and normative events are used synonymously in this case.

50

environment and to the stage-salient developmental issue of the period”
(Cicchetti, Toth, & Bush, 1988: 8).

Development Theory and Criminology

Human development theory posits that the life span is characterized by behavioral
regularities that can be predicated with some level of certainty. Increased understanding
of regularities of human developmental has lent itself to the application of developmental
theories to the study of crime and delinquency. More specifically, to the growing body of
research that has clarified the relationships between developmental patterns and the
onset”, trajectory, and persistence of deviant behavior. Developmental theories have the
capacity for conceptualizing changing patterns of criminality associated with age and
how changes in personality, peer relations, family interactions, and life circumstances
correspond to changes in criminality (Adams, 1997). Several consistent observations in
criminological research clarify the inherent associations between developmental theories
and delinquency:
0 “Criminal behavior is relatively uncommon during childhood, even though many

youngsters exhibit precursor behaviors during this developmental stage (see

Keenan, Shaw, Delliquadri, Giovannelli, & Walsh, 1998).

o The onset of actual delinquent and criminal behavior increases rapidly during late
childhood and early adolescence, roughly from the ages of ten through fourteen.

 

'3 Onset or age of onset refers to the age at which a child is first recorded as participating in delinquent
activities and has consistently been identified as an important characteristic of delinquency (Adams, 1997;
Blumstein, Cohen, Roth, & Visher, 1986; Blurrrstein et al., 1985; Catalano & Hawkins, 1996; Farrington,
1985; Farrington & Tarling, 1985; Glueck & Guleck, 1930, 1970; Loeber & Dishion, 1983; Loeber &
Snyder, 1990; Piquero, Paternoster, Mazerolle, Brame, & Dean, 1999). In some cases it refers to the first
official contact with the police (see Wolfgang et al., 1972) and in other instances it refers to first occurrence
of self-reported delinquency (see Elliot, Huizinga, & Menard, 1989). Researchers have questioned the
value of concepts such as age of onset and temporal ordering of behaviors (see Blumstein et al., 1986) but
research has generally found substantial support (Elliot et al., 1989; Elliott, 1994; Elliott & Menard, 1996;
Greenberg, 1985; Grobsnrith, 1989; Mazerolle, Brame, Paternoster, Piquero, & Dean, 2000; Nagin &
Farrington, 1992; Piquero et al., 1999; Sampson & Laub, 1997; Thomberry, 1997a).

51

0 The prevalence of criminal involvement reaches a peak during middle to late
adolescence, at about ages sixteen and seventeen.

- Following the peak, there is a rapid decline in offending, with criminal behavior
tapering off by the early twenties for most offenders” (Thomberry, 1997a: 1).

Developmental theories have several advantages over traditional etiological
theories of crime. First, they have the potential for offering better descriptions about how
developmental risk factors interact with characteristics of delinquent careers such as age
of onset, length of criminal careers, extent of specialization in certain types of crimes,
and desistence from delinquent or criminal behavior. Another important advantage of
developmental theories is that they have the potential for creating typologies of offenders
by disaggregating “pathways to delinquency” (Loeber & Stouthamer—Loeber, 1996). For
example, there may be substantive reasons to believe juveniles that initiate drug use at
age 7 are qualitatively different from those that initiate at ages 12 or 17.

Most serious delinquents demonstrate early risk factors but not all individuals
with early risk factors become delinquents (Farrington, 1985, 1992; Huizinga & J akob-
Chien, 1998; Huizinga, Loeber, Thomberry, & Cothem, 2000; Loeber & Stouthamer-
Loeber, 1996). This being the case, it is meaningful to understand the combination of
risk factors that result in future criminality, and conversely, how risk and protective
factors interact to effectively shield at-risk individuals from delinquency (Le Blane &
Loeber, 1998). Finally, developmental perspectives offer the advantage of explaining
how developmental changes during the life course explain changing patterns of
delinquent behavior (Thomberry, 1997a). The developmental approach differs from the
others because the focus on is on within individual variation compared to research that

focuses on changes between individuals.

52

Developmental Pathways

The study of “pathways to delinquency” has made a major impact on the field of
criminology and serves as a prime example of the application of developmental theory to
the study of delinquency. Delinquency often follows predictable patterns that involve
characteristics of age of onset, involvement with specific types of delinquent behavior,
and a sequencing of behavior fi'om minor to more serious forms of delinquency (Loeber
et al., 1993). Developmental pathways are typologies of delinquency “careers” that are
useful for understanding the progression from early noncriminal delinquent behavior to
more serious forms of criminal behavior. Dynamic classification schemes that use a mix
of post problem behaviors rather than a single act provide greater opportunity to
understand developmental factors and predict future behavior (Loeber, 1996).

Loeber et al. (1993) proposes that delinquency follows one of three
developmental pathways: (1) the overt pathway; (2) the covert pathway; and (3) the
authority conflict pathway. The overt pathway is represented by an escalation from
minor aggression such as bullying and behavior intended to annoy others, to more serious
delinquency such as physical fighting. A certain subset of individuals will progress from
physical fighting into more serious forms of overt delinquency such as violence. In
contrast to the overt pathway where individuals tend to be involved in delinquency to a
great extent, individuals in the covert pathway escalate from minor covert behavior such
as shoplifting and frequent lying to more serious property crime like vandalism, arson,
and destruction of property. For delinquents that “mature” into the next stage of the
pathway, serious covert behavior generally centers on more harmful property crime such

as burglary, serious theft, and fraud. Finally, children involved in the authority conflict

53

pathway initiate delinquency earlier than the other two and escalate from stubborn
behavior that is manifest before age 12. Stubborn behavior leads to defiance and
disobedience and to a more serious involvement in status offenses such as truancy,
running away from home, and curfew violations. Adolescents in the authority conflict
pathway strongly reject the role of parents and other authority figures in controlling their
behavior. Some adolescents that initiated delinquency in the authority conflict pathway
either stayed in this pathway or made the transition into either the overt or covert
pathways.

Delinquents follow predictable transitions from minor to serious forms of
delinquency as they mature through the adolescent years. With each progressive step
toward more serious crime, the number of individuals participating in such acts gets
smaller. Loeber et al. (1993) reported strong integrity to developmental pathways. The
majority of delinquents maintain “specialization” in one of the three developmental
pathways, although a sizable proportion of can be classified as fitting into more than one
delinquent path. Moreover, youth in the overt and covert pathways experience
significantly more delinquency than those in the authority conflict path.

For the developmental perspective to be empirically valid, we would expect
offenders that recidivate to primarily concentrate their delinquent activities in certain
types of behaviors and, for those that consistently recidivate, escalate offending in a
reasonably predictable manner. Wolfgang et al. (1972) provide initial support for this
conclusion. Analysis of transitional probabilities indicated juvenile recidivists were most
likely to commit a delinquent act similar to the current delinquent act. For example, the

probability that juveniles arrested for the first time and for a nonindex offense would

54

recidivate with another nonindex offense was .33. The next closest probability was for
.04 for an injury offensel4 (Wolfgang et al., 1972: 176). Similar probabilities were
experienced across types of delinquency such as injury, theft, damage, and a combination
of offenses but also as offenders continued to recidivate through the eighth offense. The
probability of desistance also decreases substantially with each subsequent rearrest
(Loeber, 1988; Wolfgang et al., 1972). Thus, Wolfgang’s research provides baseline
evidence for the developmental perspective.

There is also a close relationship between substance abuse and delinquency where
the early use of illegal substances predicts future delinquency. Among juveniles who
self-report both delinquency and substance use, involvement in delinquency usually
precedes substance abuse (Elliot & Huizinga, 1984). The developmental sequencing of
substance use starts with the use of beer or wine and escalates to cigarettes or hard liquor,
marijuana, and then other illicit drugs (Kandall, 1982). Similarly, the developmental
sequencing of disruptive child behavior begins with stubborn behavior followed by
covert delinquency (e.g., lying and shoplifting), defiant behavior (e.g., doing things own
way or refusing to do things requested), aggressive behavior (e. g., annoying others or
bullying), property damage, moderate delinquency (e. g., joyriding, pickpocketing,
stealing from car, and illegal use of checks or credit cards), serious delinquency (e.g.,
stealing a car, selling drugs, and breaking and entering), authority avoidance (e.g., staying
out late, truant, and running away), fighting, and violence (Loeber et al., 1993). When

both behaviors co-occur, involvement in minor forms of delinquency usually precedes

 

'4 The probability of desisting after the first offense was .51.

55

substance abuse. The basic premise, then, is that concurrent behavior is predictive of

future behavior. 15

Age of Onset

Age of onset is one of the most important characteristics of developmental
criminology. Wolfgang et al. (1972) reported several general findings about the
relationship between age of onset and delinquency. Using police reports (official
records) as measures of delinquency, the authors reported that individuals who began
their delinquency at age 13 committed more offenses from onset years through age 17
than did others. The data also reveal that individuals with first arrests at an older age
present more of a danger to the community. Taking into account seriousness of the
offense, the weighted offender rate increases steadily from age 7 (313.52) until age 16
(7,453) where it begins to decline by age 17 (6,151) (Wolfgang et al., 1972: Table 8.3).

Using data from 1963, 1964, and 1965 birth cohorts, Elliot, Huizinga, and Menard
(1989) reported nearly all of delinquency, including school suspension, was initiated
before age 17. Among the various indicators of delinquency, the largest percentage of
initiation past age 17 was for multi-drug use (1 1 percent). Age of onset is not only
related to the prevalence of delinquent behavior, but also connected to the types of
problematic behavior in which individuals are involved. Stubborn behavior, for example,
has an earlier onset than more serious behavior such as defiance and authority avoidance

(Loeber et al., 1993: 116). The age of onset for minor delinquency, alcohol and cigarette

 

'5 Developmental continuity in delinquent behavior should not be wholly unexpected based on research by
developmental psychologists. There is, for example, evidence of continuity in problem behaviors for
young children. Difficult temperament at 12 months is predictive of difficult temperament at 18 months,
non-compliant behavior at 24 months, and clinically diagnosed internalizing behavior at 36 months.
Similarly, noncompliant behavior at 24 months is predictive of clinically diagnosed internalizing and
externalizing behavior at 36 months (Keenan et al., 1998).

56

use, and moderate and serious delinquency is between ages nine and fifteen while the age
of onset for marijuana use and hard drugs is usually between ages thirteen and eighteen
(Loeber, Van Karmnen, Stoutharner-Loeber, & Farrington, 1996). Maxwell and Maxwell
(2000: 795-796) reported the mean age of onset into prostitution among females to be in
the early 20’s, similar to onset into drug sales (age=22) and first use of crack cocaine
(age=20-25). In general, the delinquency age-curve is evident for aggressive behavior,
fighting, and other forms of violence (Loeber & Hay, 1997).

It is possible to predict the average age of onset for particular forms of
delinquency. For property offenses, less serious property offense such as theft at home
has the lowest average age of onset (3 =10.1, s.d. = .14) followed by shoplifting
(1? =10.7, s.d. = 1.40), cheating/theft of school property (a? = 11.1, s.d. = 1.07), theft from
peers (f = 11.4, s.d. = 11.4), theft by trespass (f = 11.7, s.d. = .69), and vehicle
theft/break and enter ()7 = 12.7, s.d. = .63) (Loeber, 1988: 79). There are basic
differences between individuals that get involved in delinquency earlier rather than later.
Early starters are likely to experience additional risk factors such as problematic
parenting (Hawkins, Catalano, & Miller, 1992; Lipsey & Derzon, 1998; Loeber, 1990;
Loeber & Dishion, 1983; Simons, Wu, Conger, & Lorenz, 1994). Simons, Wu, Cogner,
and Lorenz (1994) reported poor parenting predicted development of antisocial behavior
such as oppositional/defiant disorder, a relationship that does not exist for late starters.
For late starters, poor parenting was positively associated with deviant peers. In this
case, the presence of oppositional/defiant disorder had no significant effect on the

development of delinquent peers or in the involvement in delinquent behavior.

57

Juveniles who initiate offending early in life are at risk for developing into life-
persistent offenders and being more diversified in the types of delinquency they are
involved. In comparison, adolescent-limited offenders initiate delinquency later and
remain more specialized in offending patterns (Moffitt, 1994). Juvenile offenders who
eventually develop into serious and chronic offenders begin their criminal careers
between ages 12-20, and more than half will initiate violence between ages 14 and 17.
The highest rate of onset is found for African Americans at age 5 and at age 16 for
Whites (Elliott, 1994). Violent delinquents generally initiate delinquency much earlier
than serious nonviolent and nonserious nonviolent delinquents, and serious nonviolent
delinquents also initiate their careers significantly earlier than nonserious nonviolent
delinquents (Tolan & GormamSmith, 1998). Controlling for other important predictors
of future criminality such as IQ, parental criminality and behavior, risk-taking, and
parental divorce, the authors found age of onset a significant and substantial predictor of
future convictions. Nagin and Farrington (1992: 513-514) reported the probability of
conviction at t5 as .036 for individuals with no convictions prior to age 18, .227 for
individuals with first conviction at age 16, and .318 for individuals with first conviction

at age 10.

Integrative Theory: Interactional Theory and Similar Models

Thomberry proposed interactional theory, an extension of developmental theory,
as a way of providing more dynamic explanations for delinquent behavior (Thomberry,
1987). Interactional theory is similar to other developmental theories in that it

presupposes there are important temporal and causal sequences among the predictors of

58

delinquency described above that traditional theory has not effectively conceptualized.
Interactional theory is more dynamic in two important ways:

“(1) It explicitly recognizes the importance of developmental change in

accounting for delinquency, and (2) it views human behavior, including

delinquent behavior, as a result of interactive and reciprocal causal

influences that develop over time” (Thomberry, 1996: 199).

A3 a “complete theory of delinquency” (Thomberry, 1987, 1996), interactional theory is
better able to explain why some individuals initiate delinquent conduct while others do
not, why some offenders sustain involvement in delinquency over long periods of time
while others do not, and why some offenders desist delinquency early while others
continue (Thomberry, 1996).

Although recognizing the importance of the risk factors identified earlier,
interactional theory stipulates the impact of these factors is not equal at all developmental
periods. Interactional theory also hypothesizes the impact of many risk factors are in fact
amplified when they exist concurrently with others. Thomberry (1996: 201), for
example, suggests it is plausible that the reason authors interpret their results as
suggesting parental influences are relatively weak (Elliot et al., 1985) is that data may be
drawn from too late a developmental phase to accurately reflect the causal impact of the
family.

Figure 1 depicts Thomberry’s (1987) interactional theory that hypothesizes
reciprocal relationships between many of the constructs. There are several important
features to this model that make it distinct from other conceptual models. First, the
model is integrative in that it incorporates important concepts that other theories and

research have confirmed are important precursors and predictors of delinquency. It

incorporates social control variables such as attachment to parents and commitment to

59

Figure l. Reciprocal model of delinquency.

 

 

 

Beliefin ........................ Association
Conventional with delinquent '
+ Values peers '

 

 

 

 

 

 

 

 

 

. Delinquent
............. Values

Attachment to
Parents

 

 

 

 

 

 

 

 

 

 

 

   

Commitment to

Delinquent
School

Behavior

 

 

 

 

 

Source: Figure 2, Thomberry 1987: 871.

"' Dotted lines are indicators of weaker relationships.
school, but also includes characteristics of differential association theory such as
association with delinquent peers. The most important implication to Figure l is that
human behavior, including delinquency, is dynamic and develops over time through
interaction among many different causal influences.

Other authors have proposed conceptually similar interactional models.
Furstenberg et a1. (1999) conceptualized child outcomes (e. g., delinquency) as being
directly impacted by characteristics of family history, family culture, and family
organization. These features interact in a way to produce an overall family management
style (individualistic v. collective; protective v. promotive) that affects the general

outlook children have of the world. The authors also appropriately argue that while

family management style is important, families exist in a broader community context that

60

exerts influences on both the behaviors and choices of families and individuals within
families.

The relationship between community and family structure make intuitive sense
because community context provides varying opportunity structures (e. g., access to jobs
and quality schools), degree of adult control (e.g., extent to which other adults supervise
neighborhood youth), community cohesiveness, community resources, internal networks,
and dangers. Families, especially families with financial resources, have a greater ability
to directly alter the affects of community structures by making conscious decisions about
where to reside based on these considerations, but children at the greatest risk for
delinquency often live in families with relatively fewer options. For example, although
geographic mobility is a primary family management strategy for parents who perceive
their neighborhoods to be too dangerous or without adequate resources, many are
prevented from “escaping” due to limited economic resources (F urstenberg Jr. etal.,
1999: 25).

Vuchinich et al. (1992) also noted important inactions family characteristics and
other predictors of delinquency. The authors reported antisocial behaviors of children by
age 11-12 have an adverse impact on effective disciplinary practices by parents even after
controlling for the behavior of the parents from 2 years prior. The more antisocial
behavior demonstrated the more the quality of the parenting itself suffered, something
that increased the probability of increased antisocial behavior. There is also evidence that
juveniles raised in families where parents use tobacco will associate with friends who use
tobacco (association with delinquent peers) (Melby, Conger, Conger, & Lorenz, 1993).

Moreover, Conger and Rueter (1996) reported parental alcohol use as predictive of

61

adolescent alcohol usage (drug and alcohol use). Thus, the cumulative evidence suggests
interaction effects where one set of delinquency risk factors influences the others.

In synthesizing the above information, the most likely conclusion is that the onset,
persistence, and traj ectories of delinquency are correlated with and caused by risk factors
associated with, among other things, demographics (age, gender, race/ethnicity), attitudes
toward delinquency, self-esteem, and extent of substance abuse. Moreover, delinquency
is a product of social influences such as negative peer influences, lack of bonding with
social institutions such as schools, poor family firnctioning, stressful life events, and
ecological factors relating to social disorganization (e. g., poverty, ethnic heterogeneity,
and population turnover).

Although some authors simplify the growing body of research and attribute
delinquency to a single construct like low self-control (Gottfredson & Hirschi, 1990), the
research detailed above suggests criminal propensity results from the co-occurrence and
interaction of multiple risk factors. Developmental theory has furthered criminological
inquiry by conceptualizing dynamic models of delinquency where many risk factors are
integrated into models that specify causal relationships that vary in both direction and
intensity depending on the stage of development. For example, poor family management
style or parental divorce will exert greater influences during early childhood then when a
child is in late adolescence. Conversely, associating with seriously delinquent peers will
also likely exert greater influence during early adolescence compared to peer delinquency
in late adolescence (Thomberry, 1987). Similarly, youth from impoverished
backgrounds have a greater probability of staying out of trouble if they have a strong

family environment that encourages personal development, autonomy, and problem-

62

solving regardless of the extent to which they are exposed to criminal opportunities
(Furstenberg Jr. et al., 1999; Furstenberg Jr. & Hughes, 1997).
The Need for Theory-Based Interventions

A search for the “causes and correlates” of delinquency has been profoundly
influenced by the continual increase in juvenile crime since the early 19003.
Developmental models are valuable advances in this search because they integrate
previous theories into conceptual models that elaborate the etiology and development of
delinquency through the life course. Developmental and other integrative models have
advanced the field not only from the perspective of theory testing but also by
encompassing direct implications for the structuring of interventions. Efforts to
implement comprehensive interventions have expanded since the early 19903. Since
then, policymakers and researchers have demonstrated reinvigorated efforts to identify
potentially serious offenders early on and to implement comprehensive intervention
efforts (Wilson & Howell, 1993).

The notion of early and comprehensive intervention efforts represents a departure
from the momentum after some policymakers concluded rehabilitation efforts were
generally ineffective. Lipton et al.’s (1975) analysis of methodologically sound program
evaluations found limited effectiveness in reducing recidivism or reducing risk factors
associated with delinquency. The authors concluded that although rehabilitative efforts
have the potential to work with certain populations under limited circrunstances, there is
little reason to expect consistent large-scale positive outcomes fi'om rehabilitation

programs.

63

The rhetoric of the 19803 was that rehabilitative efforts are fundamentally flawed
and that there is little hope of making major positive changes in the life-course of
delinquents (Lipton et al., 1975; Wilson, 1983; Wilson & Hermstein, 1985). Some
argued rehabilitation lacks merit because criminal propensity is stable throughout the life-
course (Gottfredson & Hirschi, 1990). However, it is just as likely the failure of the
rehabilitation agenda is at least partly related to poorly designed intervention models that
either failed to adequately conceptualize the causal mechanisms in the process of
“becoming” delinquent or limited intervention efforts to a limited set of risk factors.

To construct effective interventions, programs must have a conceptual
understanding of the problem based on theory and research. Yet implementation staff are
often left with questions like: “When in the planning process do I use theory to guide my
decisions? How do I know what theory to use? How do I make use of the experience of
others and results of other program evaluations? How do I decide what intervention
methods to use?” (Bartholomew, Parcel, & Kok, 1998). Bartholomew et al. (1998)
proposed the concept of “intervention mapping” to delineate the appropriate steps for
applying theory to a problem. Delivering an effective intervention is a difficult task that
involves conducting a needs assessment, developing and implementing a program, and
evaluating the program’s effectiveness, a process known as “intervention mapping”
(Bartholomew et al., 1998). A crucial task in intervention mapping is making decisions
about how to use theory during the developmental and implementation stages.

Like other problem-solving frameworks, Bartholomew et al. (1998) propose an
iterative process that starts with identifying the at-risk population along with a assessment

of their relative “needs.” Subsequent steps involve distinguishing the individual, social,

64

and environmental causes of the behavior (integration of theory), identifying the key
predictors of the behavior, and analyzing the resources that already exist in the
community. This process provides a comprehensive understanding of the problem and its
causes and is intended to lead to the specification of program goals that serve as a
foundation from which to begin intervention development.

In the early 19903 the Office of Juvenile Justice and Delinquency Prevention
(OJJDP) launched a national initiative to fund the development and implementation of
comprehensive intervention strategies for serious and violent juvenile offenders. This
initiative was influenced by rising juvenile crime rates and the recognition that
comprehensive strategies were needed (see Wilson & Howell, 1993). Past efforts have
been criticized for:

“. . .their narrow scope, focusing on only one or two of society’s

institutions that have responsibility for the social development of children.

Most programs have targeted either the school arena or the family

(emphasis added). Successful ...strategies must be positive in their

orientation and comprehensive in their scope” (Wilson & Howell, 1995:

39).

Based on the cumulative evidence about the causes of delinquency, OJ JDP’s
“Comprehensive Approach” initiative recognized the need for intensive efforts that
integrate efforts directed at individual characteristics, family influences, school
experiences, peer group influences, and neighborhood and community characteristics.
For example, the balanced and restorative justice approach supports the core objectives of
accountability, competency development, and community protection in an effort to
reform juvenile delinquents and make them productive and responsible citizens

(McGarrell, 2001; Wilson & Howell, 1995). Since a small percentage of juvenile

offenders are responsible for a large percentage of all crimes (Office of Juvenile Justice

65

and Delinquency Prevention, 2001; Wolfgang et al., 1972) it is highly beneficial to direct
efforts to a population designated as demonstrating sufficient need. Successful
prevention and intervention efforts are those that target and “fix” the risk factors for
delinquency.
Current Research

The data for the current research were collected as part of a multiyear community
based intervention program designed for youth who, absent the program, would have
been adjudicated to non-secure residential placement programs. The program, referred to
here as the Youth and Family Studies Program (YF S), took place in a large urban County
in a Midwestern State. This program was developed in response to substantial increases
in the number of youth under jurisdiction of the juvenile justice system, especially youth
institutionalized in out-of-home placement programs. In many cases, these were non-
secure institutions. Prior to the institution of the community-based program there were
few alternatives to residential care specifically targeted for youth newly committed to
State custody.

As an alternative to resident treatment, the purpose of the program was to provide
a community-based day treatment program that was successful in rehabilitating juvenile
offenders. In this case, rehabilitation is defined as reduced probability of firture criminal
offending and a reduction in the risk factors associated with delinquency. In the course
of this analysis, four research questions will be answered:

Q1: How do initial risk factors relate to recidivism?
Q2: How do treatment effects relate to recidivism?
Q3: How do treatment effects relate to change in risk factors?

Q4: How does change in risk factors relate to recidivism?

66

The Youth and Family Services program established a formal placement
mechanism to evaluate the eligibility of each juvenile to participate in the program
activities. The structured decision making model considered two important factors: (1)
the type of the commitment offense, and (2) risk score as determined by the Department
of Social Services’ Initial Delinquency Risk Assessment Scale (see Appendix A). The
risk assessment scale considered a variety of important background variables such as age
at first adjudication, number of prior arrests, and characteristics of peer relationships that
research has consistently identified as important predictors of future delinquency.
Juveniles were initially screened as “eligible” if they were committed for all but the most
serious offenses (e. g., homicide, attempted murder, very serious assaults, armed robbery,
and serious instances of criminal sexual conduct). The risk assessment instrtunent
constructed by the Department of Social Services was also utilized to ensure program
participants did not have background problems so severe as to negate the likelihood of
effective nonresidential, community-based treatment. Finally, information regarding the
commitment offense and risk assessment were integrated into a security matrix.
Individuals scoring “low risk (0-8 points)” on the assessment who were adjudicated for a
crime other than the most serious should then have been eligible for participation in the
community-based program (see Appendix A).

The first goal of the decision making model was to ensure the program
participants were similar in regards to the type of adjudicated offenses and background
variables. As argued by Lipsey and Wilson (1998) a problem with most programs is that
they have attempted to provide these services to target populations that are extremely

diverse. Programs are often intended to target youth with a variety of commitment

67

offenses, everything from serious felonies to status offenses. In addition, intervention
programs also tend to include participants with a variety of contextual problems such as
varying degrees of mental health, substance abuse, and academic problems.

There are several advantages of directing intervention services at defined target
populations. First, programs with narrowly defined target populations can be structured
in a way that optimizes return per unit of investment. In contrast, those that include
mixed clients such as those classified both as low-risk and high-risk clients, or clients
with extensive needs and those with few needs, run the risk of either exposing individuals
to too few or too many intervention services. Having a target population that is
consistent in regards “needs” reduces the likelihood of too little services to those
individuals requiring intensity and too much intensity to those individuals presenting less
levels of need. This is an important consideration because program resources are often
limited.

Programs with highly mixed client populations can suffer because individuals
with more problems are likely to use more resources in regards to both personnel time
and monetary investments thereby risking the relative likelihood of success for other
clients. Directing intervention services to a clearly identified population allows for more
accurate conclusions about the effectiveness of specific types of interventions for specific
types of offenders thereby increasingly the likelihood of developing successful
intervention models. Finally, intervention programs, especially non-residential
community-based intervention may also exclude high-risk offenders because of concern

they may pose a safety hazard to the general community and because of concern that

68

exposure of less-serious delinquents to hi gh-risk juveniles may actually increase

likelihood of further delinquency.

Main Treament Model

The YFS program was intended to be a comprehensive treatment program that
would serve as an alternative to residential placement. Juvenile justice professionals in
the County strongly supported the development of community-based alternative
programs because there were few other options besides residential placement prior to the
implementation of this program. The service delivery fimctions were contracted to two
service providers, and although there were some minor differences in the type of services
that were to be provided to clients in each program, the programs were similar enough to
allow for them to be considered as one program.

The model for the YFS program stipulated the development of treatment services
in the areas of education, family functioning, career services, individual/group
counseling, community service/reintegration, and intensive supervision. The main goals
of the program were to focus on improving the youth’s behavior, academic performance,
and improve family functioning. The programs were to engage juveniles in intensive and
comprehensive services that mediated the underlying causes of delinquent behavior.
Using a day treatment model, the programs were designed to be 12 months long and

included 6 months of intensive program participation followed by 6 months of aftercare.

Educational Component

Both programs included strong educational components. Educational problems

have consistently demonstrated to be important predictors of delinquency (F arrington,

69

1998; Voelkl, Welte, & Wieczorek, 1999). Research has reported that the likelihood of
delinquency is negatively related to academic achievement. Thus, the purpose of the
educational component was to increase the educational achievement of the program
participants. Increased educational achievement is expected to reduce stress in the school
environment and increase attachment to school, thereby resulting in a decrease in
recidivism.

Educational assessments were to be performed within two weeks from entry into
the program. The purpose of the testing was to assess each youth’s level of academic
proficiency with the explicit intent of tailoring the delivery and content of educational
material to fit the proficiency level of each individual. Educational services were
delivered in a modified school setting where individuals attended on a daily basis.
Students were to receive instruction in the areas of math, social sciences, science, history
and language arts. In addition, individuals were to be given pre-vocational English skills
and health education.

Accredited educational staff and “youth specialists” were responsible for the
delivery of the instructional material. The youth specialists assisted the teachers in
providing more individualized attention to each individual. There was a high staff-to-
client ratio with approximately 3 students per staff. The high staff-to-student ratio
allowed staff to have regular and substantive interaction with students during the school
day. Staff assisted students with the completion of their work assignments, regularly
reviewed class material throughout the school day, and provided regular feedback about

progress through course material. Feedback of regular evaluation efforts was an

70

important component because it established a mechanism to continually modify work

expectations based on the performance of each individual.

Counseling/Peer Counseling Component

Both programs in the YFS also developed counseling components. The purpose
of the counseling component was to provide a safe setting for youth to discuss important
issues that affect their lives. The counseling sessions were in a peer group format that
was facilitated by a group leader who was a member of the program staff. Group
sessions were held on regular basis, usually every day, to discuss important issues in the
lives of the participants. Counseling services do not serve as a “fix” for problems, but
increase general capacities that help individuals address them in a positive manner by
increasing levels of personal insight, empathy, moral reasoning, and anger management.
It is important for juveniles to have an opportunity to communicate about stressful life

events and to develop the capacity to problem-solve.

Family Component

Research has also consistently demonstrated that serious and/or chronic juvenile
offenders are more likely to experience family problems. Results from the Seattle
Development Project reported family problems, in conjunction with neighborhood,
school, and peer problems, to be a significant predictor of gang membership (Hill,
Howell, Hawkins, & Battin-Pearson, 1999). Lewis (1989) reported the interaction of
abusive family environments and neuropsychiatric vulnerabilities has substantial
predictive power of adult criminality. Moffitt (1997) similarly argued chronic

delinquency is rooted in faulty parental interaction. The majority of intervention research

71

suggests effective strategies must recognize the family as an important social unit and
direct attention at the nature and quality of family interactions.

The family component was identified as one of the most important facets of the
problem during the implementation process. Family counselors, youth, and members of
their family were to meet every week to identify important areas where services were
needed. Family counselors were responsible for linking families with the necessary
resources in the community. Families were also tied into substance abuse/drug treatment
services where needed.

Family counselors also were to work directly with families to develop transition
plans for the clients once they were released from the intensive program treatment in
Phase 1 to the aftercare program in Phase 2. Transition plans were agreements between
the juveniles and their parents about behavioral expectations, plans for increasing levels
of educational achievement, and “ground rules” for family interaction. During the
aftercare program, monthly meetings would be continued to ensure compliance with

transition plans.

Vocational Component

Program participants were offered vocational training in one of the following
areas: teaching assistant trainee, administrative assistant trainee, and food service trainee.
Participation in this program was limited to individuals that consistently demonstrated a
commitment to abide by program rules. Eligibility was based on a predetermined number
of tokens individuals would accumulate through the token economy (to be discussed

later). The vocational training was supplemented by actual work experience at the

72

various program locations. During Phase 11, individuals were also required to obtain and

maintain employment.

Recreational Component

Recreational activities were also an important aspect of the YFS program model.
The Office of Juvenile Justice and Delinquency Prevention reported a majority of
juvenile crime occurs between the hours of 3 pm. and 8 pm. (Snyder, 1997). Juvenile
arrests seem to be highest during that time period because it is when juveniles tend to
have an optimal period of unsupervised time. Research by Houston (1996) indicated self-
identified gang members perceive adequate recreational opportunities as important
components of gang prevention strategies for at risk youth. Structured recreational time
is also important because it provides the opportunity for juveniles to develop pro-social
activities. Clients regularly participated in general recreational programs including
basketball, baseball, track, aerobics, and calisthenics. Recreational activities were
conducted within the gymnasium that was located on site by one service provider where
the other service provider used a local YMCA for the delivery of this component of their

program.

Tracking Component

Accountability was an important philosophical underpinning to the program.
Supervision of youth during time spent both in programs and outside of programs is one
core aspect of accountability. The Department of Justice’s Office of Juvenile Justice and
Delinquency Prevention argue a “requisite element for an effective juvenile justice

system” is one that includes “intermediate sanctions that are centered on intensive

73

community-based supervision of juvenile offenders” (Danegger, Cohen, Hayes, &
Holden, 1999: 12).

Accountability measures were established by implementing intensive supervision
of clients, especially during non-program hours. Staff was hired to “track” individuals 24
hours a day, 7 days a week while on probation. While not participating in structured
activities at the day treatment centers, youth were to have at least one face-to-face contact
per day with a tracker and two or more face-to-face contacts each weekend day. In
addition, phone contact was to be made with each youth at least once nightly. The idea
was to keep close contact with individuals to detect deviance from behavioral
expectations and to reinforce the idea that their behavior was being monitored.

Petersilia and Turner (1990) evaluated the effectiveness of three experimental
probation programs for high-risk drug offenders“5 in California. The programs, which
were located in Contra Coasta, Ventura, and Los Angeles Counties, randomly assigned
probationers to intensive supervision programs. They reported intensive programs were
actually more effective in providing increased levels of supervision, a finding that stands
in contrast to previous research. Petersilia and Tumer’s (1990) research was a step
forward in the intervention literature because it made an important step in quantifying to
some extent the level of intervention services programs provided on average to their
clients. This strategy served as an attempt to address assumption 1 detailed above;
namely to determine the extent to which the actual implementation of treatment programs
mirror models identified in program designs thereby determining important similarities

and differences in the nature and intensity of treatment across programs.

 

'6 Offenders convicted of felony and misdemeanor drug dealing, drug use, and nonviolent, drug-related
offenses.

74

Summary

Developmental models are by no means universally accepted as meaningful or
important to the study of crime and delinquency. Gottfredson and Hirschi (1990)
generally discredit the need for frameworks that include multiple risk factors into
complex causal models. The authors attribute criminal propensity to levels of self-
control. One of the primary justifications offered by Gottfredson and Hirschi (1990) in
support of their general theory of crime is that previous research has failed to consistently
demonstrate a relationship between delinquency and other risk factors. Using lack of
consistent support as a means of discrediting the substantive role of other types of
individual, social, and environmental influences on delinquency, they argued delinquency
is reducible to this one construct that is both historically and cross culturally consistent.

This research extends the previous research by further questioning the extent to
which comprehensive and intensive community-based intervention programs actually
provide “intensive” efforts. In addition, the current research will also analyze the direct
and indirect relationships between the “quantity” of program treatment, the underlying
constructs the treatment programs are intended to affect, and outcome measures such as

self-reported and official delinquency.

75

CHAPTER THREE

METHODOLOGY

The current research proposes social developmental theory as an appropriate
framework for conceptualizing how intervention programs should be structured to
provide the necessary resources to curb future delinquency in a sample of adjudicated
youth in a large Midwestern County. As indicated in Figure 2, the current research will
not address issues pertaining to the etiology of delinquent behavior or prediction of why
certain individuals display delinquent tendencies while others do not. Since the sample
includes only adjudicated youth there is no ability to compare delinquents with
nondelinquents. Moreover, this research does not address why or how initial risk factors
develop but how risk and change in risk relates to recidivism. Instead, the purpose of this
research is to determine the effectiveness of community-based intervention in creating
positive change in the risk factors prior research has indicated are responsible for the
onset and trajectory of delinquent careers. Three research questions will be addressed:
(1) Do these delinquents share the same basic risk factors?; (2) Does risk and change in
risk mediate recidivism; and (3) Does treatment mediate recidivism.

This research represents an important addition to the existing body of literature by
investigating the relationship between intervention efforts, risk factors for delinquency,
and measures of recidivism. It is important to consider these relationships because prior
research on the effectiveness of intervention programs has generally made three broad
assumptions: (1) intervention programming is delivered in a consistent manner across
individuals and across time, (2) individuals receive the dose that is detailed in the

program design, and (3) the intervention relates to actual change in the areas or domains

76

Figure 2. Conceptual Model.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

C
5 s.
e- G
a l i l "t
is. g g
3 l l g '8
K i i 3 9‘
5 Individual 0 Social 0 Family 9 Community/Ecological ’- g
3 Characteristics Characteristics Characteristics Characteristics 3
a:
g \ \ / / 3
8
s \9
i- I
e
s.
0
Delinquency 8'
lnt rvention Criminal Offense & '53
Type and Intensity of Self-Reported 2
Specific Programming: 5'
1‘) Education. Independent effects 9
Programming : a
2-) Counseling Mediating effects 3
Programming 6
3.) Family Programming 8
4.) Substance Abuse J 33
Programming ' E
5.) Other Programming Future Delinquency L
I.) Did offender recidivate? E
2.) Did self-reported delinquency increase or a
decrease?

that are believed to underlie delinquency (e. g., educational problems, family functioning,
and peer relationships).

This research also extends the previous by further questioning the extent to which
comprehensive and intensive community-based intervention programs actually provide
“intensive” efforts. This research will analyze the direct and indirect relationships
between the “quantity” of program treatment, the underlying constructs the treatment
programs are intended to affect, and outcome measures such as self-reported and official

delinquency.

77

The Data

Data were obtained from 4 primary data sources that include structured interviews
collected over an 18-month timeframe, court disposition information, court intake data,
and census data. Research staff scheduled meetings with the youth and their guardians
to administer the interviews. Interviews were primarily conducted in the homes, but
alternative sites were also made available where necessary especially in the case where
juveniles were in short-term placement. Juveniles were interviewed at four different time
periods: initial commitment, 6 months, 12 months, and 18 months. Multiple
measurement points over an l8-month time period provides a detailed account of
incremental change in the constructs of interest, especially how they relate to program
participation. The structured interviews assessed a variety of personality, family, and
community characteristics such as extent of family involvement, school attitudes and
behavior, community involvement, attitudes and exposure to delinquency, self-
perceptions/self-esteem, attitudes toward deviance, stressful life events, neighborhood
problems, and self-reported delinquency including drug use. In most cases, multiple
indicators of each construct are then combined into factor scales.

This analysis also includes criminal history data provided by the County Juvenile
Court that details each juvenile’s referrals to the juvenile court for Court wardship,
neglect/abuse, truancy, and general violations of local or state criminal codes. Criminal
history data will be useful to both compare individuals in regard to prior official contact
with the criminal justice system before the individuals participated in the intervention
program and track the extent of recidivism after the commencement of program

involvement.

78

Criminal history ascertained from court referral data will be the source for one of
the major dependent variables, recidivism. Court referral is an indicator that reflects
multiple decision points in the juvenile justice system. First, police must discover an
apparent violation of the law. Second, police must recognize the behavior as sufficiently
serious to warrant official action and necessitate an arrest. Third, additional actors in the
local police agency (e. g., detectives, juvenile bureau) must also perceive the act
sufficiently serious and substantiated with available evidence to justify referral to the
County court system for further processing. Referral may be biased because they can be
influenced by police policy, victim cooperation, and the age and demeanor of the
perpetrator (Bittner, 1976; Wilson, 1968).

While the importance of these warnings is recognized, the occurrence of a referral
and the characteristics of a referral (e.g., type of crime) will be used an indicator of
recidivism. All criminal history data sources are potentially biased to some extent or
another by decision points that influence the characteristics of the population. For
example, court conviction data can be influenced by courtroom-specific working
relationships that affect the likelihood of conviction depending on available jail space,
extent of community-crime problems, and political agendas (Cole & Smith, 2001).

Finally, census data will be used to quantify ecological characteristics of the
neighborhoods in which individuals reside at the time of referral from instant offense.
Sociological characteristics have been long assumed to be important indicators of
criminality. Wikstrtim and Loeber (2000) argue there is reason to believe criminality
stems fiom a combination of individual characteristics and community factors. They

suggest a need to center analysis on the “person-context interaction.” Census data are

79

publicly available and aggregate indicators of socio-economic status to units of analysis
referred to as census block groups. In addition to socio-economic variables, census data
also include other measures of community disorganization.

Sample

The sample populations consist of juvenile offenders referred to “State”
Department of Social Services (DSS) by the local county juvenile court system. During
the mid-19903 the county juvenile court had the Option of referring juveniles directly to
the DSS for treatment if they determined the necessary treatment services were not
available through the local court system. This was generally reserved for the more
serious offenders such as younger offenders referred for serious offenses or older
offenders with several less serious prior offenses.

The intervention program (experimental group) was directed at male juveniles,
and the most serious offenders were excluded from the program. At the time of the
project, DSS utilized a four tier system that classified offenders to one of four security
level assignments: community-based, low, medium, or high that represented the nature of
intervention services to be provided to each youth (see Appendix A). Using a
combination of current offense(s) and calculated risk score, each level represents an
increasingly restrictive form of intervention with community-based representing the least
restrictive option. The “community-based” classification was the level that identified
placement in the YFS intervention program.

The most serious felonies such as attempted murder, first-degree rape, and armed
robbery were excluded because these offenders were, as would be expected, adjudicated

to secure placements due to the seriousness of the offenses. Eligible offenses did include

80

several felony violent, weapon, and property offense categories such as assault with a
dangerous weapon, break and enter with intent to commit to crime, and carrying
concealed weapon. In addition, the eligible offenses charges include less serious crimes
such as larceny, incorrigibility, truancy, and status offenses (see Appendix A for eligible
charges).

The YFS program established a formal placement mechanism to evaluate the
eligibility of each to participate in the program activities. The structured decision making
model considered two factors: (1) the type of the commitment offense, and (2) risk score
as determined by the Department of Social Services’ Initial Delinquency Risk
Assessment Scale” (see Appendix A). The risk assessment scale considered a variety of
important background variables such as age at first adjudication, number of prior arrests,
and characteristics of peer relationships that research has consistently identified as
important predictors of future delinquency. Juveniles were initially screened as “eligible”
if they were committed for all but the most serious offenses (e. g., homicide, attempted
murder, very serious assaults, armed robbery, and serious instances of criminal sexual
conduct). The risk assessment instrument constructed by the “State” Department of
Social Services and the National Council on Crime and Delinquency was also utilized to
ensure program participants did not have background problems so severe as to negate the
likelihood of effective nonresidential, community-based treatment. Finally, information
regarding the commitment offense and risk assessment were integrated into a security
matrix. Individuals scoring “low risk (0-8 points)” on the assessment who were

adjudicated for a crime other than the most serious should then have been eligible for

participation in the community-based program.

81

The sample for the present study includes an experimental group, a concurrent
control group, and a historical control group. The experimental group (n=135) includes
all youthful male offenders that were adjudicated between July 1, 1995 and June 30, 1996
to participate in a community-based intervention program in the County. Youth in the
community-based treatment program received 6 months of intensive program services in
a modified school setting and 6 months of aftercare treatment. The concurrent control
group (n=109) is a sample of similarly situated youth also adjudicated between July 1995
and July 1996 to traditional disposition such as residential placement. Individuals in the
concurrent control group were exposed to the same structured interviews as the
experimental group, yet were adjudicated to traditional probation services. Finally, the
historical control group (n=132) were adjudicated to traditional probation services one
year prior to the YFS program (July 1994 to June 1995).

Hypotheses

The literature review presented in Chapter 2 suggests that delinquency is a
product of the interaction of multiple domains of problem areas that are considered the
“risk factors” for delinquency. The identification of a sufficiently comprehensive model
that details the interrelationships between these problem areas provides the ability to
devise intervention efforts that make major strides in reducing future involvement in
delinquent activity. The literature presented leads us to expect the following outcomes if
the intervention efforts delivery the appropriate services in the necessary quantity:

H1: Members of the concurrent control group will exhibit more risk for
delinquency;

H2: T1 risk factors will predict recidivism and T3 self-reported delinquency;
H3: Decrease in risk factors will reduce recidivism and self-reported

delinquency;

82

H4: Increased treatment will reduce recidivism and self-reported delinquency.

Specification Of Variables Of Interest
Dependent Va_riaile_s

The primary outcome measures are both official recidivism 18 months after
referral for the instant offense and self-reported delinquency 12 months after referral.
Traditionally, criminological research has used official contacts with police or court
referrals as measures of crime and delinquency. The Uniform Crime Report is an
example of the aggregation of police data and represents official statistics collected by
police departments that indicate levels of crime in communities by reporting raw numbers
of police incidents (crimes) and arrests. Both aggregate police data and individual arrest
or court histories have been regularly used test a variety of theoretical propositions about
the relationships between individual, group, or societal processes and criminality.
Wolfgang, Figlio, and Sellin’s (1972) classic study of delinquency, for example, used

official court referrals as indicators of delinquent activity.

Oflicial Delinquency

In this instance, officially recorded authorized petitions as indicated in Court
records will be the first measure of recidivism. Court referral data do not necessarily
reflect every instance where an individual was officially detained by the police and
subsequently referred to juvenile court, but those occasions where an arrest and referral
were made and the Court substantiated the arrest to the point where charges were filed

against the individual. First, official delinquency will be operationalized as a binary

83

variable where the value 1 indicates the presence of a subsequent authorized petition and
0 indicates the lack of a subsequent referral within 18 months of the intake date.
Self-Report Delinquency

During the 19603 and 19703 criticism was leveled against the validity and
reliability of official statistics as measures of criminality. Hindelang (1974: 2) identified
seven problems associated with official statistics. Among the most notable
characteristics is the “dark figure of crime”17 (Sellin & Wolfgang, 1969). Additionally,
official police statistics can be biased because in some cases authorities fail to record all
crime reports made and definitions of crime vary from time to time.

Self-reported delinquency, a technique thought to be immune from the problems
of official statistics, was introduced during the 19503 and integrated into delinquency
research by Short and Nye (195 7) and has since become an important methodological
tool for delinquency researchers (Hindelang et al., 1981; Weis, 1988). The National
Crime Victimization Survey presents an example of efforts to move beyond traditional
measures of crime and victimization by using self-report measures. A series of self-
reported delinquency indicators were constructed that cover a range of reasonably serious
behaviors such as robbery and serious property crimes. The following indicators were
combined into a global measure of self-reported delinquency that weights18 each based on
different levels of seriousness:

1) Hit someone in anger (3);
2) Robbed someone (4);

 

'7 The “dark figure of crime” refers to the large percentage of crimes that are never reported, thus indicating
official crime statistics greatly underestimate the actual level of crime. For example, the Bureau of Justice
Statistics indicated of 1,000 serious crimes only 500 are ever reported to the police (U .S. Department of
Justice: Bureau of Justice Statistics, 1999).

'8 There is no inherent “distance” between these various indicators. The weights proposed in this coding
scheme are not proportional where the distance between 1 and 2, 2 and 3 have major substantive meaning.
They are used as a crude indication certain types of crimes are considered more serious than others.

84

Table 1. Scale Reliability Coefficients for Dependent Variable.
Scale Number of Items Time 3

Self-Reported Delinquency 5 .609

 

3) Stole something over $100 (3) ;
4) Stole something between $10 and $100 (2);
5) Damaged property that wasn’t yours (2);

Table 1 details the reliability alpha coefficients for the self-reported delinquency
scale. Reliability analysis measures the degree to which an indicator measures the
construct it is proposed to measure. The alpha value represents the amount of variance
shared among the indicators (Carmines & Zeller, 1979). Since only T3 self-reported
delinquency is considered as a dependent variable, Table 1 includes reliability diagnostics
for this time period only. The scales’ standardized item alpha is .609 at T3 which
indicates the scales are internally consistent and reliable.

Another aspect of this analysis will consider the risk factors defined below as
dependent variables. Intervention theory is based on the assumption that the treatment
will have a positive effect on the risk factors related to the onset of delinquent behavior.
Brewer et a1. (1995: 61) argue “risk and protective factors predict increased or decreased
probability of developing problem behaviors. . ..” In this situation, background variables

pertaining to the individual characteristics such as perceptions and behaviors, and social

characteristics such as exposure to delinquent peers will be considered as outcomes.

Independent Variables

The proposed study examines the effectiveness of a comprehensive community-

based intervention program in creating positive change in the life course of individuals

85

and thereby reducing the likelihood of future self-reported and official delinquency.
Research has demonstrated multiple risk factors to delinquency that exists at the
individual, family, and community levels. The structured interviews were constructed to
measure the link between these risk factors, pro gram treatment modalities, and

measurements of recidivism. The dimensions and related items are presented below.

Measures of Program Services

Intervention mapping (Bartholomew et al., 1998) is a process of structuring
responses to problems (interventions) based on the theoretical understanding of the causal
mechanisms underlying the issue. Social development theory is proposed as a theoretical
framework for understanding juvenile delinquency and implicitly suggests interventions
should be developmentally specific and address problems such as education, family
functioning, positive peer interactions, and life skills. In response, the YFS program was
designed to deliver education, job skills, counseling, and family functioning services to
program youth.

Youth in the experimental group and concurrent control group19 were asked to
identify three programs they participated in for at least 15 days during the T2, T3, and T4
interviews. After they identified the programs, the youth were asked the following
questions that were intended to quantify the extent to which they received certain types of
program services: “As part of the program, how often did you:

1.) Attend group meetings;
2.) Learn job skills (i.e., how to get a job, do the job, etc.);
3.) Have recreation time (i.e., play games, sports, etc);

4.) Attend academic classes (i.e., math, science, reading);
5.) Attend other classes (health, art, sex ed., law);

 

'9 Although youth in the concurrent control group received “traditional probation services,” in many cases
they also received similar services to the community-based experimental group.

86

6.) Attend special classes (social living skills, art);

7.) Work at the program for money;

8.) Attend individual counseling sessions;

9.) Attend family counseling sessions;

10.) Do volunteer work in the community;

11.) Receive drug/substance abuse counseling.”
The respondents were asked to record how often they participated in these programs, and
were provided with the following options and companion values: never (1), once or twice
a year (2), once every 2-3 months (3), once a month 9 (4), once every 2-3 weeks (5), once
a week (6), 2-3 times a week (7), once a day (8), and 2-3 times a day (9). “Never” was
recoded to reflect a value of zero, and the others were left in present coding scheme. The
items from the T3 interview were summed across programs, and then grouped into the
following four type of services: counseling (items 1, 8, 9, and 11), educational (items 4,
5, and 6), job skills (items 2, 7, and 10), and recreational (item 3). It is hypothesized that

individuals that receive greater amounts of treatment services will experience the largest

reductions in delinquency risk factors and also commit fewer crimes in the future.

Intake Placement Variables

Initial intake and placement variables are coded during the intake process by
personnel from the Family Independence Agency (see “INITIAL DELINQUENCY RISK
ASSESSMENT SCALE” in Appendix A).20 The placement instrument reflects issues
relating to age at first adjudication, number of prior arrests, current school status, history
of drug use, nature of current offense, prior probation commitments, number of out-of-
home placements, number of runaways from prior placements, last grade completed,
nature of parent/caretaker control, and nature of peer relationships. Juveniles in the

experimental and historical and concurrent control groups will be compared based on

 

2° Formerly the “State” Department of Social Services.

87

Table 2. Scale Reliability Coefficients for Independent Placement Intake
Variables.

 

 

 

Time
Scale Number of Items Intake
Intake Placement 12 .568

 

these individual intake factors and also on their total intake score. Total intake score is
hypothesized to be positively associated with recidivism. That is, juveniles with higher
risk scores are hypothesized to experience more instances of recidivism.

The reliability alpha scale score for Intake Placement Variables is presented in
Table 2. This data is only measured at intake and is used to compare background
characteristics of the experimental and two control groups. The alpha value of .568 is
slightly less than the value of .6 that is generally used as a standard cutoff (Carrnines &
Zeller, 1979). However, many of the intake variables have limited variability in their

responses, an issue that may limit its ability to maximize the explanation of variance.

Delinquency Risk Factors

MW Attitude Toward School. “Negative attitude toward school” is a
construct that reflects a lack of a sense of belongingness to their school and their
perceptions about levels of support from teachers and other concerned school personnel.
The construct is comprised of the indicators: (1) I can learn things at school; (2) I can’t be
successful in school; (3) I almost never expect to do well in the classes the school makes
me take; (4) The teachers and principals don’t want me in their school; (5) I get the
feeling that the school thinks I’m no good; (6) This school treats me like I’m dumb; (7)
I’m satisfied with the way I did in school; (8) I like my teachers. The scale was
summarized by computing the average scale score, a process that involved dividing the

total Stun of the scale by the number of items.

88

The responses are coded with a 7 item scale and in a manner that reflects negative
attitudes toward school. A value of 1 indicated they perceived the statement to be
“completely false” and 7 indicated the statement was “completely true.” Where
necessary (items 1, 7, and 8), responses were reverse coded so that they were
conceptually consistent with the other indicators. Attitude toward school is hypothesized
to be positively associated with recidivism. Meaning, juveniles with more negative
attitudes toward school will exhibit more recidivism. It is also expected that increased
program services will decrease negative attitudes toward school. Table 3 indicates that
the reliability measures are highly reliable at both T1 (.710) and T3 (.795).

Negative School Belrayior. The construct “negative school behavior” is an
indicator of the extent to which individuals participate in certain negative behaviors while
at school. The construct is comprised of the following indicators: Do something on
purpose that you knew would make a teacher angry or annoyed or interrupted class;
Smoke in or around school when you weren't supposed to or in a place where smoking
wasn't allowed; Cheat on a test; Leave the school grounds without permissions when you
weren't supposed to; Cut class without permission; Get into a serious fight with a student
at school; Skip a day of school without a real excuse; Damage or messed up school
property; Get suspended from school; Get sent to the principal’s office; Get expelled
from school; Get required to bring your parent(s) to school. The items were coded on a 9
item scale that included never21 (1), once or twice a year (2), once every 2-3 months (3),
once a month 9(4), once every 2-3 weeks (5), once a week (6), 2-3 times a week (7), once
a day (8), and 2-3 times a day (9). Individuals reporting higher levels of negative school

behavior are hypothesized to be more likely to recidivate than their counterparts. There

 

2’ Never was recorded to value of zero.

89

is also a hypothesized negative relationship program participation and negative school
behavior. Average scale scores were computed by dividing the total scale sum by the
number of items.

The 12 items that comprise this construct are highly reliable across time periods
(Table 3). The T3 alpha value (.902) is substantially higher than that of T1 (.782)
suggesting more internal consistency in the indicators of the former.

F_avgable Attitude Toward Delinquency. The construct “favorable attitude
toward delinquency” measures “how wrong” they perceive a series of acts to be for
someone their age. The scale includes the following indicators: Cheat on school tests;
Purposely damage or destroy property that did not belong to them; Use marijuana or
hashish; Steal something worth less than $5; Hit or threaten to hit someone without any
reason; Use alcohol; Break into a vehicle or building to steal something; Sell hard drugs
such as heroin, cocaine, and LSD; Steal something worth more than $50. Available
responses included very wrong (1), wrong (2), a little bit wrong (3), and not at all wrong
(4). It is hypothesized those that perceive delinquent acts to be acceptable would
participate in more delinquent acts. Thus, there is a hypothesized negative relationship
between attitudes toward delinquency and recidivism. Moreover, a positive relationship
is hypothesized between measures of program participation and attitudes favorable to
delinquency. That is, individuals exposed to more treatment services will perceive the
identified actions unfavorably. Average scale score were computed by dividing the total
scale sum by the number of items.

The reliability analyses for the nine items that represent attitude toward

delinquency also indicate the measures are highly reliable. Alpha values for the scales at

90

T1 (.765) and T3 (.851) suggest the indicators are internally consistent and represent good
measures of the same underlying construct.

Self-esteem. The “self-esteem” construct operationalized in this research is more
analogous to the global measure suggested by Rosenberg et al. (1989). It reflects
individuals’ perceptions of themselves and their level of personal acceptability. The
indicators include: “I accept myself the way I am,” “I am proud of the qualities I have,”
“I am an irresponsible person,” “I feel useless at times,” and “On the whole, I am
satisfied with myself.” Respondents were asked to indicate their level of agreement with
the questions with one of the following responses: (1), disagree (2), agree (3), and
strongly agree (4). The construct is coded in a manner that indicates positive self-esteem.
Items 2 and 4 were subsequently reverse coded to be consistent with this coding scheme.

Self-esteem is hypothesized to be negatively associated with recidivism. That is,
higher values of self-esteem are expected to result in lower measures of recidivism. It is
also hypothesized that levels of program services received will be positively associated
with self-esteem. The reliability analysis presented in Table 3 suggests the items are
internally consistent across T1 (.692) and T3 (.607).

Pro-socialfirlues. The construct “pro-social values” reflects the extent to which
individuals ascribe to conventional values such as the value of hard work and showing
respect to adults. Respondents were asked to respond to several items with the following
responses: strongly disagree (1), disagree (2), agree (3), and strongly agree (4). The scale
included the following four items: “Hard work is the only way to get ahead,” “It is
important to help people who are in trouble,” “You should always show respect to your

parents,” “It is wrong to deliberately hurt someone else.” Higher values indicate stronger

91

agreement with conventional, positive value systems. Pro-social values is hypothesized
to be negatively associated with measures of recidivism. It is also hypothesized that
measures of program participation are positively associated with pro—social values.

Reliability analysis for the four items detailed in Table 3 indicate higher internal
consistently at T3 (.658) compared to T1 (.587). Although the value of .587 is slightly
lower than the conventional cutoff of .6, it will be maintained in the analyses because it is
only slightly lower than expected.

Exposure to Delinquent Peers. The construct “exposure to delinquent peers”
measures the perceptions of the respondents as to how many of their close friends
participated in a series of delinquent acts during the previous 6 months. The indicators
included: cheated on school tests; purposely damaged or destroyed property that did not
belong to them; used marijuana or hashish; stolen something worth less than $5; hit or
threatened to hit someone without any reason; used alcohol; broken into a vehicle or
building to steal something; sold hard drugs such as heroin, cocaine, and LSD; stolen
something worth more than $50; suggested you do something that was against the law;
beat someone up; carried a gun; shot at someone; participated in gang activities; carried a
knife; stabbed someone. Responses were measured on a 5 item scale that included: none
of them (1), very few of them (2), some of them (3), most of them (4), and all of them
(5). Higher values indicate greater exposure to delinquent peers. Exposure to delinquent
peers is hypothesized to be positive associated with recidivism. A negative relationship
is hypothesized between measures of program participation and exposure to delinquent
peers. Average scale score were computed by dividing the total scale sum by the number

of items.

92

The reliability alpha values for exposure to delinquent peers is unusually high for
both T1 (.917) and T3 (.940). While the high values suggest the measures share a
substantial of amount of variance, the high values are also indicative of a scale that might
lack sufficient variation in indicators to represent the true complexity of the construct of
interest. An alpha value of l, for example, would suggest the indicators are exact
duplicates of one another and do not represent multi-dimensional construct. However,
visual inspection of the measures suggests they do in fact cover a wide range of behaviors
from violent and property crime to the sale and use of illegal drugs.

Positive Family Involvement. “Family involvement” indicates the extent to
which individuals participated in positive activities with members of their family or other
caregivers. Items asked how often during the past six months the respondents spent time
with family members or other care givers engaged in the following activities: “How often
did you spend time doing things with your family/caregiver?,” “How often did you spend
time with sisters or brothers?,” “How often did you spend time with other relatives?,”
“How often did you spend evenings at home?,” “How often did your family/caregiver
know what you were doing with your time?,” “How often did your family/caregiver talk
with you about school or work?,” and “How often did your family/caregiver and you do
fun things together?” Respondents answered with one of the following: never (1), once
or twice a year (2), once every 2-3 months (3), once a month 9(4), once every 2-3 weeks
(5), once a week (6), 2-3 times a week (7), once a day (8), and 2-3 times a day (9). The
value for “never” was subsequently recoded to zero.

Higher values on this scale are taken as indications of more opportunity to be

involved in positive activities with family members or other caregivers. It is

93

hypothesized that “positive family involvement” is negatively associated with measures
of recidivism. Moreover, aggregate amount of program services is expected to be
positively associated with this construct. The scale is reliability across T1 (.745) and T3
(.838) suggesting the items are internally consistent and share substantial amounts of
explained variance.

Alcohol and Drug Use. The construct “alcohol and drug use” summaries the
frequency of use of illegal substances. The items for this scale includes: alcoholic
beverages (beer, wine, hard liquor); marijuana/hashish (“grass”, “pot”, “hash”);
hallucinogens (LSD); amphetamines (speed); cocaine (crack). The items were coded on
a 9 item scale that included never (1), once or twice a year (2), once every 2-3 months
(3), once a month 9(4), once every 2-3 weeks (5), once a week (6), 2-3 times a week (7),
once a day (8), and 2-3 times a day (9). Alcohol and drug use is hypothesized to be
positively correlated with recidivism. Average scale score were computed by dividing
the total scale sum by the number of items.

Stressful Life Events. “Stressful life events” represent situations in the life course
of individuals that have the potential to cause significant levels of trauma and
subsequently relate to the escalation of delinquency. Hoffman and Cerbone (1999)
reported stressful life events were responsible for a significant intra-individual “growth”
in delinquency that is independent of age effects. The indicators are binary measures
with 1 indicating the respondent experienced the stressful life event during the previous 6
months and 0 that the event was not experienced. Measures include: fiiends getting into
gangs; family members getting on drugs; teachers hassling you; getting picked up by

police; friends getting on drugs; family having money problems; other students talking

94

Table 3. Scale Reliability Coefficients for Independent Variables.

 

 

Time

Scale Number of Items 1 3

Attitude Toward School 9 .710 .795
School Behavior 12 .782 .902
Attitude Toward Delinquency 9 .795 .851
Self-esteem 5 .692 .607
Pro-social Values 4 .5 87 .658
Exposure to Delinquent Peers 16 .917 .940
Positive Family Involvement 8 .745 .838
Drug/Alcohol Use 5 .597 .525
Stressful Life Events 31 .782 .900
Community Integration/Attachment 8 .623 .746

 

junk; police pointing guns at you; fiiends getting someone pregnant or getting pregnant;
family members getting killed; doing tests in school; going to court; friends getting
locked up; arguing among family members; being bored in school; getting locked up;
fiiends getting killed; family member not accepting your friends; following school rules;
getting caught in drug raids; a parent dying; brother or sister dying; close friend dying;
parents getting divorced or separated; failing one or more subjects in school; being
arrested by the police; flunking a grade in school; family member having trouble with
alcohol; getting into drugs; getting into alcohol; parent or relative in your family getting
very sick. A positive relationship between exposure to stressful life events and
recidivism is hypothesized. Average scale scores were computed by dividing the total
scale sum by the number of items. Table 3 suggests the scale that includes 31 indicators
of stressful life events was also highly reliable at both T1 (.782) and T3 (.900).
Community Involvement. The construct “community involvement” is a measure
of participation in positive community activities. The term “community” is used loosely
to refer to home, school, and neighborhood environments. Participation in community

activities was captured by asking respondents to indicate how often during a normal week

95

they participate in the following activities: “looking for work,” “doing chores,”
“attending classes,” “doing homework,” “organizing athletics at school, “organizing other
athletic activities,” “attending church or church related activities,” and “clubs or groups
away from school (boys/ girls club). Available responses were: never (1), once or twice a
year (2), once every 2-3 months (3), once a month 9(4), once every 2-3 weeks (5), once a
week (6), 2-3 times a week (7), once a day (8), and 2-3 times a day (9). The value for
“never” was recoded to zero. The reliability analysis in Table 3 indicates the items are
internally consistent across both T1 (.623) and T3 (.734).

Social Disorganization. Measures of “social disorganization” are taken from
1995 Tiger census block group files provided by the US. Department of Census. The
measure is available only for the experimental and concurrent control groups because
“home address” was not provided with the intake data for the historical control group.
The first step of this process included mapping each individuals’ home address that was
recorded at time of intake. After locating each individual in a particular census block
group, the following data elements were used as measures of social disorganization:
percent vacant housing; percent renter occupied housing, and percent of homes built
before 1940. It is hypothesized that individuals residing in highly disorganized
communities will exhibit more delinquent tendencies and will be less likely to respond
positively to treatment.

Measures of neighborhood social disorganization were only measured at time of
intake and are not hypothesized to have changed as a result of intervention efforts. The
purpose of this measure is to compare the three different groups of individuals based on

background characteristics, but it also serves as a control for outcome measures. The

96

Table 4. Scale Reliability Coefficients for Neighborhood-Level Independent
Variables.

 

 

 

Time
Scale Number of Items Intake
Social Disorganization 3 .640
Socio-economic Status 4 .743

 

alpha value of .640 detailed in Table 4 indicate the 3 measures are reliable measures of
the same construct.

The items were combined using principle components factor analysis. All three
indicators were hypothesized to measure the same underlying factor, social
disorganization. The principle components analysis revealed only 1 factor as
hypothesized, and all three items loaded consistently high on that factor. Table 5 presents
the factor loadings, eigenvalue, and percent explained variance for the analysis. The data
indicate the combined items explain nearly 60 percent of the variance in the social
disorganization construct.

Socio-economic Status of Neighborhood. Similar to measures of social
disorganization, neighborhood socio-economic status will also be taken from the 1995
Tiger file provided by the US. Department of Census. Similar to the measures of social
disorganization, the measures of socio-economic status are available only for the
experimental and concurrent control group because “home address” was not made
available for the historical control group. Measures of socio-economic status include
percent nonwhite residents, percent female headed household, percent low income,
percent unemployment, and average home value. It is also hypothesized that individuals
residing in more impoverished neighborhoods will experience more recidivism. Similar

to the social disorganization scale, neighborhood socio-economic status is used as an

97

Table 5. Exploratory Factor Analysis for Social Disorganization.

 

Item Factor 1
Percent vacant housing .764
Percent of residents renters .750
Percent of housing built before 1940 .790
Eigenvalue 1 .770
Percent of variance 59.00

 

explanatory variable and is not expected to change as a result of intervention efforts.
Table 4 suggests the 4 items included in the scale are highly reliable with an alpha value
of .743.

Indicators of socio-economic status were combined using principle components
factor analysis. All three indicators were hypothesized to measure the same underlying
factor, socio-economic status. The principle components analysis revealed only 1 factor
as hypothesized, and all three items loaded consistently high on that factor. Table 5
presents the factor loadings, eigenvalue, and percent explained variance for the analysis.
The data indicate the combined items explain nearly 60 percent of the variance in the

socio-economic status construct.

Expected Relationships

Direct Effects Of Intervention Treatment 0n Risk Factors

The YFS community-based intervention program was designed to reduce risk
factors associated with delinquency and simultaneously increase the presence of
protective factors. The conceptual model depicted in Figure 2 (see page 77) indicates the
impact of intervention treatment is hypothesized to have both a direct and indirect effect

on recidivism with indirect effects operating through risk and protective factors. This

98

Table 6. Exploratory Factor Analysis for Socio—economic status.

 

Item Factor 1
Percent of residents nonwhite .816
Percent of residents female headed household .912
Percent of residents low income .872
Percent of residents unemployed .864
Eigenvalue 3.003
Percent of variance 75.08

 

will be analyzed with paired samples t-tests. Since the YFS program was designed to
provide intervention treatment services in the areas of educational achievement, family
functioning, peer association, and community attachment, the first stage of the analysis

will assess if the treatment had any affect on these intended outcomes.

Direct Effects Of Intervention Treatment 0n Characteristics And Extent 0f Recidivism
The present research attempts to separate the direct and indirect effects of
intervention treatments on recidivism. More specifically, this research will measure if
individuals have records of additional court petitions for all types of crimes on an
aggregate basis, but also for specific types of crimes such as status offenses, property
offenses, and violent offenses. This is an important consideration because the juvenile
justice system has grown increasingly concerned about juvenile involvement in serious
crime, especially serious violent crime. As indicated in the introduction to this study, the
perception that juveniles have been increasingly involved in serious and violent crime
was responsible for the growth in “get tough” legislation directed at juvenile offenders.
This analysis will include not only official measures of delinquency as dependent
variables but also self-reported delinquency. A direct negative relationship is expected

between aggregate indicators of program participation and measures of official and self-

99

reported recidivism. The experimental group will be compared with both the concurrent
and historical control groups for measures of official delinquency and the concurrent

control group only for measures of self-reported delinquency.

Indirect Eflects 0f Intervention Treatment 0n Recidivism Through Changes In Risk
Factors

This research will also test if the background variables previously identified in
this manuscript mediate the effectiveness of community-based intervention treatment
services. Although the theory underlying intervention research suggests individuals are
generally better treated in the community as opposed to institutional treatment, there is
evidence treatment may be more or less successful depending on characteristics of the
offender. As discussed above, it is expected that the treatment effects will create positive
change in the domains general found to be underlying delinquent behavior. These
domains can be considered “risk factors” for delinquency. The treatment is expected to
reduce risk factors and increase protective factors. The hypothesis is that the treatment
effect will only reduce the characteristics and extent of recidivism of delinquency in as
much as it reduces the presence of risk factors and increases the presence of protective
factors. Similar relationships will be tested as in the section titled “Direct Effects Of
Intervention Treatment On Characteristics And Extent Of Recidivism.” In this case,
however, the analysis will determine how treatment effects influence the relationships
between risk factors and outcome measures.

Scaling Procedures

Three scaling procedures were used to construct factor scores for the measures of
self-reported delinquency and delinquency risk factors (attitude toward school, school

behavior, attitude toward delinquency, self-esteem, pro-social values, exposure to

100

delinquent peers, positive family functioning, alcohol and drug use, stressful life events,
community integration/attachment, social disorganization, and socio-economic status).
The self-reported delinquency scale was determined by using the coding scheme
identified on page 86. Respondents were asked how many times during the previous 12
months they participated in a series of delinquent acts that included measures of violent,
property, and drug crimes. Each of the indicators was assigned a weight that indicates a
level of seriousness. Although the values assigned with the weighting scheme do not
create proportional distance between measures based on objective criteria, it makes a step
in creating a seriousness hierarchy among the indicators. Each indicator weight was then
multiplied by the value provided by the respondents with regards to the number of times
they participated in the actions over the previous 12 months. A cumulative scale score
was computed by summing the values. The original distribution was negatively skewed
so the log of the cumulative scale score was be computed to normalize the measure. The
summary value was computed for T1 and T3 measures.

Except for measures of social disorganization and socio-economic status, scale
scores for the delinquency risk factors were computed by dividing the sum of the items in
the scale and dividing by the total number of items. Finally, principle components factor
analysis was used to compute the factor score for the measures of social disorganization
and socio-economic status.

Analysis Plan

Respondents were interviewed by trained staff that administered the same series

of instruments to each respondent at initial intake, and after 6 months, 12 months, and 18

months. The instrument includes theoretical variables that are both important to the onset

101

of delinquency and also important to the occurrence of recidivism. Scales were
constructed that operationalize constructs such as attitude toward school, self-esteem, and

problem behavior at school in ways that are consistent with previous research.

Comparing Measures of Centrallendency

The first step in the analysis plan will involve descriptive analysis that compares
important characteristics of the experimental group, concurrent control group, and
historical control group. The groups will be compared to determine how similar they are
in regards to basic demographic characteristics such as age, race, ethnicity, and
neighborhood characteristics. Neighborhood characteristics are limited to measures of
neighborhood social disorganization and socio-economic status. The descriptive analyses
will involve contingency table analyses and measures of association. In most cases, the
one-way analysis of variance (AN OVA) test will be used to compare the means of the
three groups. Similar to significance tests comparing two groups (e. g., t-test), ANOVA
uses the F distribution for detecting differences among means in situations where there
are more than two groups (Agresti & Finlay, 1997).

The experimental group, concurrent control group, and historical control group
will also be compared on their total score recorded on the “State” Department of Social
Services’ risk assessment classification scheme (Appendix A). Referring to section titled
“Initial Delinquency Risk Assessment Scale” compiles a total risk score based on issues
such as age at first adjudication, number of prior offenses, current school status, and
history of drug use. Individuals scoring between 0-8 are classified as “low risk,” 9-11 as
“moderate risk,” and those 12-19 as “high risk.” This information is used in conjunction

with the severity of the most serious offense for which the individual were adjudicated to

102

determine the overall security level assignment (see section titled “Initial Security
Matrix”). Security level assignments fall on a continuum that includes “high,”
“medium,” “low,” and “community based.” Those classified as “community based” were
eligible for participation in the YFS program.

Descriptive analysis will also be used to compare characteristics of the risk factors
and self-reported delinquency reported in the structured interviews for the experimental
and concurrent control group. The purpose of this research is to determine if
comprehensive community-based alternatives represent more effective intervention
strategies for serious adjudicated youth, and if so, why. One of the first steps then is to
determine how similar they are based on aspects research and experience indicate are
important to program success.

The experimental and concurrent control groups will also be compared in regards
to the amount of services received during the program year. The extent of program
services will be added across the possible three different programs identified by the
respondents and will be grouped into the following categories: education, job skills,
counseling, and family functioning services. These comparisons will be made using t-test
procedures.

Finally, the experimental and concurrent control groups will be compared based
on changes in between T1 and T3 delinquency risk factors. Change scores will be
computed by subtracting T3 measures from T1. For negative attitudes toward school,
negative school behavior, favorable attitude toward delinquency, exposure to delinquent
peers, alcohol/drug use, and stressful life events decreases from T1 to T3 represent a

decrease in risk, changes considered favorable. Similarly, increases in self-esteem,

103

prosocial values, positive family involvement, and community involvement are also
considered favorable and indicate increases in protective factors. T-tests will be used to

measure the significance of the change scores for the two groups.

Logistic Regression

Logistic regression is the appropriate model to use in cases where the variance in
the dependent variable is limited to dichotomous, binary values (0/1). In this case, the
value of the dependent variable is no longer of intrinsic interest but the “odds” or
probability that one or the other category of the dependent variable can be predicted
based on combined effects of the independent variables (Menard, 1995). In contrast to
the OLS parameter estimation used in linear regression, logistic regression utilizes the
maximum likelihood technique to maximize the value of a function. Evaluation of the
logistic regression model typically involves answering the questions: First, Can the
combination of variables used to predict the dependent variable do so in a manner above
and beyond what would be expected by chance alone? Second, if the model works, how
important are each independent variable? Third, does the form of the model appear to be
correct? (Menard, 1995: 17).

Logistic regression will be used to test the effects of demographic and
delinquency risk factors on recidivism for any criminal offense with 18 months of intake.
Several different logistic regression models will be presented. The first analysis will
combine all three samples (experimental, concru'rent control, and historical control
groups) in a single model and will include demographic characteristics, age at arrest for
instant offense, instant offense, and placement variables as independent or explanatory

variables. A dummy variable will be created to denote group membership. In this case,

104

“experimental” will be coded with a value of 1 and the concurrent and historical control
groups with a value of 0. The historical and concurrent control groups will be combined
because there is no substantive reason to hypothesize members of the group should
respond any differently to traditional probation services. Inclusion of the historical
control group provides the added value of increasing the sample size for greater
explanatory power. Three binary dummy variables will also be constructed for instant
offense that reflect if the offense was for a violent crime, property crime, or other type of
offense.

The second analysis will use data from the experimental and concurrent control
groups only. This model will include the same variables indicated above in a stepwise
fashion, and will also incorporate T1 delinquent risk factors from the structured
interviews.22 Change scores for the delinquency risk factors will be stepped into the

regression model after the T1 risk factors are entered.

Within Subiects Change

It is hypothesized that a comprehensive community-based program design based
on social development theory will be more effective in reducing delinquency risk factors
than traditional probation services. If the YFS program better conceptualizes the
mechanisms that “produce” delinquency and delivers the appropriate intervention
services to offset these processes, it is expected juveniles adjudicated to the YFS program
will experience greater decreases in delinquency risk factors compared to those given

traditional probation services.

 

22 The historical control group is excluded from this analysis because they were not subject to the structured
interviews.

105

The within subjects repeated measures design is the traditional statistical method
used for determining the extent to which subjects changed as a result of a particular
intervention. This method is commonly used, for example, in the situation where two or
more groups (e.g., experimental group and control group) are exposed to a treatment
condition (e. g., new drug versus placebo). In this scenario, subjects would be tested and
then retested on certain characteristics (e. g., presence of a particular disease condition) to
determine if the drug had an effect on the characteristic of interest.

A modified form of a within subjects repeated measures was used in this research.
The primary concern was to determine if change in risk factors effect likelihood of
recidivism and levels of self-reported delinquency at T3. Instead of considering “change”
as a dependent variable as is usually the case in repeated measures design, delinquency
risk factor change scores were independent variables. The formula used to calculate the
extent of change was: XiT3-Xm where “X” represents individual respondents, “i”
represents each delinquency risk factor, and T1 and T3 represent the time periods of
interest (T1=initial risk; T3=12 months). Thus, change will be assessed between the
structured interviews administered at intake and after the 12 month follow up. A 12
month follow-up period was used to allow for adequate time to have passed for the
intervention effects to take bold but not so much time that the intervention effects
dissipated. Higher values would be considered increases in risk and decreases as
reductions in risk.

While useful for the current analysis, the within subjects designs, repeated

measures or otherwise, are not without potential disadvantages. The principal problems

are with systematic changes caused by practice effects and diflerential carryover eflects

106

(Keppel, 1991: 334-343). Practice effects are systematic changes that result fiom the
process of taking the test (in this case, structured interview) itself. For example, the
repeated process of delivering the same structured interview itself might sensitize
respondents to certain answers. While there is always a concern with the lack of
truthfulness during structured interviews, especially when asked about sensitive subjects
such as self-reported delinquency, practice effects are not perceived to be a serious
problem in the current situation. First, the interviews are delivered in 6 months intervals
thereby reducing practice effects that might exist if given tests in close temporal
proximity. Second, the nature of the questions are not such as to suggest a “right” or
“wrong” answers that would be evident from taking the same “test” on multiple
occasions.

Diflerential carryover effects also present a situation where a subject’s
participation in the administration of an earlier treatment influences the subject’s
performance in the current treatment. In this case, the concern is that subjects might have
been provided earlier intervention services by the Court system or “State” Family
Independence Agency (F IA) that may systematically predispose them to differential
outcomes. Although a potentially serious problem, there is little ability to control for it in
the current analyses since data on prior family involvement with F IA or other related
agencies is not available. “Number of prior arrests” can be used as a conservative proxy
for differential carryover effects since it would suggest some level of court involvement
in the past. It is assumed that prior court involvement is indicative of some level of

official intervention. However, this measure is less than perfect for several reasons. The

107

most glaring is that it excludes prior services delivered other non-court agencies such as

FIA.

108

CHAPTER FOUR: FINDINGS

The purpose of this research was to assess the efficacy of community-based
intervention services for adjudicated juvenile offenders. Using a quasi-experimental
design, the proposed research questions test the extent to which community-based
programs were more effective in creating positive change in delinquency risk factors and
reducing recidivism than traditional probation services. The research design compared a
sample of 135 experimental youth assigned to a community-based intervention program
between July 1, 1995 and June 30, 1996, a concurrent control sample of 109 juveniles
assigned traditional residential probation services during the same period, and an
historical sample of 132 youth processed through the same Court during the previous
year (July 31, l994-June 30, 1995). There were a total of 376 individuals in the three
groups.

The three study populations were compared on several dimensions including
demographic characteristics" such as race, age at instant offense, and offense type for
instant offense. The three sample populations were also compared on placement
variables coding as part of the initial placement risk assessment (see Appendix A).
Following comparisons of background characteristics for the three sample populations,
logistic regression analysis was used to make an initial assessment of the relative merits
of community-based versus residential treatment programs when controlling for
demographic and placement variables. After baseline assessments were made about the
relative merits of each treatment modality, a series of statistical models will be presented

that compare characteristics of the experimental and concurrent control groups. These

 

2’ Gender is excluded from the analysis because the sample population was restricted to male offenders.

109

Table 7. Descriptive Information by Study Sample Groups (n=376).

 

 

1:53:1- Concurrent Historical Total xz df Sig.
n % n % n % n %
Race
Black 88 65% 77 71% 97 74% 262 70% 2.243 2 .326
Non-black 47 35% 32 29% 35 27% 1 14 30%
Total 135 100% 109 100% 132 100% 376 100%
Offense”
Personal 29 22% 31 29% 30 23% 90 24% 10.790 4 .029
Property 63 48% 62 58% 72 55% 197 53%
Other 39 30% 13 12% 30 23% 82 22%
Total 131 100% 106 100% 132 100% 369 100%

 

T Significant at p s .05.
' Intake placement data missing for 7 individuals.

two were subject to the same series of structured interviews that will establish a
framework for drawing comparisons on the initial delinquent risk factors (T1), changes in
risk factors during the intervention period, and how both relate to recidivism measures.
Theoretically, it is expected that juveniles who are at greater “risk” for delinquency as
indicated with the T1 risk scores will experience higher levels of recidivism. Similarly,
juveniles reporting positive changes in delinquency risk factors during the intervention
period are also expected to experience less recidivism.
Sample Characteristics

The sample population is comprised primarily of minority youth. Race was
dichotomized into the categories “black” and “non-black?“ Although it would have
been useful to break down racial characteristics into more discrete categories, the small
numbers in other categories required the data to be combined (see Table 7).

Approximately two-thirds of the entire population (11 = 262) were African American. The

 

2’ Disaggregated racial characteristics for entire sample population: Black (n=262), White (106), Hispanic
(5), and Other (3).

110

Table 8. Mean Intake Age by Group Membership (n = 369).'

Mean

 

Experimental Concurrent Historical Total d.f. Sq F Sig.
Age
Mean 15.39 15.22 15.28 15.29 1 1-561 1-424 .233
Std. Dev 1.096 1.087 1.00 1.048
Range 13-17 13-17 13-17 13-17

 

a Intake placement data missing for 7 individuals.

experimental group included the largest percentage of non-black participants (35 %) yet
the mean differences between the populations were not significant.25

Table 7 also presents characteristics of the referral intake offense. Intake offense
was categorized into three groups and represents the criminal charge for which
individuals were referred to the Court for the instant offense. In the case where multiple
charges were filed against an individual, the instant offense was the most serious
substantiated charge.26 “Personal crimes” were assaultive crimes such as murder, rape,
robbery, and assault. “Property crimes” included among others, burglary, larceny,
receiving and concealing stolen property, and malicious destruction of property. The
group “weapons, status offense, other” was a very broad category that included crime
types that did not fall into the previous.27 Fifty three percent (n=197) of the population

were referred for property offenses, 24% (n=90) for personal crimes, and 22% for other

 

2’ The breakdowns for the “non-black” categories are as follows for each sample population: Experimental
group (White: n=44, 30 percent; Hispanic: n=1, 1 percent, Other: n=2, 2 percent), Concurrent control group
(White: n=29, 27 percent; Hispanic: n=3, 3 percent; Other: n=0), and Historical control group (White:

=33, 25 percent; Hispanic: n=1, 1 percent; Other: n=1, 1 percent).
26 See Table B-1 in Appendix B for more specific details of intake offense categories.
27 See Hindelang, Hirschi, and Weis (1991) for concerns with scales that include wide continuums of
behaviors. “The trivial items in self-report scales swamp more serious items when, as is common, global
simple sum scales are used. To the extent that the correlates of serious items differ from the correlates of
trivial items, global scales will reflect the correlates of trivial delinquency. Similarly, to the extent that the
correlates of certain types of delinquency differ, global scales will match these differences” (Hindelang et
al., 1991: 90). Although these comments refer specially to self-reported delinquency scales, the sentiments
expressed are directly applicable to official delinquency scales. However, since recidivism in a juvenile
population is a rare occurrence in the short term, this type of method will be used.

111

offenses. Approximately 44% (n=66) of the “other” category consists of weapon
offenses such as carrying concealed weapon and 33% (n=51) for status offenses such as
truancy, runaway, and incorrigibility. The remaining 24 percent (n=37) represents a wide
range of less serious criminal offenses.

The experimental, concurrent, and historical control groups did not differ
significantly in regards to mean age at referral for instant offense. The mean age for the
entire population was 15.33, a value that was similar for all three populations (Table 8).
Intake placement data were missing for seven individuals who were subsequently
excluded from relevant analyses.

The relationship between age at intake for instant offense and offense type for the
full sample (n=369) is presented in Table 9. A larger percentage of juveniles age 13 (33
percent) were referred for violent crime compared to juveniles age 17 (21 percent).
Conversely, a smaller percentage of older offenders were referred for crimes in the “other
category” keeping in mind that the category “other” is composed primarily of weapon
related offenses such as carrying a concealed weapon and status offenses such as truancy
and curfew violations. It was expected that younger offenders might be more likely to be
referred for more serious offenses since they were referred to the Department of Social
Services precisely because they were considered more serious offenders.

Referring to the column percentages in Table 9, it is also evident that there is not
a substantial level of variation in the age composition for the separate crime types. For
example, only eleven percent of all individuals referred for personal crimes were
juveniles age 13, a figure nearly identical for juveniles referred for “other” offenses.

Although the percentage of juveniles age 13 for property offenses was slightly smaller at

112

Table 9. Age At Intake for Instant Offense and Offense Type (n=369)‘.

 

 

Age Personal Property S rim/33,31,” Total
13 Count 10 9 1 1 30
Row % 33% 30% 37% 100%
Col % 1 1% 5% 13% 8%
14 Count 13 35 10 58
Row % 22% 60% 17% 100%
Col % 14% 18% 12% 16%
15 Count 21 36 23 80
Row % 26% 45% 29% 100%
Col % 23% 18% 28% 22%
16 Count 37 92 30 159
Row % 23% 58% 19% 100%
Col % 41% 47% 37% 43%
17 Count 9 25 8 42
Row % 21% 60% 19% 100%
Col% 10% 13% 10% 11%
Total Count 90 197 82 369
Row % 24% 53% 22% 100%
Col % 100% 100% 100% 100%
Model Statistics
1:2 12.963
df 8
Sig. .1 13

 

3 Seven excluded due to missing data.

5 percent, the differences among the three figures amounts to very little. The model
statistics indicated no significant differences in the mean percentages across the
measures.

Table 10 presents similar data yet disaggregates the information by population
group type. There are a few differences evident in this analysis. As indicated by
previous analyses, offenders in the experimental group are generally less serious
offenders. The most important differences between the experimental and control groups

is among younger offenders. Younger offenders in the experimental group are

113

 

 

 

 

 

$2: $2: $2: $2: $2: $2: $2: $2: $2: $2: $2: $2: $ .00
$2: $3. $wm $3 $2: $3 $3 $3 $2: $2” $3 $33 $ 33.
2: m . No 3 an. cm 3 cm 3 . am 3 3 E30 .35.
$w $o $2 $3 $3 $2 $3 $2 $2 $2 $3 $5 $ .00
$2: $0 $3 $3 $2: $3 $3 $3 $2: $3 $8 $2 $ 33:
m o o m m: v o. v m : v a 3 “550 3
$2. $. m $36 $3 $9 $3 $9. $3 $3 $3 $24 $2.. $ .oU
$2: $2 $39 $3 $2: $2 $6 $3 $2: $3 $~m $3 $ 33.
me e 3 3. ..m .. mm .3 co m: 3 3 Esau 3
$3 $. m $3 $3 $3 $2” $w. $3 $3 $3 $3 $3 $ .00
$2: $0. $3 $3 $2: $3.. $3. $3 $2: $3 $3 $3 $ 33.
3 v 3 h 3 a m. c 3 o. a w .880 m:
$3 $3 $3 $3 $2 $0 $3 $m. $3 $m. $2 $3 $.60
$2: $5. $3 $3 $2: $o $3 $3 $2: $3 $3 $3 $ 33.
3 v m. o 3 o o: w 3 o 3. m “goo 3
$.. $m $m $0. $3 $3 $0 $2 $0 $3 $m $5 $ .00
$2: $3 $3 $3 $2: $2. $3 $3 $2: $3 $3 $3 $ 33:
N. _ m m m . c v m w v 3 3 .550 m.
.850 .856 .856
.55. .333 .395 .2398: .33. 335m deem .283.— .3:.—. 2.3% 395 .3998.— 0?.
>393? E333 E333
E25950 33.835 355.53”.—

 

.mseumv via—am .3 «Santa: 22H. 8:0qu 3:. 85:0 «:32: ..e.— 8.3:. 2 om< .3 «Sun.

114

Table 11. Comparison of Neighborhood Characteristics (n=178).‘II

 

x 32:: $2.222. F

Social Disorganization 3.370 4.529 0.035
Concurrent 0.467 0.867 0.098 -1.237 2.871
Experimental 0.190 0.859 0.086 -1.343 2.885
Total 0.313 0.871 0.065 -1.343 2.885

Socio—economic Status 2.598 3.367 0.068
Concurrent 0.635 0.960 0.108 -1.228 2.380
Experimental 0.392 0.808 0.081 -1.224 2.303
Total 0.500 0.884 0.066 -1.228 2.380

 

" Analysis restricted to experimental (n=99) and concurrent connol groups (n=79).

substantially less serious offenders than those in the control. For example, only 25%
(n=2) of juveniles age 13 and 14% of those age 14 in the experimental group were
referred for a personal. This is substantially lower than both the historical control group
(33% and 29% respectively) and concurrent control group (43% and 26% respectively).
Moreover, 29% of those age 13 in the experimental group were referred for a crime in the
“other” category, a category dominated heavily by status and other low level offenses, yet
this compares to 17 % for the concurrent control group and none for the historical control
group. Thus, regardless of factors associated with age at referral, those in the
experimental group continue to appear to be less serious offenders.

The experimental and concurrent control groups were also compared on
neighborhood characteristics such as levels of social disorganization and socio-economic
status.28 It is difficult to establish effective institutions in areas characterized by high
levels of social disorganization and poverty (Komhauser, 1978). This is particularly the
case for important institutions such as schools (Welsh et al., 2000). Not only are chronic

juvenile offenders more likely to come from high crime communities (Thomberry et al.,

 

28 This analysis was restricted to the experimental and concurrent control groups because “home address at
arrest” was available for these individuals only.

115

1995), but children living in high poverty communities are reported to experience more
court referrals (F arrington, 1995).

Neighborhood characteristics were operationalized from 1990 census tract data
provided by the United States Census Bureau. The construct “social disorganization”
was operationalized by percent vacant housing, percent renter occupied housing, and
percent of housing built before 1940. Similarly, “socio-economic status” is an index
variable comprised of percent of population nonwhite, percent female headed households,
percent of persons low income, percent of population unemployed, and average home
value. Higher values on both scales indicate greater risks associated with living in such
neighborhoods. The values for both indicators were computed using varimax rotated
exploratory factor analysis. According to Table 11, the social disorganization group
mean for the concurrent control group was significantly higher than the experimental
group. The concurrent control group also experienced a larger group mean on the
socioeconomic status measure. Although not statistically significant, the concurrent
control group seems to come from more highly disorganized neighborhoods and
neighborhoods characterized by higher levels of poverty.

Comparisons of Placement Intake Data

Placement data were collected as part of the Court intake procedures for juvenile
offenders and was coded by case workers employed by the Department of Social
Services.29 Placement data were intended to capture important aspects of individuals’
background as a means of determining placement categories. At that time, the placement
data were used as a strategy for determining placement levels. Except where overrides

were made, individuals who scored a total of 8 or less were assigned to the Youth and

 

29 See Appendix A for a duplication of the Intake Placement information.

116

Family Study community-based intervention program. The remainder of the individuals
classified as “low” were considered the concurrent control group for this analysis.

Tables 12a and 12b present the descriptive findings for each placement variable
(n=3 69). The tables present the mean scores and other aspects of central tendency
measures for all three samples. The variables were categorical without any inherent
sense of distance between the measures. In many cases, the variables are truncated to
only two available options and the limited operational definitions can make it difficult to
determine actual between-group differences.

Age at first adjudication was an ordinal variable that denotes the age range when
the individual was first referred for a criminal violation.30 The variable ranges from 0-3
with higher values reflecting younger ages. Juveniles first involved in the juvenile justice
system at younger ages were considered more serious because such behavior is
developmentally out of place (see Steffensmeier et al., 1989). Delinquency, even
moderately serious delinquency, is considered “norma ” among slightly older adolescents
(Hindelang et al., 1981; Lipsey & Derzon, 1998). The concurrent control group scored
the highest value (f = 1.45) and the historical the lowest (J? = 1.22 ). The three group
means were not significantly different (d.f.=2, F =2.524, p.>.05). There were also no

’9 ‘6

significant differences for “seriousness of current offense, probation at time of

’9‘

offense, ‘prior FIA placement,” and “level of parental commitment.”

,9 6‘

Group means were significantly different for “prior arrest history, school

99 “

status,” “drug use,” “prior placement, runaway,” “last grade completed,” and nature of
“peer relations.” Prior arrest history is an ordinal variable that assigns a value of “0” if an

individual had no prior, “1” if there were 1-2 priors and “2 if there were more than 2 prior

 

3° See “Variable Coding Table” in Appendix D for further details on dependent and independent variables.

117

Table 12a. Descriptive Characteristics of Placement Intake Variables (n=369)‘.

 

 

Placement — Std. Std. . Mean .
Variables X Dev. Error Mm Max dJ’ Sq. F Slg
Age 1st Adjudication 2 1.827 2.524 0.082
Hist 1.220 0.868 0.076 0 3
Concurr 1.453 0.806 0.078 0 3
Exper 1.250 0.868 0.076 0 3
Total 1.297 0.854 0.044 0 3
PriorArrest 2 2.239 4.196 0.016
Hist 0.970 0.741 0.064 0 2
Concurr 1.245 0.715 0.069 0 2
Exper 1.083 0.731 0.064 0 2
Total 1.089 0.736 0.038 0 2
School History 2 5.229 8.572 0.000
Hist 1.318 0.894 0.078 0 2
Concurr 1.736 0.652 0.063 0 2
Exper 1.553 0.755 0.066 0 2
Total 1.522 0.797 0.041 0 2
Drug Use
Hist 0.273 0.464 0.040 0 2 2 1.331 5.772 0.003
Concurr 0.481 0.502 0.049 0 1
Exper 0.333 0.473 0.041 0 1
Total 0.354 0.485 0.025 0 2
Offense
Seriousness 2 1.383 1.685 0.187
Hist 1.477 0.878 0.076 0 2
Concurr 1.491 0.876 0.085 0 2
Exper 1.303 0.957 0.083 0 2
Total 1.419 0.908 0.047 0 2
Prob at Time of
Arrest 2 0.290 1.185 0.307
Hist 0.379 0.487 0.042 0 1
Concurr 0.477 0.502 0.048 0 1
Exper 0.430 0.497 0.043 0 1
Total 0.426 0.495 0.026 0 1
Prior Placement 2 0.553 2.543 0.080
Hist 0.303 0.478 0.042 0 2
Concurr 0.387 0.489 0.048 0 1
Exper 0.250 0.435 0.038 0 1
Total 0.308 0.468 0.024 0 2
Runaway 2 0.289 3.288 0.038
Hist 0.083 0.304 0.026 0 2
Concurr 0.151 0.360 0.035 0 l
Exper 0.053 0.225 0.020 0 1
Total 0.092 0.298 0.016 0 2

 

' Seven cases excluded due to missing data.

118

Table 12b. Descriptive Characteristics of Placement Intake Variables (n=369)‘.

 

 

9:323: g 18;: Std. Error Min Max d.f. "g?“ F Sig
Last Grade
Completed 2 1.144 3.372 0.035
Hist 1.568 0.644 0.056 0 2
Concurr 1.764 0.526 0.051 0 2
Exper 1.674 0.559 0.049 0 2
Total 1.662 0.586 0.030 0 2
Parental
Commitment 2 0.309 2.193 0.113
Hist 0.115 0.320 0.028 0 l
Concurr 0.217 0.437 0.042 0 2
Exper 0.167 0.374 0.033 0 1
Total 0.163 0.377 0.020 0 2
PeerRelations 2 3.858 5.546 0.004
Hist 2.290 0.873 0.076 0 3
Concurr 2.557 0.731 0.071 0 3
Exper 2.205 0.871 0.076 0 3
Total 2.336 0.844 0.044 0 3
Total RiskScore 2 128.629 21.263 0.000
Hist 9.992 2.567 0.223 0 l6
Concurr 11.972 2.573 0.250 4 17
Exper 10.311 2.248 0.196 4 16
Total 10.673 2.591 0.135 0 17

 

' Seven cases excluded due to missing data.

arrests. Although the mean value for the historical control and experimental groups were
similar, the concurrent control group was significantly higher (d.f.=2, F=2.236, p.<.05)
on prior arrest history, a relationship driven by the substantially higher value for the
concurrent control group. “Current school status” measures the extent to which the
individual regularly attended school. A value of “0” was assigned for individuals who
regularly attended, “1” for those who dropped out, and “2” for those individuals who
were expelled or were chronically truant. The three groups were significantly different
(d.f.=2, F =8.572, p<.05) with the highest value reported for concurrent control group

(J? = 1.74) and the lowest for the historical control group (J? = 1.132 ).

119

The groups were also significantly (d.f.=2, F=1.321, p<.05) different on the
indicators of drug use. “Drug use” is a binary variable where a value of “0” indicates
either no use or only experimental use and a value of “1” indicates regular use of illegal
substances. The highest value was reported for the concurrent control group (J? = .481)
and the lowest for the experimental group (J? = .333 ). “Runaway” is also a binary coded
variable with a value of “0” indicative of no prior runaways from F IA placement and “l”
reflecting one or more occasions of running away. Similar to previous findings, the
highest value was reported for the concurrent control group (J? = .151) and the lowest for
the experimental group (J? = .053 ). “Last grade completed” is coded where higher
values represent a lower grade last completed. Individuals were assigned a value of “0”
if they last completed tenth grade, “1” for ninth grade, and “2” for eighth grade. The
coding was based on research findings that low school achievement is a risk factor for
current and fiJture delinquency (Denno, 1990; Farrington, 1989; Maguin et al., 1995).
The concurrent control group was significantly (d.f.=2, F=1.14, p<.05) higher than both
the historical control (3? = 1.568) or experimental groups (a? = 1.674 ). “Peer relations”
characterizes the nature of relationships between offenders and their peers. Higher values
were assigned to individuals with delinquents among their close associates. For example,
the value “3” was assigned when the evidence suggested the majority of close associates
were gang members or involved in other forms of delinquency. The experimental and
control groups were significantly different (d.f. = 2, F=5.55, sig. <.05) on this measure.
Finally, the variable “total risk score” reflects the cumulative placement score as
determined by the values of the previous 11 indicators. The study populations were

significantly different (d.f.=2, F=128.63, p<.05) with the highest mean value for the

120

concurrent control group (J? = 11.972) and the lowest for the historical control group
(J? = 9.992 ).

These placement variables were used to differentiate criminal offenders on
characteristics that theory and practice indicate are important to delinquency careers. In a
sense, review of the entire scale is consistent with the theories of delinquency identified
earlier. Social development theory, for example, stipulates that delinquency is a product
of the co—occurrence of multiple delinquency risk factors such as drug use, poor family
functioning, negative peer relations, and academic problems. The overall conclusion
reached was that in many ways the historical and experimental groups “looked” similar.
Moreover, the concurrent control group included slightly more serious offenders yet these
differences were not unexpected. The historical control group was defined based on the
classification of “community-based.” Since the determination of “community-based”
was contingent on the total placement risk score it was expected that these individuals
would be similar to the experimental group which were also, by definition, community-
based. It was expected that the concurrent control group would also be considered to be
more serious than the other two groups because, again by definition, these were offenders
too serious to qualify for community-based intervention services. While these data
revealed some important differences between the groups, the expectations were that the
differences between the concurrent control group and the experimental and historical
control group would actually be greater. Regardless, the availability of these data was
important because it provides the opportunity to control for these basic differences and to

determine if they are important factors in the production of recidivism.

121

Table 13a. Correlation Matrix: Placement Variables (n = 369).

 

 

X1 X2 X3 X4
Age 1St Adjudication (X1)
Prior Arrest(X2) 0179*
School Status(X3) -0.039 0085*
Drug Use(X4) ~0.146** 0.037 0.163*
Offense Seriousness(X5) 0.011 0.165* 0.160" 0.121"
Prob. Time of Offense(X6) 0.204" 0.339M 0.000 -0.l68**
Prior Placement(X7) -0.032 0. 155” -0.039 -0.03 1
Runaway Status(X8) 0.017 0089* 0.026 0.142"
Last Grade(X9) 0.342M 0.033 0.170" -0.055
Parental Commitment(X10) -0.037 0.022 0.052 -0.012
Peer Relations(X11) -0.025 0.153" 0.278" 0.211"
Total Risk Score(X12) 0.400" 0.539" 0.476” 0.246"

 

* Correlation significant at the 0.05 level (2-tailed).
** Correlation significant at the 0.01 level (2-tailed).

Tables 13a and 13b present correlation matrices of the 11 variables that comprise
the cumulative total placement score and the total placement score itself.31 As expected,
the total placement score was significantly correlated with all of its component measures.
The bivariate correlations were highest between total risk score and for prior arrest record
(r=.539), and total risk score and offense seriousness (r=.557). Although significant
(p<.05), the smallest bivariate correlations for total placement score were between itself
and drug use (r=.246) and parental commitment (r=.236).

There were substantial levels of inter-item correlations between several of the
individual indicators that comprise total placement score. In addition to the correlations
between total placement score and other indicators, the largest significant bivariate
correlations were between “last grade completed” and “age at first adjudication” (r=.342),
and “probation at time of instant offense” and “prior arrest history” (r=.339). Except for
“prior placement,” “parental commitment” was not significantly correlated with other

placement characteristics. The relationship between prior social service placement and

 

3 ‘ A full correlation matrix of all variables included in various models Appendix C.

122

Table 13b. Correlation Matrix: Placement Variables (n=369).
X5 X6 X7 X8 X9 X10 X11 X12

 

X1

X2

X3

X4

X5

X6 0200*

X7 0.044 0.128“

X8 0.036 0.030 0.258M

X9 0.104" 0073* -0.037 -0.049

X10 0.017 -0.005 0.107* 0.103" 0.026
X110.162** 0.021 0.003 0.068 0.032 0.098“l
X12 0.557" 0.404“ 0.259“ 0.253M 0.386'MI 0.236" 0.503"

* Correlation is significant at the 0.05 level (2-tailed).
"”" Correlation is significant at the 0.01 level (2-tailed).

parental commitment was not surprising considering social service placement is often a
result of problems associated with families, especially parental problems.

There were numerous instances of negative correlations between the placement
variables. “Age at first adjudication” was negatively correlated with school status (r=-
.039), drug use (r=-.146), prior placement (r=-.032), and peer relations (r=-.250). Age at
first adjudication should be expected to be negatively correlated with the other
indicators. The variable was coded to assign higher values for younger offenders, and
characteristic that reflects greater presumed risk.

Intake placement characteristics indicated the historical and concurrent control
groups were slightly more seriousness offenders when compared to the experimental
group. The differences were most pronounced for age at first adjudication, prior
placements, offense seriousness, and peer relations. Delinquency theory has generally
reported characteristics identified in the initial risk classification scheme differentiate

delinquents from nondelinquents and serious delinquents from less serious delinquents.

123

Table 14. Recidivism for Study Groups - Binary Coding Scheme (n=376).

 

 

Study Group None 1+ Total
Historical Count 89 43 1 32
Row % 67% 33% 100%
Concurrent Count 81 28 109
Row % 74% 26% 100%
Experimental Count 83 52 135
Row % 62% 39% 100%
Total Count 253 123 376

Row % 69% 31% 100%

Model Statistics
2:2 4.512
d.f. 2
Sig. .105

 

Initial Recidivism Measures

The first measures of recidivism are provided for all three sample populations.
This first section analyses the relationship between study group membership, court
identified delinquency risk factors, and recidivism. It was hypothesized that individuals
assigned to the community-based intensive intervention program (experimental) would
experience less recidivism than those assigned to traditional program services. The
hypothesis was based on the understanding that delinquency is caused by the interaction
of multiple delinquency risk-factors, but also because the experimental and historical
control groups exhibit fewer delinquency risk factors. In order to make positive change
in “delinquent careers,” intervention services must be directed at the antecedent factors
that underlie the problem (Howell, 2001).

Recidivism is a global measure of at least one arrest within eighteen months of the

referral date for the instant offense.32 This type of global measure is not the most

 

32 In this case “recidivism” is restricted to new criminal violations and does not reflect court petitions for
administrative action.

124

Table 15. Recidivism for Study Groups (n=376).

 

 

Study Group None 1-2 Times 3+ Times Total
Historical Count 89 4O 3 132
Row % 67% 30% 2% 100%
Concurrent Count 81 21 7 109
Row % 74% 19% 6% 100%
Experimental Count 83 45 7 135
Row % 62% 33% 5% 100%
Total Count 253 106 17 376
Row % 67% 28% 5% 100%
Model Statistics
x2 8.510
d.f. 4
Sig. .075

 

preferred method for operationalizing recidivism. It would be valuable to create
operational definitions for different types of recidivism such as rearrests for serious
personal crime, less serious personal crime, serious property crime, less serious property
crime, status offense, and other violations. This would create the ability to understand
not only if one program method is more effective for preventing recidivism, but more
effective for preventing crimes that cause greater harm to society. However, only 123 of
the original 376 individuals included in the analysis recidivated. When disaggregated by
group membership and offense type, it could be difficult to establish a measure with
substantial enough variation and numbers for inferential statistical analysis. More
importantly, serious crime was a rare exception for this population. As such, the global
measure will be used in this instance.

The recidivism measure coded in Table 14 was a binary coding scheme where a
value of “1” indicated at least one court referral for a new criminal offense afier referral

for the instant offense. A larger proportion of juveniles adjudicated to the community-

125

based program (experimental group) experienced one or more authorized court referrals
after program intake. For example, 39% (n=52) of the members of the experimental
group had one or more referrals compared to approximately 27% for the historical control
group and 33% for the concurrent control group. Although these differences do not reach
statistical significance (x2=4.512, d.f.=2, p.>.05), the differences served as an initial
indication that individuals adjudicated to traditional, residential probation services
recidivate less than those in community-based alternatives. This appears to be the case
even though the previous review of the placement characteristics suggested those in the
concurrent control group were more serious offenders than the others. It was originally
hypothesized that community-based treatment would be a more effective form of
intervention services since it would likely deliver more comprehensive services in a
community setting.33

Table 15 further specified the bivariate relationship between group membership
and recidivism by decomposing the number of future crimes into “none,” “1-2,” and
“3+.” The purpose of this analysis was to determine if there were differences in
proportion of offenders who not only recidivated, but who were considered among the
chronic recidivists, with chronic refening to those arrested and referred on three or more
occasions.

A similar proportion of both the experimental and concurrent control groups were

considered chronic recidivists (5% and 6% respectively), while the historical control

 

33 An additional analysis was considered that operationalized recidivism in a rrranner that excluded arrests
for status offenses. The concern was that members in the community-based treatment program might have
been more likely to be arrested for minor offenses while they were on probation in the community. The
analysis revealed similar findings that, regardless of whether status offenses were included or excluded,
individuals in the experimental group recorded proportionately more instances of recidivism than either the
historical or concurrent control groups.

126

group recorded proportionately fewer (2%). The group means were not significantly
different (x2=8.510, d.f.=4, p.>.05) but did approach statistical significance. It is likely
larger cell sizes may have revealed statistically significant differences. This finding
suggests there may be a cohort effect where offenders referred during one time period, in
this case July 1, 1995 to June 30, 1996, were referred more than offenders in a previous
cohort. This may be indicative of increased levels of criminality among a new cohort of
juvenile offenders or of policy changes among the police or court personnel that
increased the likelihood of arrest and referral.34

Table 16 provides more detailed recidivism information for different types of
crimes. The purpose of this analysis was to provide an overview of differences among
the three populations in type of future offending. Separate bivariate significance tests
were computed for personal crime, property crime, and other offenses. The three
outcome measures were binary variables with “1” indicating the presence of at least one
instance of recidivism for that type of crime. The final column “any” is the same as
computed in Table 15 and was included for comparison purposes. Except for property
offenses, the three groups were not significantly different. Referring to property
offenses, it was evident that individuals in the experimental group exhibited a
significantly greater number of property offenses than both the historical and concurrent
control groups.

One of the primary outcome measures considers the relative efficacy of

community-based treatment in preventing future delinquency. Table 17 presents a

 

3‘ It unlikely the differences are a result of increased levels of criminality between the arrest cohorts since
the cohorts were in such close proximity (12 months apart).

127

Table 16. Recidivism by Crime Type (n=376).
Personal Property' Other Any

 

Historical Count 8 1 8 24 43
Row % 6% 14% 18% 33%
Concurrent Count 7 16 10 28
Row % 6% 15% 9% 26%
Experimental Count 14 33 21 52
Row % 10% 24% 16% 39%
Total Count 29 67 55 123
Row % 8% 18% 15% 33%

 

' Groups significantly different (x2=6.358, d.f.=2, p<.05).

logistic regression model predicting the onset of recidivism from a series of independent
variables that include age at instant offense, offense type, and placement variables.35
Like linear regression, the logistic regression coefficient can be interpreted as the change
in the dependent variable associated with a one-unit change in the independent variable.
The model differs in that the slope of the curve varies depending on the value of the
dependent variable (Menard, 1995: 43). The column labeled “B” represents the
standardized regression coefficient. Standardizing the regression coefficient allows the
measure to be compared across different independent variables coded on varying scales.

The odds ratio represents the relationships between the independent variables and the

 

3’ “Total placement risk score” was excluded from the analysis because of problems associated with multi-
collinearity. Problems of multi-collinearity arise when independent variables are highly correlated with
one another. High multi-collinearity produces large variances for the slope estimates and large standard
errors (Lewis-Beck, 1980: 59). Diagnostic tests available through linear regression analysis were used to
test problems of collinearity. It is appropriate to employ diagnostic tests such as tolerance levels and
variance inflation factors (VIF) from linear regression because the analysis tests linear combinations of the
independent variables separate from their effects on the dependent variable (Lewis-Beck, 1980). Tolerance
levels as high as .9 and VIF values nearing 2.5 can be considered the upper levels of acceptability to
determine collinearity not to be a problem. VIF values as high as 2.78 are indicative of correlations nearing
.8, levels generally considered high (Fox, 1991: 12). Models A1 and B1 in Table B-1 (Appendix B) detail
collinearity diagnostics for the placement variables and total risk score for models that both exclude and
include consideration of treatment modality. The VIF value for total placement risk is 75.750 (Model Al)
indicating almost perfect collinearity (1.0). Similar VIF value is evident in Model Bl (76.995). Thus, it is
excluded from this analysis. Models A; and B; are the same as above but exclude total risk score. There
are no apparent problems with multicollinearity when total risk score is excluded.

128

Table 17. Multivariate Logistic Regression Model. Official Recidivism as
Dependent Variable (Full Sample, n=369).a

 

 

. Odds
B S.E. Wald d.f. Srg. Ratio
Constant 5.140 2.224 5.341 1 0.021 170.640
Age at Instant Offense a -0.456 0.132 11.914 1 0.001 0.634
Offense Type a -0.000 0.120 0.000 1 0.998 1.000
Placement Characteristics b
Age 1st Referral 0.759 0.831 0.834 1 0.361 2.136
Prior Arrests 1.228 0.752 2.667 1 0.102 3.413
School Status 0.947 0.717 1.744 1 0.187 2.578
Drug Use 1.109 0.868 1.635 1 0.201 3.033
Offense Seriousness 0.730 0.595 1.503 1 0.220 2.074
Probation 0.119 0.871 0.019 1 0.892 1.126
Prior Placement 0.170 0.882 0.037 1 0.847 1.186
Prior Runaway -1.282 1.476 0.754 1 0.385 0.278
Last Grade Completed -1.074 1.088 0.974 1 0.324 0.342
Parental Commitment -0.954 1.1 19 0.726 1 0.394 0.385
Peer Relations 0.525 0.786 0.446 1 0.504 1.691
Model Summary
Model x2 31.491
d.f. l3
sig. 0.003
R1 0.114

 

' Seven cases excluded due to missing intake placement data.
b Variables transformed using natural log.

“odds” of being a recidivist. A value greater than “1” indicates increased odds, and a

value less than “1” decreased odds (Menard, 1995).”

6‘1”

Recidivism was coded as a binary variable with indicating the occurrence of

at least one instance of recidivism during the 18 month follow-up period and “0”

 

3° The “odds ratio” calculation has one major disadvantage. In the case of large unstandardized regression
coefficients, the estimated standard error is inflated that can result in a Type 11 error, or a failure to reject
the null hypothesis when the null hypothesis is in fact null. This problem is rrrinimized in this situation by
using log transformed values for the independent variables.

129

indicating no instances of recidivism.37 The model was for the firll sample population
and did not incorporate sample characteristics (experimental or control groups) features.
The model summary statistics indicate the chi-square distribution for the full model was
statistically significant (x2=31.491, d.f.=13, p<.05). Of all of the variables included in the
model, age at instant offense was the only variable significantly associated with the onset
of recidivism. The relationship was negative suggesting that older offenders recidivate
less and younger offenders more. Older offenders were almost 40% less likely to
recidivate than younger offenders. More specifically, a one unit decrease in age at instant
offense was associated with a 40% increase in recidivism. This relationship was
statistically significant even after controlling for intake placement factors such as drug
use, peer relations, and parental commitment. This finding was consistent with previous
that indicate age at arrest is highly predictive of future delinquency (Hawkins et al.,
1998). The pseudo-R2 indicates the linear combinations of the independent variables
explained only a moderate amount of variation in the dependent variable.38

Table 18 builds on the previous analyses by entering study group membership
into the model.39 It was hypothesized that after controls were entered for background risk
characteristics (intake placement variables), individuals exposed to YFS intervention
program would recidivate less than those exposed to traditional probation services. To
test this hypothesis, two dummy variables were entered into the model in a stepwise

fashion. The first assigns a value of “1” for members of the historical control group, and

 

37 See Appendix B for multicollinearity tests.

38 “Although one is tempted to think of this quantity as the proportion of variance explained by the model,
it is not quite correct to do so” (Demaris, 1992: 53). Instead, it can be considered a rough approximation of
the predictive efficacy of the model. The pseudo-R2 tends to underestimate the explanatory power in
logistic regression, a consequence of the binary coding for the dependent variable that limits variability.

39 See Model B in Table B-1 (Appendix B) for multicollinearity tests. The model excludes “total risk
score” from the logistic regression analysis because of apparent problems with multicollinearity. Exclusion
of the variable reveals no apparent problems.

130

Table 18. Multivariate Logistic Regression Model of Official Recidivism
Controlling For Experimental Group Membership. Official Recidivism as
Dependent Variable (n=369).a

 

 

. Odds
B S.E. Wald d.f. Srg. Ratio
Constant 5.212 2.290 5.182 1 0.023 183.498
Age at Instant Offense b -0.465 0.135 11.866 1 0.001 0.628
Offense Type b -0154 0.129 1.423 1 0.233 0.857
Placement Characteristics b
Age lst Referral 1.234 0.860 2.060 1 0.151 3.437
Prior Arrests 1.036 0.782 1.755 1 0.185 2.818
School Status 1.220 0.755 2.608 1 0.106 3.386
Drug Use 1.158 0.919 1.586 1 0.208 3.182
Offense Seriousness 1.157 0.632 3 .353 1 0.067 3.181
Probation 0.522 0.903 0.334 1 0.563 1.686
Prior Placement 0.750 0.914 0.674 1 0.412 2.1 17
Prior Runaway -0.548 1.518 0.130 1 0.718 0.578
Last Grade Completed -0.985 1.132 0.758 1 0.384 0.373
Parental Commitment -0.857 1.161 0.544 1 0.461 0.425
Peer Relations 1.014 0.837 1.470 1 0.225 2.758
Historical Group -0.552 0.290 3.618 1 0.057 0.576
Concurrent Group -1.049 0.327 10.303 1 0.001 0.350
Model Summary
Model x2 41.743
d.f. 15
sig. 0.000
R1 0.151

 

' Seven cases excluded due to missing intake placement data.

1’ Variables transformed using natural log.

the second a value of “1” for those in the concurrent control group. Entering both

variables into the model simultaneously established “experimental” as the reference

category.

The model summary statistics indicated a significant “goodness of fit”

(x2=41.743, d.f.=15, p<.05) among the independent and the dependent variables. Thus,

the linear combination of the independent variables predicted a significant amount of

variation in the dependent variable. When group membership was entered into the

131

model, the relationship between age at instant offense and recidivism remained
significant. Regardless of treatment modality or other background characteristics,
individuals arrested at a younger age were significantly more likely to recidivate.

The final preliminary recidivism measure stepped time in detention40 into the
model. Number of days in detention, a control analogous to “time at risk,” an adjustment
recommended (Winner, Lanza—Kaduce, Bishop, & Frazier, 1997) to account for time out
of the community that reduces the ability to participate in certain type of criminal
activities. The findings from Table 19 did little to clarify the relationships between
background characteristics, delinquency risk factors, treatment modality, and official
recidivism. With the addition of detention days, the overall model goodness of fit (see
Table 19)“ remained significant (x2=42.734, d.f.=16, p<.05) but there was no effect on
the pseudo-R2 (.15). Referring to the independent variables, age at instant offense
remained significant and the odds ratio was also unchanged. The addition of detention
days had a modest effect on the significance level for prior arrests and drug use but little

other practical significance. Membership in the concurrent control group remained

 

‘0 Days in detention refers to number of days detained in the County’s juvenile detention facility. This
measure, however, is problematic in that it does not account for other type of social service placements that
may, and ofien did, occur in institutions besides the County juvenile detention facility. Thus, the measure
may systematically underestimate the actual “time at risk,” especially for the concurrent control group
which would have experienced the most out of County placements. However, the measure does serve as a
baseline control for time at risk.

Mean detention days were calculated for all three groups. As expected, the most number of days
in detention was reported for the concurrent control group (J? = 54 s.d.=62, std. error-=59) followed by the
historical control group ()7 =45, s.d.=51, std. error=4.44) and the experimental group (J? = 40 s.d.=62, std.
error=4.44). These group means are not statistically significant (F=1.41, p.>.05). The large standard
deviations in conrparison to the means suggest a potential problem with high levels of variance. To
account for this problem, the variable was transformed using the natural log. Transforming the variable
effectively reduced the variance in the distribution. The group mean differences of the transformed
variable are significantly different (F=15.3 l, p<.001).

‘1 The Variance Inflation Factors and Tolerance tests presented in Table B-2 (Appendix B) indicate
multicollinearity is non-problematic in this model.

132

Table 19. Multivariate Logistic Regression Model Controlling For
Experimental Group Membership, Days in Detention Included (n=369).a

 

 

. Odds
B S.E. Wald d.f. Srg. Ratio
Constant 5.576 2.260 6.085 1 0.014 263.907
Age at Referral b -0.492 0.134 13.435 1 0.000 0.612
Offense Type b -0.078 0.125 0.388 1 0.533 0.925
Placement Characteristics b
Age lst Referral 1.070 0.843 1.61 1 1 0.204 2.914
Prior Arrests 1.282 0.774 2.745 1 0.098 3.603
School Status 1.086 0.737 2.172 1 0.141 2.964
Drug Use 1.620 0.905 3.203 1 0.074 5.055
Offense Seriousness 1.072 0.622 2.964 1 0.085 2.920
Probation 0.299 0.900 0.1 10 1 0.740 1.348
Prior Placement 0.420 0.911 0.213 1 0.645 1.522
Prior Runaway -0.840 1.518 0.306 1 0.580 0.432
Last Grade Completed -1.086 1.108 0.960 1 0.327 0.338
Parental Commitment -0.749 1 .141 0.431 1 0.512 0.473
Peer Relations 0.838 0.818 1.049 1 0.306 2.312
Det Days b 0.083 0.201 0.170 1 0.680 1.087
Historical -0.299 0.288 1.080 1 0.299 0.741
Concurrent -1.073 0.333 10.382 1 0.001 0.342
Model Summary
Model x2 42.734
d.f. 16
sig. 0.000
R2 0.152

 

' Seven cases excluded due to missing intake placement data.
5 Variables transformed using natural log.

significant. Members of this group were almost 65% less likely to recidivate than the
experimental group.

This initial recidivism analysis revealed mixed support for the argument that the
delinquency risk factors identified by the developmental perspective were important
considerations in the understanding why certain individuals recidivate more than others.

The findings from Tables 18 and 19 suggested some basic support for the developmental

133

perspective based on the finding that individuals who offended earlier (e. g., age at instant
offense) were more likely to recidivate than those who were older when referred for the
instant offense. Yet, in general, the independent variables stepped into the equation and
hypothesized to be important predictors of recidivism failed to do so. From the
developmental perspective, it was expected that stage-salient issues such as school and
peer issues would have been important factors in predicting recidivism. The mean age
for all three populations was 15 so school and peer issues should manifest themselves as
stage-salient issues if the theory explains delinquent behavior as posited. Except for age
at referral for instant offense, none of the stage-salient issues found by previous research
to be important to the onset of delinquency were predictive of recidivism for adjudicated
juvenile offenders.

These findings may indicate a need to go no further in regards to understanding
why some offenders recidivate and others do not. However, several questions remain
unanswered in determining the efficacy of comprehensive community-based treatment
programs. First, was the seeming failure of the alternative treatment program explained
by delinquency risk factors that were not adequately reflected in the intake placement
variables? Ifthe intake placement assessment did not effectively operationalize the full
dynamics of “risk,” inclusion of these variables would not enter in the appropriate
controls for background characteristics. Second, do neighborhood characteristics mediate
the effectiveness of intervention programs? If individuals from extraordinarily
disorganized or impoverished neighborhoods were intentionally or unintentionally over
sampled to the experimental group it could be they are subject to more extensive negative

neighborhood influences. Third, are some programs more effective than others because

134

they are better able to “fix” or change delinquency risk factors than others? Program
models often posit intervention domains that are linked to theoretical fimneworks
stipulating specific causal mechanisms. It would be expected that individuals
experiencing greater levels of positive change in delinquency risk factors would
recidivate less. Finally, is “success” attributable to differences in the amount of
programming received? Some have argued rehabilitation programs are not effective
because they fail to deliver any meaningful rehabilitation services. If so, it could be
differences in quantity of treatment that explains differences in relative success.

The questions posited above shape the current research. Evaluation research of
intervention programs often fails to address these questions. The typical evaluation
proposes a program model and then measures relative “success” comparing multiple
groups. This research will delve into these questions by performing additional analysis
on the concurrent control and experimental groups. Individuals in these two groups were
subjected to a series of structured interviews that provided the opportunity to consider the
potential effected of these issues.

Comparison of T1 Delinquency Risk Factors

A series of four structured interviews were administered to the experimental and
concurrent control groups. The first interview was conducted at or near intake (T1), the
second after 6 months of program participation (T2), the third after 12 months (T3), and
the fourth 18 months (T4) after adjudication. The interviews were assessments of
delinquency risk factors, background and characteristics, and self-reported delinquency.
This provided a much more comprehensive picture of each individual than was available

from review of the intake placement data alone. The experimental and concurrent control

135

groups were compared to determine the extent to which they shared T1 delinquency risk
factors. Research reports individuals with increased delinquency risk factors demonstrate
earlier onset of delinquency, chronic offending (Hawkins et al., 1992; Hawkins et al.,
1998; Heilbrun et al., 2000; Loeber, 1990; Maxwell & Maxwell, 2000; Scholte, 1992;
White, Loeber, Stouthamer-Loeber, & Farrington, 1999; Wilson & Howell, 1995), and
also respond less favorably to treatment efforts (Davidson et al., 1990; Lipsey, 1999;
Lipsey, 2000; Lipsey & Wilson, 1998). Values for the risk factors were computed for
negative attitudes toward school, negative school behavior, favorable attitude toward
delinquency, self-esteem, prosocial values, exposure to delinquent peers, positive family
involvement, alcohol/drug use, stressful life events, and community involvement (Table
20). Index scores for each risk factor were determined by computing the mean scale
score over a series of indicators.42

Negative attitude toward school represents the degree to which individuals
believed they failed in their school environment. Lower values indicate more affirmative
belief that success can be achieved in the school environment, and higher values indicate
that success is less likely to be achieved. The group means were approximately 2.4 and
were not statistically distinguishable (F=.563, p<.05).

Negative school behavior reflects levels of participation in problematic behaviors
while on school property. The scale covers a wide-range of problematic behavior in
school, everything fiom cheating on tests and being truant to committing acts of
vandalism. Higher values on this scale indicate an individual participated in higher levels

of problematic school behavior. Members of the concurrent control group reported

 

‘2 See “Scaling Procedures” in Chapter 3.

136

Table 20. Descriptive Statistics for T1 Risk Factors (n=178)‘.

 

 

“r. Risk Factors x 18):“; 5:; Min Max 824:2; Sig.
Negative Attitude Toward School (A8“) 0.549 0.563 0.454
Concurrent 2.448 0.939 0.106 1.00 5.13
Experimental 2.336 1.024 0.103 1.00 5.00
Total 2.386 0.986 0.074 [.00 5.13
Negative School Behavior (88“) 9.655 5.63 0.019
Concurrent 3.466 1.365 0.154 1.17 6.17
Experimental 2.997 1.264 0.127 1.00 7.00
Total 3.205 1.327 0. 099 1.00 7. 00
Favorable Attitude Toward Delinquency (SD11) 0.039 0.169 0.681
Concurrent 3.480 0.427 0.048 2.33 4.00
Experimental 3.510 0.518 0.052 1.22 4.00
Total 3.497 0.479 0. 036 [.22 4. 00
Self-esteem (SETI) 0.149 1.015 0.315
Concurrent 2.932 0.407 0.046 2.00 3.60
Experimental 2.990 0.363 0.036 2.20 3.60
Total 2. 964 0.383 0.029 2. 00 3. 60
Prosoclal Value (PVTl) 0.004 0.021 0.884
Concurrent 3.437 0.444 0.050 2.00 4.00
Experimental 3.428 0.386 0.039 2.50 4.00
Total 3.432 0.412 0.031 2. 00 4. 00
Exposure to Delinquent Peers (EPn) 2.346 3.435 0.065
Concurrent 2.351 0.829 0.093 1.00 4.19
Experimental 2.120 0.825 0.083 1.00 4.50
Total 2.222 0.832 0.062 1.00 4.5 0
Positive Family Involvement (FIP-n) 0.552 0.248 0.619
Concurrent 6.435 1.501 0.169 0.00 8.63
Experimental 6.323 1.486 0.149 1.75 8.88
Total 6.373 1.489 0.112 0. 00 8. 88
Alcohol/Drug Use (DUn) 13.383 7.946 0.005
Concurrent 1.952 1.424 0.160 0.00 6.00
Experimental 1.400 1.187 0.1 19 0.00 5.40
Total [.645 1.323 0.099 0. 00 6. 00
Stressful Life Events (SLTI) 0.461 22.351 0.000
Concurrent 1.361 0.132 0.015 1.10 1.74
Experimental 1.463 0.152 0.015 1.13 1.81
Total [.418 0.152 0.011 1.10 1.8]
Community Involvement (C111) 0.437 0.739 0.391
Concurrent 2.648 0.776 0.087 0.33 4.1 1
Experimental 2.748 0.763 0.077 0.78 4.78
Total 2. 703 0. 768 0. 058 0.33 4. 78

 

' Experimental (n=99), concurrent (n=79).

137

significantly (F=5.63, p<.05) more negative school behavior (3? = 3.47) than the
experimental group (J? = 2.99 ).

The groups were not different on favorable attitudes toward delinquency, levels of
self-esteem, prosocial values, exposure to delinquent peers, or positive family
involvement. Higher values for attitudes toward delinquency were suggestive of attitudes
approving of delinquent activity. Those with higher values for self-esteem are said to
think more positively about themselves and their abilities than their counterparts.
Prosocial values measure the extent to which mainstream values such as the need to work
hard are shared. Positive family involvement operationalizes different levels of
participation in positive activities with family members.

The groups did, however, differ significantly on measures of alcohol and drug
use, and exposure to stressful life events. Members of the concurrent control group
reported significantly more drug use (F =7.95, p<.05) than those in the experimental. This
finding should be considered important because prior research has found substance abuse
is predictive of delinquency (Dembo, Williams, Fagan, & Schmeidler, 1993; Grobsmith,
1989; Hawkins et al., 1992; Stice et al., 1998; Teplin, 2001; White et al., 1999). In
contrast, members of the experimental group reported significantly more cumulative
stressful life events (F=22.351, p.<.001) than those in the concurrent control group.
Exposure to stressful life events increases aggression (Attar, Guerra, & Tolan, 1994;
Hoffinan & Cerbone, 1999), and thus was expected to effect delinquency and recidivism.

Consideration of T1 delinquency risk factors indicates the groups were remarkably
similar to one another. There were no significant differences between group means on

most of the indicators except for negative school behavior, alcohol and drug use, and

138

stressful life events. Except for stressful life events, the data indicate members of the
concurrent control group exhibit greater risk for delinquency and increased future
delinquency. This finding was consistent with findings from the analysis of intake
placement variables. However, members of the experimental group report exposure to
significantly more stressful life events than their counterparts.

Reductions in Risk

Effective intervention programs are those that reduce the risk factors associated
with delinquency (Catalano, Arthur, Hawkins, Berglund, & Olson, 1998). It was
hypothesized that the Youth and Family Studies program would be less intensive than
residential placement in terms of intensive monitoring of the individuals since youth
would only be supervised by staff while receiving services. However, was also
hypothesized the community-based program would still be a more effective intervention
model because it delivered targeted, stage-salient services to problem areas commonly
associated with delinquency. This systematic service delivery model was hypothesized to
not only produce a greater reduction in recidivism than traditional services, but was also
hypothesized to be more effective in reducing the risk factors associated with
delinquency.

Data presented in Table 21 indicated changes in delinquency risk factors targeted
by intervention services. The measures represent changes in the raw scores and were
computed with the following formula (T31, — Tn.) where “k” represents the construct of
interest, and represent changes in risk. As discussed in Chapter 3, change scores were
computed using T1 and T3 because it provided enough time for an intervention to take

hold but not so long as to expect dissipation of the effect. Research that employs too

139

Table 21. Mean Charge in Delinquency Risk Factors (n=178).

 

T. Risk Factors x S‘d' 8‘“ Min Max ”e" F Sig.
Dev. Error Square

Negative Attitude Toward School (AS Change) 0.666 0.459 0.499
Concurrent -0.3 86 1.169 0.131 -3.500 2.630
Experimental -0.263 1.231 0.124 -3.250 2.880
Total -0.318 1.202 0.090 -3.500 2.880

Negative School Behavior (SB Change) 3.478 1.320 0.252
Concurrent -l .185 1.759 0.198 -5.000 4.170
Experimental -0.903 1.506 0.151 -4.920 2.920
Total -1.028 1.625 0.122 -5.000 4.170

Favorable Attitude Toward Delinquency (AD Change) 0.343 1.171 0.281
Concurrent 0.020 0.525 0.059 -1.330 1.440
Experimental -0.068 0.553 0.056 -2.1 10 1.440
Total -0.029 0.541 0.041 -2.1 10 1.440

Self-esteem (SE Change) 0.421 2.188 0.141
Concurrent -0.157 0.464 0.052 -1.400 1.400
Experimental -0.059 0.417 0.042 -1.200 0.800
Total -0.103 0.440 0.033 -1.400 1 .400

Prosocial Value (PV Change) 0.019 0.072 0.789
Concurrent -0.027 0.519 0.058 -1.500 1.170
Experimental -0.007 0.506 0.051 -1.000 1.500
Total -0.016 0.510 0.038 -1.500 1.500

Exposure to Delinquent Peers (EP Change) 0.074 0.087 0.768
Concurrent -0.197 0.985 0.1 1 1 -2.940 1.600
Experimental -0.156 0.868 0.087 -3.250 2.130
Total -0.174 0.919 0.069 -3.250 2.130

Positive Family Involvement (FIP Change) 6.290 1.493 0.223
Concurrent 0.822 2.271 0.256 -6.880 7.880
Experimental 0.444 1.860 0.187 -4.250 5.880
Total 0.611 2.055 0.154 -6.880 7.880

Alcohol/Drug Use (DU Change) 20.810 8.496 0.004
Concurrent -0.927 1.780 0.200 -5.200 2.250
Experimental -0.238 1 .370 0.138 -5.400 2.800
Total -0.544 1.598 0.120 -5.400 2.800

Stressful Life Events (SL Change) 0.342 4.824 0.029
Concurrent 0.181 0.275 0.031 -0.580 0.710
Experimental 0.092 0.259 0.026 -0.690 0.680
Total 0.131 0.269 0.020 -0.690 0.710

Community Involvement (Cl Change) 2.702 2.803 0.096
Concurrent -0. 183 0.972 0.109 -3.220 2.1 10
Experimental 0.065 0.990 0. 100 -2.670 3 .890
Total -0.046 0.987 0.074 -3.220 3.890

 

140

short a follow-up period runs the risk of over inflating the effect Size of treatment
outcomes (Lipsey & Derzon, 1998). Both the experimental and concurrent control group
experienced several reductions in delinquency risk factors, or conversely, an increase in
protective factors. For example, both groups experienced decreases in negative attitudes
toward school, negative school behavior, exposure to delinquent peers, and decreases in
drug use. Individuals in the experimental group also reported a modest decrease in
favorable attitudes toward delinquency. Members of the concurrent control group
reported significantly greater reductions in alcohol and drug use than experimental youth.
This may be a function of less opportunity stemming from residential placement.

Both groups also reported increases in positive family involvement. While the
experimental group reported a modest increase in community involvement, the
concurrent control group experienced a modest decrease. The decrease makes sense
since members of the concurrent control group were in residential placement programs
and therefore unable to participate in community activities to the same extent as the
experimental group. For unknown reasons, both groups also reported decreases in self-
esteem and prosocial values. From the perspective of self-esteem, it is possible the
process of being labeled delinquent is responsible for this phenomenon, something Leger
(1981) has suggested as a possible outcome of formal processing in the juvenile justice
system. One delinquency risk factor increased for both groups, stressful life events, and
the mean increase for the concurrent control group (5? = .18) is significantly (F =.342,
p<.05) higher than the experimental (J? = .09 ).

Notwithstanding a few notable exceptions, both groups experienced risk factor

changes in the expected directions. However, it was hypothesized that the changes would

141

be significantly more positive for members of the experimental group. As can been seen
in Table 21, the differences were only significant for reductions in alcohol and drug use
and an undesired increase in stressful life events. While the significant differences in
alcohol and drug use are noteworthy and important, the expectation was that the
differences would be more comprehensive. It is important to try to understand why there
are not more substantial differences in risk factor changes between the two populations.
To do so, two related analyses were conducted. The first compared the experimental and
concurrent control group on levels of programming they received while on probation.
Next, bivariate correlation analysis detailed the relationships between varying levels of
programming services and risk factor change. This step was important in deciding if the
“failure” to create significantly more positive change is a result of a failure in treatment
modality or a failure to effectively deliver intervention services.
Programming Services

Several research questions have been proposed that seek to determine the
relationship between levels and types of programming and both changes in delinquency
risk factors and recidivism outcome measures. Lipsey and Wilson (1998) argued
intervention programs can be effective in reducing recidivism, but characteristics of the
program itself are important considerations in determining why some programs are more
effective than others.

Levels and types of programming were determined by a series of questions
administered during the structured interviews. Respondents were asked to identify up to
three specific programs they had participated in since they were placed on probation. A

series of twenty-two questions were then asked about each program individually. For

142

Table 22. Comparison of Programming Totals (n=178).

 

Pro rammin — Std. Std . Mean .
Sgervices g N X Dev. Error Mm Max Square F Srg.
Counseling 70.204 0.392 0.532
Concurrent 79 17.6 12.4 1.4 14.8 20.3

Experimental 99 16.3 14.2 1.4 13.5 19.1

Total 178 16.9 13.4 1.0 14.9 18.8
Educational 207.106 0.700 0.404
Concurrent 79 22.7 17.9 2.0 18.7 26.8

Experimental 99 20.6 16.6 1.7 17.3 23.9

Total 178 21.5 17.2 1.3 19.0 24.1
Job Skills 656.091 6.978 0.009
Concurrent 79 12.2 9.7 1.1 10.1 14.4

Experimental 99 8.4 9.7 1.0 6.4 10.3

Total 178 10.1 9.9 0.7 8.6 11.5
Recreational 439.234 1.489 0.224
Concurrent 79 24.1 17.5 2.0 20.1 28.0

Experimental 99 20.9 16.9 1.7 17.5 24.3

Total 178 22.3 17.2 1.3 19.7 24.8

 

example, after providing the names of each program, respondents were asked how often
they “attend group meetings,” “learn job skills,” and “attend academic classes” for each
identified program. Programming services were grouped into counseling services,
educational programs, job skills training, and recreational opportunities. Frequency was
measured on a nine point scale that ranged from “never” to “2-3 times a day.” Each
construct (e. g., counseling) is comprised of 3 to 5 indicators that were aggregated within
and then across programs.

Table 22 summarizes the program services received by members of both the
experimental and concurrent control groups.43 It was hypothesized that individuals in the

community treatment program would receive more comprehensive services than those in

 

‘3 Respondents were asked to identify up to 4 different programs they participated in over the time period
covered in the interviews. They were then asked a series of questions about the specific types of services
received for each program separately. The indicators were then combined to represent the following types
of services: counseling services, educational services, job skills training, and recreational services. Thus,
each measure represents multiple indicators across one or more programs. The distributions for all four
measures were large and not normal. Future analyses used the natural log transformed figures.

143

traditional probation services. As a broad based intervention program that targeted
family, education, and social causes of delinquency the experimental group was expected
to receive more services, experience positive changes in delinquency risk factors, and
ultimately recidivate less. Although similar, the mean scores indicate the concurrent
control group received more average programming services in all categories. The job
skills mean for the concurrent control group’s (i = 12.23) mean is Significantly higher
(F=6.978, p<.05) than the experimental (i = 8.36 ).

If, as hypothesized, levels of programming are important predictors of changes in
risk factors and recidivism, it should be expected that those receiving more services
should exhibit more favorable outcomes measures. The evidence accumulated thus far
provides support for this conclusion. Although both the intake placement variables and
the delinquency risk factors indicated the concurrent control group exhibit increased
levels of “risk” for recidivism, they experienced significantly less recidivism than their
counterparts in the community-based treatment model. At this point the relationship
remains speculative. Bivariate and multivariate models will follow that address this
relationship in greater depth.

Risk Factor Changes and Programming Nexus

Tables 23a and 23b detail bivariate correlations between delinquency risk factor
change scores and measures of intervention programming. Many of the change scores
were significantly correlated and in the expected directions. Decreases in negative
attitude toward school were significantly correlated with negative school behavior
(r=.308) and indicates individuals with decreases in negative attitudes also reported

decreases in negative school behavior and that increases in negative attitudes were

144

Table 23a. Bivariate Correlations Among Risk Factor Change Scores and

Programming (n=283).

 

Correlations X1 X2 X3 X4 X5
Negative Attitude

Toward School(X1)

Ne ative School

BefiaviodXZ) 0308"

Favorable Attitude Toward

Delinquency(X3) '0'30‘" 1' '0' 195 * *

Self-esteem(X4) -0. 136* -0.074 0.009

Prosocial Values(X5) -0.084 -0.097 0.281" 0.086

Exp. Delinquent Peers(X6) 0.248" 0.410" -0. 149* -0.079 -0.120*
Positive Family Involvement(X7) -0.079 0.055 0.149* -0.008 0.104
Drug Use(X8) 0.284" 0.559" -0.312** -0.193** -0. 132*
Stressful Life Events(X9) -0.148* -0.512** -0.034 0.149* 0.050
Community Involvement(XlO) -0.199** -0.171** 0.273"”'l 0.082 0.121*
Counseling(X1 1) -0.159** -0.267** 0.092 0.092 0.087
Educational(X12) -0.189** -0.282** 0.081 0.055 0.076
Job Skills(X13) -0.1 10 -0.241** 0.075 0.089 0.097
RecreationaKX 14) -0.003 -0.1 16 0.029 0.136 0.070

 

*

Correlation is significant at the 0.05 level (2-tailed).
*"' Correlation is significant at the 0.01 level (2-tailed).

correlated with increases in negative school behavior. Similarly, the drug use change

score was also positively associated with negative attitudes toward school (r=.284) and

negative school behavior (r=.559). A change in exposure to delinquent peers was

significantly correlated with drug use (r=.335). This suggests that a decrease in exposure

to delinquent peers is significantly associated with decreases in drug use. Visual

inspection of the delinquency risk factor change scores does not indicate problems with

multi-collinearity.

Tables 23a and 23b also present bivariate correlations between the programming

levels and risk factor change scores. Aggregate measure of counseling services were

significantly and negatively correlated with negative attitude toward school (r=-.159),

negative school behavior (r=-.267), and drug use (r=-.255). Thus, the more counseling

someone received the more they were to experience decreases in school and drug use

145

Table 23b. Bivariate Correlations Between Risk Factor Change Scores and
Programming (n=283).

X6 X7 X8 X9 X10 X11 X12 X13 X14

 

X6

x7 -0.075

x3 0.335" 0.056

x9 0468" -0.095 -0.424"

x10 -0050 0.217** -0.299** 0.049
x11 -0.075 -0.120* -0255" 0136* 0.062

x12 -0.076 -0110 -0.278“ 0147* 0.071 0.972**

x13 -0045 -0.137* -0276" 0127* 0.113 0.895** 0.88**

x14 0.073 0.120 -0.255** -0.016 0251* 0.521** 0.604“ 0.413**

 

* Correlation is significant at the 0.05 level (2-tailed).
** Correlation is significant at the 0.01 level (2-tailed).

problems. Counseling was expected to increase positive family involvement, yet for an
unknown reason the data indicate a significant negative correlation (r=-. 120).
Educational services affected the change scores in a similar fashion. Educational
programming was negatively correlated with negative attitudes toward school (r=-.189),
negative school behavior (r=-.282), and drug use (r=-.278). Both counseling and
educational services had significant, positive relationships with stressful life events. Job
skills training also had a Significant, negative correlation with negative school behavior
(r=-.241) indicating problematic school behavior decreased as individuals were exposed
to job skills training. A significant, negative relationship was also evident between job
skill training and drug use. Finally, recreational opportunity was significantly and
negatively correlated with drug use (F-.276) and community involvement (r=.251).

As hypothesized, there were significant correlations between characteristics of
programming services delivered to the entire population and changes in delinquent risk
factors. Comprehensive intervention services resulted in significant reductions in

delinquency risk factors and increases in protective factors. The most consistent

146

reductions were reported for counseling, educational, and job skills services. Although
significant reductions in delinquency risk factors were not linked to levels of recreational
services in as consistent a fashion, the data support the preliminary argument that it is
important to consider the operational characteristics of services and outcome measures.
Final Recidivism Measures

Recidivism analyses presented earlier suggested that, compared to the
experimental group, both the historical and concurrent control groups were nearly 65%
less likely to recidivate within the 18 month follow-up time frame. This relationship was
apparent even when controls were established for important background characteristics
such as age at arrest for instant offense, instant offense crime type, and intake placement
characteristics. The initial findings indicated the treatment modality itself, independent
of these important characteristics, was a significant predictor of “success.” It was
hypothesized, however, that the apparent failure of the Youth and Family Studies
program to provide intervention services that were more effective in reducing
delinquency compared to traditional services was explained by characteristics not fully
conceptualized by previous models. Several hypotheses were advanced arguing relative
levels of success or failure in treatment efforts is explained by consideration of “risk
factors” at the time of intake, the level of programming services individuals received
during programming efforts, and any apparent change in risk factors that is expected to
be attributable to treatment services.

Tables 20-23 specified that there are several important differences between the
experimental and concurrent control group. Similar to the earlier conclusions, the

concurrent control group generally included more serious offenders. They exhibited

147

more significant problems with delinquency risk factors in the areas of prosocial values
and alcohol/drug use. However, the concurrent control group experienced statistically
significant decreases in stressful life events. Although the concurrent control group
reported greater problems with delinquency risk factors, there were few Significant
differences in the extent of “rehabilitation” as measured by changes in the risk factors.
The experimental group experienced moderate success in reducing consumption of drugs
and alcohol, but also surprisingly experienced more stressful life events than their
counterparts in traditional probation services. The relationships between these
characteristics remain unclear at this point.

Tables 24-27 integrate the additional elements collected through the structured
interviews into logistic regression models. These models allow for consideration of the
complex reasons why some intervention modalities were more successful than others in
rehabilitating juvenile offenders. The models will consider the impact of each grouping
variables independently, and then a fully integrated model. For each model, the
relationship between the predictor variables and recidivism will be considered
independent of treatment modality, and then secondary models step in treatment
modalities to determine how they mediate the relationships. The first model will enter
neighborhood characteristics into a model similar to Table 19 to determine the extent to
which neighborhood context effects recidivism. Subsequent models will then consider
the independent relationships between T1 risk factors, change in risk factors as measured
by the formula Xm — Xm where X, represents each delinquency risk factor, and the
effects of intervention programming. The final model then represents an integrated

model that considers the effects of all simultaneously.

148

Table 24 is Similar to the model presented on the full sample (historical control,
concurrent control, and experimental groups) in Table 19, but several additional features
were integrated into the prediction models that were not available for the historical
control group. The two neighborhood characteristics, social disorganization and socio-
economic status, were considered in this model as was the variable “total intake.” Total
intake is an index variable of all of the intake risk classification variables.44

Model A indicated that when the model was reduced to the experimental and
concurrent control groups, the model failed to achieve statistical significance. Thus, the
linear combination of independent variables did not explain a significant amount of
variance in the dependent measure, recidivism.45 In addition to the lack of statistical
significance for the model, none of the independent variables achieved the level of
statistical significance. The variable closest to achieving statistical Significance (p=.054),
age at instant offense, was negatively related to recidivism signifying that younger
offenders recidivate more than older offenders. Model B in Table 24 was the same as the
previous but entered group membership (concurrent) into the equation. “Concurrent”
was a binary coded variable where the value “1” reflects membership in the concurrent
control group and held the experimental group as the reference category.

Model B in Table 24 also did not achieve statistical significance, but there were a
few notable changes in significance and direction of relationship for several of the
independent variables. In Model B, age at referral for instant offense was Significantly
positive indicating those referred at younger ages were approximately 30% more likely to

recidivate. Total intake score was also significant and indicated those that demonstrated

 

“ The intake risk classification variables are highly correlated with the total risk score (see Tables 12a and
12b), and the use of a single measure reduces the complexity of subsequent models.
‘5 Variance Inflation Factors in Table 84 (Appendix B) indicate minor multi-collinearity problems.

149

 

 

 

3.66 .5— 366 .5.

636 .36 69.6 .65

e a... m .6...

$66 - 626 ~u
5.2 83 _ a: ~_e.~ 662 68an 666.2 e-6 _ £66 663 3.2 :5an
6.3.6 366 _ 6666 -v6 -w6- aoeeoeeo - - - - - - .8585
£66 866 _ ~86 3~6 :66- sense 656 E66 _ 6:6 5~6 83. season
~36 56 _ ~36 83 6666- $5 $66 366 _ :66 33 686- mmm
$66 ~86 _ .666 £~6 :66- 85 8m 6666 866 _ :66 6-6 ~86- ...5 8m
66~._ 6~66 _ 69.6 686 ~66 835369 :3 686 _ 6666 686 26 £553.;
:3 :66 _ 366 6:6 :66 85.60883 68.— 4666 _ 6:6 86 ~86 oaoboaas:
626 ~86 6 6%... ~26 636- Seems owe. 9:6 :66 _ 6666 626 86. 8383?.
8.5 .66 .6 25$ dd m 8.5 Em .6 2.3 .66 a

£6.52 «Jove—z

.321: 85:33.5 6529.8: 32829: o3»...— o..t§£=2 ...~ 2.3.

150

greater risk via the intake instrument were 20% more likely to recidivate. The addition of
group membership resulted in a Slight increase in explained variance, but an increase that
was of little practical importance. For both Models A and B in Table 24, the measures of
neighborhood disorder and low socio-economic status were negatively associated with
recidivism. This provided evidence that individuals residing in highly disorganized and
impoverished communities at the time of their arrest are more likely to recidivate.

The next set of regression equations incorporates basic background variables and
T. risk factors (Tables 25a and 25b). Background variables include age at instant
offense, instant offense type, total intake placement score (index value), neighborhood
characteristics, days in detention, and an index of T1 delinquency risk factors. Although
it would have been desirable to consider all T1 delinquency risk factor independently, it
was not possible due to multi-collinearity in the model. To account for problems with
multi-collinearity two series of models were analyzed. In the first (Table 25a), T1 risk
was represented by an index of all risk factors that account for the combined effects of
risk. In Table 25b, the analysis was restricted to include only T1 negative school
behavior, T1 alcohol/drug use, and T1 stressful life events. These were the only three T1
risk factors on which the two populations differed46 (see Table 20). For both Tables, the
first model (Model A) details the relationships without entering group membership into
the equation and the second (Model B) represents the complete model.

Model A did not reach statistical significance in Table 25a, and except for age at
instant offense which only nears statistical Significance (p=.055), the remainder of the
predictor variables failed to yield significance levels close to statistical significance.

Little was added by considering T1 negative school behavior, drug/alcohol use, and

 

’6 Collinearity diagnostics in Table B-5 (Appendix B).

151

 

 

 

6666 as. 866 Sam

3:6 .Hm e86 .68

w .66 \- .2.

~65: "x ~86 ax
~32 8~6 _ $2 6~e.~ 82 :5an 852 -m6 _ :36 :3 $2 8388
86 866 _ £66 686 6666- :8588 - - - - - - “8:380
866 366 2 6666 8~6 ~666- 395 866 3.6 _ :6 5~6 .86. Season
486 ~56 _ 686 83 :66- mmm 866 866 _ ~86 686 ~86- 88
~86 $66 _ 366 «86 866. 85 8m 866 866 a :66 6-6 866- ...5 8m
:66 866 _ $~6 :26 866- 82? 6666 666 a 686 E6 _86- 82:.
82 866 _ 426 686 6:6 oassfifi 8: 866 H 83 :66 :6 35536.6
663 £66 _ 866 E6 6666 oaoeoasms E: 6666 _ 686 86 ~86 oaouoaafi
86 ~86 _ 68.6 86 636. ass: .82 686 866 6 ~86 26 686- .32: o»...
8.5 05 8 25$ .3 m 8.5 .68 .6 255 6.6 m

mtg—z «So—32

.3515 :38... and 3 338.26% anus-— 3.58.62 88 2.3

152

 

..lll

 

 

8.6 "a :66 ..Nm

23 .5 636 am

2 .6... 6 .6...

62.2 ~x 636 ~x
68.2 ~36 _ v-6 ~6v.~ §~ 8388 $3” 366 _ v26 ~6~.~ 662 3588
63.6 636 _ -~6 $46 :66- 68588 - - - - - - .8588
366 £66 _ 6666 3~6 ~666- SB :5 -~6 ~6~6 _ ~26 -~6 62 6- 25 “on
686 :66 _ _ S6 83 _~66- mmm 866 6666 _ :66 83 366- mum
6666 ~86 6 ~86 -~6 :66- 3 8m 556 366 _ :66 _-6 366. an 8m
6666 636 _ 866 -~._ :66- gmfi 5: 6666 6 6:6 SS 83 gm fi
366 -66 _ 866 E6 266- :9; ~36 6&6 _ ~26 626 «86. so 3
~62 E~6 _ 32 ~26 ~26 mm; a: -~6 _ ~62 626 ~26 mm fi
662 £66 _ 866 686 2. _ 6 835 365 a: 366 _ 6-.~ ~86 E 6 335 was
~66.“ $66 6 ~86 ~: 6 ~86 8860 :53 ~86 :66 _ 3~6 626 ~86 cacao 3586
~86 -66 _ -3 62 6 6-6- :59: ow< -~6 $66 _ 6:..~ 3:6 3~6. =53 ow<

~25 .fim z. 235 6.6 m 3.5 .umm S 255 6.6 m
m .282 < .232

 

ézus team .2: a. 2.282»:— zfiufi 33232:: 6m~ 2......

153

stressful life events independently (Table 25b). Although the pseudo R-square increased
30-40 percent across the models, the model chi-square still failed to achieve statistical
significance. None of the risk factors significantly predicted recidivism and the
standardized slope for negative school behavior was in the unexpected direction.

The addition of group membership into Tables 25a and 25b (Model B in Table
25) presents a few minor clarifications to the findings above. The addition of group
membership into the models added some explanatory power and total intake placement
score also developed into a significant predictor of recidivism afier its addition into the
models. For both Tables the variable denoting group membership itself was also a
significant predictor of recidivism. Members of the concurrent control group were
approximately 60 percent less likely to recidivate, a finding contrary to what was
expected. However, delinquency risk factors do not have significant effects on official
measures of recidivism regardless of treatment modality.

The models in Tables 26a and 26b presents the relationships between background
variables and changes in delinquency risk factors.47 It is possible that positive outcomes
are only expected in the situations where positive change is evident in the factors
conceptualized to cause the problem, in this case the delinquency risk factors. Table 26a
includes an index measure for all change scores as an independent variable, and the drug
use change score is included as the only change score in Table 26b because that was the
only score where the groups significantly differed (see Table 21).

Similar to Tables 25a, neither models nor any of the individual predictor variables
reached statistical significance in Table 26a. The variables closest to reaching were age

at instant offense (p=.055) and total placement score (p=.079). Model B considered the

 

‘7 Change scores for each risk factor computed with formula Change = Tx3-Tx1.

154

 

 

 

~86 .5. ~86 .82

~26 .~_~ 8~6 .5

~ 8 ~ 8

68.: ~x ~8.~ ~x
~82 8~6 2 82 :3 8~.~ 3388 88.2 «86 _ ~86 83 v-.~ 8388
886 866 _ ~_~.~ ~86 v8.9 8:880 - - - - - - .5588
~86 ~86 _ ~86 83 266- 2395 ~86 -v6 _ ~86 _-6 83. 25.5
~86 ~86 _ 686 ~86 ~86- 88 $66 ~6~6 _ ~86 -~6 ~86- mmm
86 ~86 _ 86 -~6 686- an 8 ~86 ~86 _ :66 83 ~86- an 8m
.86 ~86 _ ~86 6:6 266- ”~55 :5 ~86 686 _ 286 -_6 86. 0826 “7.2
8~._ 686 _ 8% 886 -_6 fisfisfi 8: 886 _ ~86 686 -_6 issfioe
682 686 _ ~86 E6 886 0880.32 «~62 ~86 _ .26 226 :66 08808882
8~6 ~86 _ 83. ~26 686- 83: a? -~6 ~86 _ ~8~ ~26 ~86- :52: 09..
8.5 .5 8 25$ .5 a 8.5 .ufi 8 23$ .m.~ m

5282 5252

 

.85 W5 8.39m “aw—=30 .3 no.5— 3380391 932qu 33.3.53: .aen 93:.

155

 

 

 

-26 .52 ~86 .52

~86 .228 ~26 .228

~ 8 ~ 8

~82 ~x 82.62 ~x

-~6~ 2-6 2 -~2 8~.~ -6.~ 25288 82 ~86 2 ~82 -~.~ ~2.~.~ 25288

-~6 ~86 2 ~2~.~ 2.2.6 ~82- 28588 28588

~86 2.-6 2 286 82 2.2.66- 2839 6-6 -~6 2 ~86 -~6 2-6- 93922

822 E6 2 226 83 826 88 ~82 -~6 2 ~86 ~86 ~86 88

~86 8~6 2 686 -~6 ~86- 222 8m ~86 -~6 2 686 2.-6 686- an 8~

-~6 286 2 -~.~ 23 ~86. an 85220 o-6 ~86 2 2-.~ SS -26- an 822220

22 ~86 2 8~.~ 286 826 2252222s§ -2.2 826 2 ~2.~.~ 286 226 32522228£

686 ~86 2 ~86 -26 686- 388022582 ~82 83 2 ~26 ~26 ~86 0880222382

-~6 ~86 2 ~82. ~26 -~6- 28282 -< -~6 686 2 -~.~ ~26 886- .8282 -<
825 .~2~ 8 2.255 2.28 22 825 .~2~ 8 2.35 .m.~ .2

222252 «.2822

 

.3» —u5 935 992.225 89 MEG "228892qu 932»?— 3222-2233—2232 den 0325.

156

same background and change variables but also simultaneously considered the impact of
the treatment modality. The model remained nonsignificant yet there was a modest
increase in the pseudo-R2. The addition of treatment modality resulted in a significant
relationship between the dependent variable and both age at instant offense and total
intake placement score. This suggests younger individuals and those with higher
placement scores recidivated more than their counterparts. Similar to the T1 index, the
change score index was not a significant predictor of recidivism, a cumulative finding
that indicates risk and change in risk is not important in the production of recidivism. An
interesting finding emerges in Table 26b. The change score for drug/alcohol use was the
only change score on which the experimental and concurrent control groups significantly
differed. Model B in Table 26b reaches statistical significance and most importantly
indicates decreases in drug and alcohol use significantly reduce recidivism by almost 25
percent. In contrast to the delinquency risk factors that demonstrated no relationship with
recidivism, the findings suggested changes in risk factors were important to determining
if individuals recidivated. The relationship between reductions in risk and reductions in
delinquency was particularly important for reductions in alcohol and drug use.

This research is also concerned with the extent to which the type and quantity of
program services mediates recidivism in an adjudicated juvenile population. It was
hypothesized that individuals receiving more intensive intervention services will be less
likely to recidivate. In their meta-analysis of 200 intervention studies, Lipsey and Wilson
(1998: Table 13.2) reported that nearly 50 percent of total explained variance in

recidivism was explained by the amount of treatment and the type of treatment received.

157

 

 

 

226 2:2 62 26 22.22

~86 .5 «86 .5

~ 8 ~ 8

-~62 22 ~22 22
-6.~ ~86 2 ~86 -~.~ 2-6 222588 -6.~ 8~6 2 686 8~.~ -~6 22588
82.6 .286 2 -~.2. 22.2.6 686- 8588 - - - - - - 28588
~86 866 2 ~86 ~86 686- 83 an 6-6 83 2 6-6 -~6 -~.? 928 222
68.2 ~2.~6 2 ~86 8~6 ~86 88 ~82 2.-6 2 -66 ~86 686 .88
~86 ~86 2 ~86 8~6 2.86- ...5 8m ~86 6-6 2 62 26 -~6 866- an 8m
8~.~ ~86 2 8~.~ -~6 886 88:8 238. -2..~ 686 2 -~.~ 22-6 8226 885m 233
82 ~86 2 2.2.~.~ 686 -26 329222 2328 -22 ~86 2 8~.~ 286 2 26 325222 2so~
~86 -~6 2 686 - 2 6 ~86- 3.2280 22582 ~82 2.8 2 82 6 6~26 ~86 3.22820 222582
6-6 ~86 2 8~.~ 826 2 3. 222882 -< -~6 ~86 2 2-.~ ~26 2-6- 25282 -<
825 .~2~ 8 2.2.23 226 22 825 88 8 2.2.23 228. m

22 2882 < 2232

 

.85 W5 892.com Eauwoum 222289231 282qu 3222-2952:: .5 35:.

 

158

The most notable differences between Models A and B in Table 27 with the previous was
that both reach statistical significance. In Model A, “total services” was the only variable
statistically significant, yet these findings were contrary to what was originally
hypothesized. For example, it was expected that individuals received more services
would respond more favorably to treatment but the data show clients that received more
programming services were over 2.5 times more likely to recidivate. The relationships
between age at instant offense and total intake score remained in the expected directions,
but the variables continue to lack statistical significance in the prediction of recidivism.
Similarly, the measure of social disorganization was negatively associated with
recidivism.

A few themes emerged from the combined findings across Tables 24-27. The
majority of the regression models did not explain a significant amount of variance in
official recidivism. Also, the addition of treatment modality generally increased the
pseudo R-squares but the increases were often modest. The hypotheses that successful
rehabilitation is contingent on the extent of initial risk, changes in risk, and types and
quantity of treatment were not supported. Initial risk did not matter in terms of
recidivism and treatment services actually appeared to increase recidivism. The only
finding consistent with what has been hypothesized relates to the relationship between
change in risk and recidivism. Although the risk change score index did not significantly
reduce recidivism, decreases in drug use did significantly reduce recidivism when it was

considered independent of other risk factor change scores.

159

Self-Reported Delinquency

Official crime reports are only one measure of juvenile involvement in
delinquency. Self-report surveys are other measures that offer some advantages to
official data. As a supplement to official data, self-report surveys can cover a wider
range of behavior and are not limited to those crimes that come to the attention of the
police (Osgood, O'Malley, Bachman, & Johnston, 1989). Although early comparisons of
self-reported offending and official statistics concluded significant disparities between
the two (Hindelang et al., 1981), subsequent studies concluded similar pictures emerged
when a careful attempt was made to develop adequate measures and focused on similar
behavior (Elliot & Ageton, 1980).48

Self-reported delinquency measures included in this analysis were reported by the
respondents at T3, or 12 months after referral. The descriptive statistics for T3 self-
reported delinquency are reported for both the experimental and concurrent control group
in Table 28. The mean scores for the measures were highly distributed. The standard
deviation for the experimental group was over 1.5 times the mean (9? =6.928) and 3.3
times the mean (f =56.907) for the concurrent control group which was indicative of a
larger distribution. T3 self-reported delinquency was not statistically different between
the experimental and concurrent control group, although the group mean for the
concurrent control group was 2.5 times that of the experimental. One possible
explanation for the lack of significant difference was due to the large distributions of the
group means. The natural log was computed in an effort to normalize the values and

reduce the distributions. Computation of the natural log did not normalize the measure, a

 

‘8 Self-reported delinquency measures reflect data collected at T3 and reflects indicators of reasonably
serious types of behavior (see page 84 for indicators).

160

result of a large percentage (30 percent) of the total population reporting no self-reported
delinquency at T3. Thus, the concurrent and experimental groups did not differ
significantly on levels of T3 self-reported delinquency.

As a follow-up to the descriptive analyses, the next step involved the regression of
the same series of independent variables from Tables 24-27 on T3 self-reported
delinquency. Independent of change in self-reported delinquency, it was important to
understand how delinquency risk factors and treatment modality affect self-reported
delinquency. The Models presented in Tables 29 are similar to Tables 24-27 except that
treatment modality is not regressed into the model separately for series of independent
variables.

In Model 1 of Table 29 the simultaneous effects of background variables
including age at instant offense, instant offense type, total intake score, social
disorganization index, socio-economic status index, days in detention, and treatment
modality were considered. The relationship between age at instant offense and T3 self-
reported delinquency was negative which indicated that younger offenders reported
higher levels of later delinquency. Also consistent with expectations, the relationship
between total intake score and T3 was positive and consistent with what was
hypothesized. Narnely those individuals with higher placement scores were more heavily
involved in delinquent activity. The variable denoting concurrent control group
membership was also positive. In contrast, the relationship between group membership
and self-reported delinquency was not significant but the direction of the relationship
offers a notable distinction with the official recidivism models. Finally, days in detention

was negatively associated with self-reported recidivism and indicated those in detention

161

Table 28. T3 Self-Reported Delinquency 13! Group Membership (n = 178).

Experimental Concurrent Total d.f. Mean Sq F Sig.

 

Untransformed

Mean 6.928 17.475 11.695 1 4877-081 3073 -081
Std. Dev 15.107 56.907 40.072

Range 092 0-378 0-378

Log

Mean .467 .530 .496 l -176 -480 -489
Std. Dev .564 .652 .604

Range 01.97 0-2.58 0-2.58

 

longer reported less delinquent behavior. The model, however, was not significant and
explained almost no variance in the dependent variable (R2 = .033).

In Model 2, the index for T1 delinquency risk factors was stepped into the
equation. The model summary remained non-significant but the amount of explained
variance increased slightly to .063. The relationships between the background variables
and self-reported delinquency remained in the expected direction, but there were
inconsistent relationships between the T1 delinquency risk factors and self-reported
delinquency. The negative relationship indicated those at greater risk at T1 reported less
delinquent behavior at T3. Also, after controlling for T1 risk, age at instant offense was
now significantly predictive of T3 self-reported delinquency. Thus, the relationship was
specified by the addition of risk factors and indicates age at instant offense was important
when considered in conjunction with risk.

The three T1 risk factors on which the experimental and concurrent control group
significantly differed were stepped into the regression equation in Model 3. As expected,
negative school behavior at T1 was positively and significantly related to T3 self-reported

delinquency and the standardized slope (B=.285) suggested it was comparatively more

162

Save 5 5565mm :3
8.5 5 552.25% ._

dorm 93ng @8585 mono? 33vaqu a

 

 

55. NM _. m2. 2:. Go. m8. 552:5 a
882 LS...- Lsfi .83 82 mm: "m
58. an: 88. Se. 8%. m3. 55:5 =82
bufifiav. 33.:

33:8. - - - - - 5655.0- 38-
: _ 3:8. 32 .58. E: .58. 5 328. a: :58.- 2 2 $28. 565680
28.58.- €838.- EQ: 8.- 88.58.- 68.58.- 3838.. 295 5n
38.82:. $8.58. $8.882. $8.52. 685:. - mmm
€8.52.- G858c $8.87 $83: _.- 88.52.- - 6.5 8m

- £88.58. - - - - an amuse

- - ...GSQSN. - - - 552 556m 5.56

- - - 32.x: 2. - - Am F

- - - $3.58.- - - an F

- - - .3353. - - mm :-

- - - - $3.52 .- - 555-2 52 E
88.58. 88.58. $88.88. 85:8. 88.58. 88.58. 355 55a
63.888. $3.58. $3.58. 33888. @338.- ..Geezo. 686:0

235.52.- .3352.- ..a352: 235.52.- 83.52.- agtflc 558. 6w<

.- u u 1 1 1 35800

e .252 m 5.52 v 5.52 n 5.52 N 5.52 _ 5.52

 

emu—5.55.2. octagon-u—om nu. 035.3.» «no—Eamon— .nemmmouuod mac .3 «Ban.

163

predictive of self-reported delinquency than the other independent variables. In addition
to T1 negative school behavior, age at instant offense still remained predictive of T3 self-
reported delinquency. Although not significant, stressful life events at intake were
positively related to self-reported delinquency. This indicated juveniles reporting more
stressful life events also reported more delinquent behavior at T3. Contrary to
expectations, T1 drug use was negatively associated with the self-reported delinquency.
The relationship between T1 risk factors and future levels of self-reported
delinquency was, according to the above findings, less clear than hypothesized. Models 4
and 5 test the extent to which change in delinquency risk factors affected recidivism.
Since the two populations were at similar risk as evidenced by the similarities in T. risk
factors, recidivism may be reflective of changes in risk. Change scores were computed to
reflect the difference between T3 and T1 risk factors that were measured by the structured
interviews administered to each participant. Positive values were indicative of increases
in risk and negative values reduction in risk, and it was expected that a reduction in risk
would reduce T3 self-reported delinquency. Contrary to some earlier findings, the model
summary was significant (p<.05) and the R2 =.123. Two of the independent variables
were significant predictors of T3 self-reported delinquency, age at instant offense and
change score index. The positive direction of the relationship between the change score
index and the dependent variable indicates those at increased levels of T3 risk report more
delinquency. This finding is consistent with the notion that reductions in negative
behavior would only be expected if the risk factors causing the behavior were changed in

a favorable direction.

164

One additional OLS regression model is analyzed where T3 self-reported
delinquency was regressed on the same background variables and the change score for
drug/alcohol use. Since Model 4 suggested change in risk was an important factor in the
“production” of recidivism, the purpose of Model 5 was to determine if the drug/alcohol
use change score is particularly important in the production of delinquency. This
specified change score as identified as particularly relevant since it was the only change
score for which the experimental and concurrent control group differed significantly.
The model reached statistical significance and the explained variance also increased
modestly to .132. The drug/alcohol use change score was significantly predictive of T3
self-reported delinquency which suggests that reductions in drug and alcohol use were
predictive of latter decreases in self-reported delinquency. It is interesting to note this
single change score results in a similar amount of predictive power as the index change
score.

Finally, the index value for total program services was stepped into the model
along with background characteristics (Model 6). Like the first two models, Model 6
explained very little variance in self-reported delinquency. Expected for age at instant
offense which continued to be negatively associated with T3 delinquency, none of the
independent variables were significant predictors of later delinquency.

Self-reported delinquency proved to be a complex phenomenon that was
explained by combinations of both where individuals start in terms of their “risk” for
delinquency, but also explained by how much positive change occurs during the time
they are under the jurisdiction of the Court. There were several important distinctions

between measures of self-reported delinquency and measures of official recidivism. In

165

comparison to previous models, Table 29 indicated that while the aggregate measure of
T1 delinquency was not important to T3 self-reported delinquency, negative school
behavior was a significant predictor. Moreover, although the total services index
predicted official recidivism it does little in terms of self-reported delinquency. Finally,
treatment modality consistently had little predictive power of self-reported delinquency.
This stands in sharp contrast to Models 24-27 where treatment modality consistently
predicted recidivism. ’
Summary

Hypotheses were proposed throughout this analysis positing that members of the
experimental group would react more favorably to intervention strategies upon
adjudication to the community-based treatment compared to those youth in the historical
and concurrent control groups. Contrary to what has been hypothesized throughout this
research, risk factors, change in risk factors, and levels of intervention programming did
not seem to be as important in the production of recidivism. Moreover, in several of the
cases where a significant relationship was evident its direction was in the unexpected
direction. For example, the negative slope for T1 risk factor index in Table 25a and
negative slope for T1 drug use in Table 15b initial risk actually reduced recidivism as
opposed to increasing recidivism. Similarly, Tables 26a and 26b also presented
paradoxical conclusions that change scores were negatively associated with recidivism.
This was considered paradoxical because negative change score values (T3-T1) would
indicate a reduction in risk, something expected to reduce recidivism. Yet this finding

suggests reductions in risk actually increase recidivism, a finding consistent when

166

changes in drug/alcohol use are considered independently. Again, contrary to
expectations, total program services are also positively associated with later recidivism.
To add further confusion to the mix, reverse findings were evident when T3 self-
reported delinquency was considered as the outcome variable. In these situations,
delinquency was reduced for individuals with lower initial risk, especially lower levels of
problematic school behavior (e. g., negative school behavior). However, the unexpected
effects of T1 drug/alcohol use were consistent. Additionally to the finding that initial risk
may be important considerations in later self-reported delinquency, change in risk may
also be an important consideration in this situation. Individuals with more favorable
change, especially favorable change in drug use, may very well have reduced levels of
later problematic behavior. Finally, consistent with official recidivism models, total
services was positively associated with T3 self-reported delinquency yet the relationships

are not significant.

167

Chapter 5: Conclusion

The United States has experienced several distinct “waves” of juvenile crime over
the past 100 years. Since there is no objective standard for what constitutes a crisis of
crime, a perspective on the relative seriousness of a particular crime problem is created
by comparing contemporary crime figures with those of years gone by. It is generally not
absolute figures that demarcate a crisis of one kind or another, but the relative extent of
the problem when compared to previous periods.

In one sense defining crises or problems in the relative sense is effective because,
by definition, it provides a barometer for the problem. If our society is based on the
notion of continual “progress,” progress can be best measured by assessments of
innovations between two time periods. However, the strategy has also resulted in a near
universal condemnation of each successive generation of youth as substantially “worse”
than the previous and demands for social policy that effectively corrects the problems of
errant youth. Commentators argue fervently for a return to the “good old days” when
youth were behaved and reverent of their social responsibilities. Paradoxically, reviews
of media sources fiom the “good old days” reveals a social climate where out of control
youth were similarly identified as pressing social ills (Bernard, 1992). For example,
Platt’s (1991) account of the impetus behind the Child Saver’s Movement of the early
twentieth century provides meaningful insight into these issues.

Yet is it these relative comparisons of crime problems that drive public policy —
policies intended to obtain some nostalgic sense of yesterday. This should not be a
surprise because public policy usually deals with problems that communities experience

“today” and not those of previous generations. Moreover, crime policies are usually also

168

not effective in addressing unanticipated problems of the future. The continual stream of
state and federal crime legislation exemplify this reality, a reality not likely to subside
any time too soon (see Olson-Raymer (1984) and McGarrell (1988) for further discussion
on post-19603 crime legislation). Since the early twentieth century, the types of
responses have vacillated to some extent or another between punishment and treatment
depending on both the political ideology of the time and the “scientific” understandings
of the “causes” of delinquency.

The purpose of this research has been to assess the efficacy of nonresidential,
community-based intervention programs for adjudicated juvenile offenders. Non-
residential, community-based alternative treatment models are often touted as some of the
most promising initiatives in juvenile justice.

“Community-based programs offer several advantages above

and beyond the traditional adult treatments that are often applied

to juvenile offenders. First, juvenile offenders are adolescents-

they are still children-and they should be treated differently from

adults. Second, these children are products of the communities

in which they live and deserve to have the community care for

and be responsible for them. Third, incarcerated inmates,

whether child or adult, learn strategies for being a more skilled

criminal. A fourth advantage to community-based approaches is

that they can be more cost-effective than institutional

approaches” (Stanton & Meyer, 1998: 205).
Keeping and treating children in the communities from which they come offer the
possibility of reducing recidivism, limiting the “harm” done to juveniles through
exposure to the horrors of institutionalization (see Miller, 1991), and building on existing
connections with social institutions such as families, schools and communities. Yet

Stanton and Meyer (1998) aptly warn that while there might be some intuitive advantages

to community-based alternatives, there is a need to determine their effectiveness in

169

producing the “desired outcome” associated with “treatment,” namely a reduction in
recidivism.

Afler demonstrating some initial success in rehabilitating juvenile offenders
community treatment models have been put forth as effective “intermediate
interventions” (Lundman, 2001). nonresidential treatment programs have proliferated
during recent years as the Office of Juvenile Justice and Delinquency Prevention (OJJDP)
and other governmental organizations have championed the need for intensive and
comprehensive services in community settings (Howell, 1995). OJJDP’s Comprehensive
Communities Program initiatives have been heavily influenced by social development
theory, a theoretical framework that has implications for both understanding the causes of
delinquency but also has implications for the structuring of intervention services.
Strategies that appear to be the most promising are those that provide “comprehensive
programs that include treatment strategies targeting the systems that affect the juvenile’s
development” (Stanton & Meyer, 1998: 221).

Delinquency is a product of multiple, interacting risk factors (Thomberry, 1997b).
Instead of searching for a single cause of delinquency, the implication of social
development theory is that individual involvement in delinquency increases among those
who experience multiple attitudinal, behavioral, and contextual stressors (risk factors).
More importantly, the impact of any specific stressor or group of stressors is
developmentally specific and thus possibly inconsistent throughout the life course. For
example, association with drug using peers is likely to be more damaging at age 8

compared to age 38.

170

Several goals were established at the onset of this research initiative. The first
goal was to trace the development of the treatment ideology in the juvenile justice
system, and ultimately set the stage for why many believe community-based alternatives
to traditional probation services can be one of the most effective treatment strategies for
young offenders. The second goal of this manuscript has been to argue the form and
fimction of responses to juvenile crime are directly related to the governing theoretical
understanding of the “causes of delinquency.” For example, when the family
environments of delinquents were thought to offer nothing but criminogenic influences,
juvenile delinquents were universally moved to institutions to “protect” them from
negative influences. Using social development theory as a theoretical foundation for the
type of intervention services advocated by OJJDP, the argument was advanced that the
effectiveness of intervention programs in reducing delinquency is contingent on the
extent to which it is developmentally specific and delivers program services that address
the causes of delinquency.

Using data collected as part of the Youth and Family Studies (YF S) intervention
program for adjudicated juvenile offenders, a framework was established for testing the
efficacy of community-based alternative treatment programs compared to residential
placement programs. This research is a necessary addition to the current literature
because it moves beyond traditional program evaluation models. The classic evaluation
design details a program model and then typically tests the effectiveness of one program
over another in offsetting recidivism. In this scenario, several assumptions are made.
First, there is an assumption anything actually “happened” as part of the program and that

the participants received the program as described in its ideal form. The second

171

assumption is program services create favorable change in the delinquency risk factors
that underlie the program’s theoretical foundation. Third, there is also an assumption that
change in delinquency risk factors will reduce recidivism.

These assumptions lie at the core of this research. For straightforward,
compressed programs these concerns might not pose significant problems. For example,
assumptions that programs are true to their ideal form may be sufficient in many cases for
a short-term program where everyone receives exactly the same service. In these
scenarios it is less likely major program failure will occur short of the program not being
administered. Take, for example, the case of the challenge course model where clients
report to a predefined location and spend one or more days involved in team building
exercises on a military-style obstacle course. Similar “off the shelf” type programs can
be structured to the point where there is little concern about the level of variation in
services received among individuals. However, program failure is more likely in the case
of long term, intensive programs that tailor services to fit individual needs.

Summary and Discussions

Social development theory offers several intrinsic advantages over other theories
of delinquency. The theoretical model offers advantages not only because it is a more
complete picture of the onset and trajectory of delinquency careers, but because the
model also offers implications for developing and implementing intervention services for
juvenile offenders. Social development theory is not a theoretical framework that posits
distinct direct and indirect relationships between a unique set of independent variables
and delinquency, but instead integrates the most prominent existing theories into a more

complete theoretical model. Delinquency is best represented by the interaction of

172

multiple domains of risk factors that originate from individual, social, and ecological
influences. Moreover, the relative importance of a particular risk factor or set of risk
factors is contingent both on when in the individual’s development the risk factor is
realized and the extent to which certain risk factors co-occur. Social development theory
also offers implications for how intervention treatment services should be delivered to
delinquent youth. . To be effective, stage-salient services must be delivered that intervene
in the delinquency “process” by providing intensive, sustained services that offset the

impetus for delinquency.

Impact of Delinquency Risk Factors

Delinquency is a product of discernible risk factors that increase the likelihood of
initial participation in delinquent activity. The independent and cumulative effects of
delinquency risk factors explain not only the onset of delinquency, but also levels of
sustained involvement and escalation in delinquent careers (Blumstein et al., 1988b;
Blumstein et al., 1986; Blumstein et al., 1985; Farrington, 1992).

The experimental group, historical control group, and concurrent control group
were compared on delinquency risk factors using both the Court intake placement risk
classification data and the delinquency risk factors measured as part of the structured
interviews. The purpose of these analyses was to determine if characteristics of risk and
change in risk predicted later levels of official recidivism (18 months) and self-reported
delinquency (12 months) after court referral to the Department of Social Services. Of the
three groups considered in the initial set of analyses (experimental, and historical and
concurrent control groups), the concurrent control group was at greatest “risk” based on

the initial placement scores. There were some significant differences in risk between the

173

three groups evident in the Court intake data (see Tables 12a and 12b). Members of the
concurrent control group were first arrested at significantly younger ages compared to
both the historical control group and the experimental group. A larger percentage of the
concurrent control group also had prior arrests, had difficulties in school, used drugs, was
arrested for more serious offenses, were on probation at the time of arrest, had a prior
FIA placement, and were previously in a runaway status. However, it was expected that
there would have been greater differences between the three groups apparent in the intake
classification data and, in general, the differences were less than expected. From a
theoretical point of view, it was still important to determine if these differences in risk
explained recidivism.

Based on these differences it was hypothesized the concurrent control group
would respond less favorably to intervention services than others due to apparent
increases in risk, yet this was not the case. Compared to the experimental and historical
' control groups, the concurrent control group has the largest proportion of non-recidivists
(see Table 14). Of all of the variables considered in the preliminary recidivism (Tables
17-19) models, age at instant offense was the only variable that consistently demonstrated
a significant relationship with recidivism. Since behavior can be characterized by
developmental progression, this finding was consistent with the social development
perspective. From this standpoint, there is “normal” sequencing to behavior that makes
its occurrence and temporal ordering relevant to certain developmental phases (Loeber,
1988). Behavior inconsistent with expected temporal ordering can be classified as
abnormal and expected to hold the possibility of adversely affecting normal development.

The impact of initiation into delinquency is not even across all developmental phases.

174

The significant relationship between age at instant offense and recidivism implies
younger offenders and older offenders initiated into delinquency at younger ages should
be considered more serious and subject to additional scrutiny by the juvenile justice
system.

The predictive power of the intake placement variables was more limited than
expected. Except for drug use and offense seriousness, none of the other intake
placement variables approached statistical significance for any of the models. The lack
of meaningful relationships may be partly caused by their limited variability or due to
problems with multicollinearity. To account for these problems, further analysis was
performed on delinquency risk factors coded from the structured interviews (see Table
20).

Findings from Table 20 were similar to those in Table 14. The concurrent control
was considered more at risk compared to the experimental group on a few of the T1
delinquency risk factors. More specifically, they reported significantly more negative
attitudes toward school, alcohol/drug use, and stressful life events. Although not
statistically significant, members of the concurrent control also reported higher levels of
exposure to delinquent peers. Again, it was surprising the differences between these two
groups were not more apparent on other measures included in the analyses.

It was evident that T; delinquency risk factors had little effects when regressed
into prediction models. Even when controlling for T1 risk, age at instant offense
remained the most predictive of recidivism (see Tables 25 and 26) and self-reported
delinquency (Table 29). The only relationship continually supported in the current

research was the relationship between age at first arrest and delinquency, a finding

175

indicating earlier offending was predictive of recidivism. Younger juveniles in the study
groups were consistently more likely to recidivate. This finding complements those by
Blumstein (1985), Loeber (1996), Loeber, Farrington, and Washbush (1998), and
Thomberry (1997b) that consistently identify age as one of the most stage-salient
considerations. In this case, the research is extended to suggest that earlier delinquency
not only predicts involvement in the criminal justice system but also predicts recidivism.
However, contrary to what was expected, baseline delinquency risk factors actually
explained very little in terms of variance in later delinquency.

Somewhat different findings were evident when T3 self-reported delinquency was
considered as the dependent variable (see Table 29). When stepped into a OLS
regression model with the background variables, the T1 risk index was negatively
associated with later self-reported delinquency. However, when the three T1 delinquency
risk factors were considered independently the findings suggested increases in T1
negative school behavior significantly increased T3 self-reported delinquency.

Although the findings relating to the role of substance abuse was inconsistent
throughout this research, its role in the initiation and progression of delinquency presents
important policy considerations. Juvenile substance abuse poses a serious problem to the
juvenile justice system. There are many ways the use of illegal substances is connected
to crime (F agan & Pabon, 1990; Goldstein, 1995). In some situations pharmacological
effects of substance use results in aggression not present absent drug use. Since drug and
alcohol, especially among younger teens, tends to be concentrated among a small
percentage of the population, its use may bring people into contact with those already

predisposed to delinquency.

176

Many delinquents caught up in the juvenile justice system use illegal drugs at
least occasionally (Dembo, Schmeidler, Nini-Gough, & Manning, 1998; Grobsmith,
1989; Hawkins et al., 1992). Although alcohol/drug use is not a sufficient precursor to

9, many youth involved in delinquency use alcohol or

more serious delinquent behavior4
drugs at least occasionally (ADAM, 2000; Dembo, Dertke et al., 1987; Dembo et al.,
1998; Johnson, 1995; Kelley, Huizinga, Thomberry, & Loeber, 1997; Lipsey & Derzon,
1998). For the developing young adult, drug and alcohol abuse undermines motivation,
interferes with cognitive processes, contributes to debilitating mood disorders, and
increases risk of accidental injury or death” (Hawkins et al., 1992: 64).

There are discernible individual, social, and ecological risk factors for alcohol and
drug abuse that often share the same etiological causes as other forms of delinquency.
Similar to findings from previous research, alcohol and drug abuse were dependable
predictors of recidivism throughout this research. Juveniles using illegal substances
experience more problems with their families and many report other types of negative
consequences such as missing school, having automobile accidents, becoming depressed,
or getting into trouble with the police because of their drug or alcohol use (Dembo,
Dertke et al., 1987).

Other delinquency risk factors that were operationalized from the T1 interviews
were excluded from further analyses because of problems with multicollinearity.
Although the relationships between these and risk factors and delinquent behavior were

not included, preliminary analyses revealed insight into the relationships between these

and delinquency. Contrary to previous research clearly establishing consistent

 

‘9 Alcohol/drug use in many cases may by definition be considered delinquency. Measures of drug use, for
example, are often included on self-reported delinquency scales. In this case alcohol/drug use is considered
independent of other delinquent behavior such as property or personal crime.

177

relationships between peer networks and delinquency50 exposure to delinquent peers
almost universally failed to predict onset of recidivism. The null finding was puzzling
because the argument is generally accepted that delinquency is a group phenomenon
(Elliot & Menard, 1996; Erickson & Jensen, 1977 ; Farrington, 1996; Warr, 2002) and
that middle adolescence is the height of peer importance (Warr, 2002). Since the mean
age of the population is approximately 15 years (see Table 8), peers would be expected to
play a more important role in the production and trajectory of delinquency.

Delinquent peers can be expected to influence delinquent patterns through one of
two paths (Elliot & Menard, 1996). Using the social learning approach, the first suggests
exposure to delinquent peers leads to greater delinquency (Burgess & Akers, 1966;
Sutherland & Cressey, 1978). In this case delinquency is learned through a process of
regular interaction with individuals more knowledgeable of the tricks of the delinquency
trade. This represents the traditional mode of thought that delinquency is a learned
behavior. The second suggests involvement in delinquent behavior precedes regular
association with delinquent peers (Glueck & Glueck, 1950; Hirschi, 1969). This is based
on the argument that crime is a simple process that requires little specialized knowledge
or skills (Gottfredson & Hirschi, 1990) so learning the tricks of the trade is unnecessary
because there is little to learn. As juveniles become repeatedly involved in delinquency
their conventional ties to society are dramatically reduced. During the process of
increased exposure and involvement in delinquency youth either reject conventional
norms or choose to ignore them. Elliot and Menard (1996) favor the first approach,

especially among younger delinquents. The temporal ordering of delinquent peers and

 

5° Warr (2002) provides a comprehensive assessment of the theoretical and empirical relationships between
peer networks and juvenile delinquency.

178

delinquency could not be discerned here because longitudinal data be_for§ the onset of
delinquency were not available.

It was also surprising that school level variables mattered very little in terms of
recidivism or T3 self-reported delinquency. Mid-adolescence is a time when schools play
a prominent role in the lives of juveniles (Gottfredson, 2001). Juveniles with little
bonding to their schools report higher levels of delinquency (Maguin et al., 1995), and
those involved in minor forms of negative school behavior such as truancy also report
higher levels of self-reported and official delinquency (Farrington, 1989). Outside of
families, schools represent one of the strongest conventional attachments students have to
their community (Hirschi, 1969). Institutional level characteristics such as school climate
have strong influences not only on aggregate crime rates, but also influences individual
delinquency (Farrington, 1993; Olweus, 1993; Sharp & Thompson, 1994; Smith & Sharp,
1994)

There are also two other possible alternative explanations for the lack of
consistency in the predictive power of the risk factors for both official and self-reported
delinquency outcome measures. It could be that juveniles become “better” criminals,
perfecting their involvement to the point where they simply do not get caught. In this
case levels of delinquency are not reduced but remain undetected. However, the
impulsive and unskilled nature of crime, especially juvenile crime, suggests this is
probably not the case (Gottfredson & Hirschi, 1990). A more viable explanation is that
low level offending is just more common among all juvenile offenders and that 18
months follow-up for official delinquency is not sufficient to get at the full picture of

recidivism.

179

The delinquency risk factors from the structured interviews added very little
predictive power to the logistic regression models. Although the vast majority of
delinquency research would suggest these variables, in isolation or in combination with
one another, should be important predictors of recidivism no clear picture emerges. Most
models failed to achieve statistical significance, and in general, accounted for little
explanatory power. These findings stand in striking contrast to considering the predictive
power of the initial intake placement variables (see Table 18). The intake placement
scores are analogous to the delinquency risk factors. It was originally believed these
variables, although important, were not the best measures because of their limited

variability.

Impact of Change in Delinquency Risk Factors

If delinquency is understood to be caused by particularly criminogenic risk
variables in the domains of attitudes, behavior, the social world of individual offenders,
how should we understand “success” to be measured? To understand delinquency as a
product of a multitude of causes implies success, at least at some level, and must be
measured by favorable changes in the risk factors program services are intended to “fix.”
Success then is multi-dimensional and can be measured using a variety of outcome
measures that would be expected to change on account of the type of “response” initiated
for the problem. If the theoretical foundation was properly specified, we would not
expect reductions in levels of delinquency unless there are significant reductions in the
causes of the behavior (e.g., delinquency risk factors). Consideration of delinquency risk

factor change should be a critical piece of any evaluation research where the opporttmity

180

exists. OJJDP’s “risk-focused” prevention and intervention strategies hinge on reduction
of risk, reductions expected to reduce delinquency (Howell, 1995).

The direction and significance level of the change score index indicated the
variable mattered little in understanding recidivism (see Table 263). Contrary to what
was expected, not only was the relationship between initial risk and recidivism in the
unexpected direction, so was the relationship between change in risk and recidivism. In
this case, individuals who experienced increases in risk were less likely to recidivate.
The variable also failed to reach statistical significance regardless if treatment modality
was included or excluded. The relationship between the change score index and T3 self-
reported delinquency are quite different. First, the relationship is positive and, as
hypothesized, indicates those with greater reductions in risk were less delinquent in the
future. This is supportive of the argument that reductions in risk reduce delinquent
behavior.

The findings presented above do little to clarify the relationship between changes
in drug/alcohol use and delinquency. The findings in Table 26b suggested official
recidivism actually increased as drug/alcohol use decreased, a finding that was
completely counterintuitive. While the relationship was not significant in the first model,
it reached significance in Model B when treatment modality was simultaneously
considered. The relationship took on a different dynamic when T3 self-reported
delinquency was regressed on the variable in Table 29. In that case, decreases reductions
in drug/alcohol use resulted in significantly lower levels of self-reported delinquency.

There is little doubt to think that reductions in drug and alcohol use would serve

as a positive change on the fiiture life course of the user. However, the magnitude of the

181

change remains unclear and, in this research, was contingent on which dependent variable
was used. It is unclear if the damage of substance use is effectively done once substance
abuse is initiated, or if it is possible to steer juveniles away from drug using behavior
after use has begun. The concurrent control group reported a significantly greater
reduction in substance use but it was not possible to disentangle if the reduction was a
result of reduced opportunity to use due to the residential nature of the program or if the
treatment strategies themselves were more effective

As suggested earlier, the discrepancies in the conclusions reached about the
predictive power of risk factors and changes in risk factors when different outcome
measures were considered may be partly explained by the very nature of the outcome
measures. Although the measure of self-reported delinquency reflects relatively serious
behavior, official recidivism generally reflects some of the most serious behavior. In this
case, it takes a certain level of seriousness, harm, and consistency for problematic
behavior to be “discovered” and punished by actors in the criminal justice system. For
example, arrests are made only 10% of time when a serious crime occurs.51 A longer
follow-up would likely to have revealed increased instances of official recidivism, but the
limited period yielded populations where approximately 33 percent were considered
recidivists. It could also be the inherent relationships between the independent variables
and self-reported delinquency can be more easily detected because of the manner in
which it was operationalized. The larger distribution of the dependent variable
(compared to binary coding scheme for official recidivism) was likely more sensitive to

change.

 

5 ' Of 1,000 serious crimes that occur, 500 go unreported and of those reported, 400 remain unsolved (US.
Department of Justice: Bureau of Justice Statistics, 1999).

182

Another possibility is that official recidivism and self-reported delinquency were
two conceptually distinct types of behaviors and thus not subject to the same causal
influences. Some have argued self-reported and official measures are dissimilar
measures and should be not used interchangeably (Tittle, Villemez, & Smith, 1978).
Hindelang, Hirschi, and Weis (1991) questioned this argument and instead reported both
types of measures effectively tap similar domains of behavior.

The self-reported delinquency measures used in the current research reflect
reasonably serious crimes. For example, the indicators include “robbed someone”, “stole
something over $100,” and “forced someone to be sexual with you.” These indicators
were serious indicators of delinquency and would be expected to be highly correlated
with official measures of delinquency. The two most nonspecific indicators in terms of
seriousness were “hit someone in anger,” and “damaged property that wasn’t yours.”
While the self-reported scale is open to criticism because it may have covered too wide a
range of behavior, delinquency is a multi-dimensional type of behavior. It is believed
that the range of behavior covered on the self-reported index makes the scale analogous
to official measures of delinquency, thereby mitigating the impact of this explanation.

The final possibility is self-reported delinquency was more sensitive to the
independent variables because the behavioral measures are more similar. In this
possibility, independent measures such as negative school behavior and alcohol/drug use
may be measuring the same phenomenon and thus would be expected to be related. This
was a possibility, yet to account for this potential problem all measures of low level

delinquency and drug use were intentionally excluded from the T3 self-reported

delinquency scale.

183

The fact that juveniles exposed to the YFS program experienced reductions in
delinquency risk factors similar to the concurrent control group, neither consistently more
nor consistently less, could also be interpreted as a positive. While it was hypothesized
the community-based alternative would be more successful in reducing risk since it was
both comprehensive and kept youth in a familiar environment, youth in the experimental
group were less serious offenders to begin with. Since the community-based experienced
less risk at the beginning of the program, there was an inherent cap on the level of
positive change they could experience. Taken to an extreme, it could be the community-
based group did not experience reductions in risk because they did not have substantial

levels of risk to begin with.52

Impact of Program Efforts

There was reason to believe the quantity and type of intervention services
youthful offenders received would be an important predictor of desistance from future
criminality. Treatment matters, and those who received more treatment services directed
at the etiology and trajectory of delinquent behavior were expected to experience
substantially less occurrences of delinquent behavior. Comparing differences in mean

effect sizes for treatment and control groups, Lipsey and Wilson’s (1998) meta-analysis

 

’2 Maltz, Gordon, McDowall, and McCleary (1980) addressed the possibility of statistical artifacts of
pretest-posttest research designs. Reductions in T1 delinquency risk factors, self-reported delinquency, and
official delinquency are possibly explained by the concept “regression to the mean,” a statistical artifact.
The data presented suggest the concurrent control group were significantly more serious offenders. This
finding was substantiated by initial intake placement data and T1 delinquency risk factors. It was also
evident based on the higher intake placement score that the concurrent control group included more
individuals with more serious prior offenses and a greater total number of offenses. Thus, members of the
concurrent control group could be considered as existing on one end of the continuum of offending. Since
this more serious group was near the “ceiling” they had no place to go but down. If this was so, the
changes would be a statistical artifact. It seems likely that “regression to the mean” may very well explain
part of the findings detailed in this research. However, the significant reductions in delinquency for the
historical control group, a group presumed to closely resemble the experimental group, indicate there are
substantive reasons to believe the reductions in recidivism are also at least partly explained by
characteristics of the treatment modalities.

184

of 200 previously published studies found differences were best explained by four
different clusters of variables: characteristics of juveniles, amount of treatment, treatment
type, and general program characteristics.

According to Lipsey and Wilson (1998), the combined effects of the four clusters
of variables for non-institutionalized youth explained a significant proportion of variation
in the mean differences in effect sizes. Juvenile characteristics explained the largest
proportion (.40), followed by treatment type (.26), amount of treatment (.20), and general
program characteristics (.15). In total, the combined effects of the clusters explained
55% of the variation (Lipsey & Wilson, 1998: 320). The most effective types of
treatment included interpersonal skills, individual counseling, and behavioral programs.
Programs that used a combination of the above treatment services demonstrated
meaningful effects on the standardized mean differences, but to a lesser degree (324).

There are marked contrasts for treatment programs targeting institutionalized
offenders compared to those for non-institutionalized offenders. Among the four clusters
identified earlier, characteristics of juveniles accounted for the smallest proportion of
explained variance (.10). The largest proportions were accounted for by general program
characteristics (.36), amount of treatment (.27) and treatment type (.26). Although the
cluster of variables explained a similar amount of total variance (R-square = .51), the
clusters for institutionalized offenders behaved differently (328). For institutionalized
offenders, interpersonal skills treatment continued to be most beneficial, followed by
family services. The authors concluded types and levels of treatment are important
measures for explaining the success of some programs over others in rehabilitating non-

institutionalized and institutionalized juvenile delinquents.

185

Based on these findings, it was hypothesized individuals receiving more
programming would experience less recidivism than their counterparts. This is important
because it further questions if the level of “failure” for rehabilitative efforts in the past
was a result of the inadequacy of rehabilitation as a concept, or because individuals that
failed to demonstrate “success” did not receive the same level of services than others.
Indicators of services received were self-reported by the participants during the structured
interviews. There were reasons to expect the concurrent control group might receive
more services than the experimental group because those participating in residential
pro grams are in the company of treatment specialists throughout the entire day. As both
a literal and figurative “captive audience,” those in residential placement might be
expected to have more time available for treatment services and receive more intensive
supervision throughout the day since they do not leave the confines of the treatment
facilities.

Yet there was also reasons to expect the experimental group would receive more
program services. The YFS community-based day treatment model was conceptualized
as a comprehensive service delivery program able to provide directed services in a more
effective and efficient manner. Justified as a program designed specifically to deliver
comprehensive treatment services, it would be plausible to hypothesize services to be
more intensive. The data suggest few notable differences between the experimental and
concurrent control groups in the level of services received (see Table 22). For example,
both groups received statistically indistinguishable levels of counseling, educational, and
recreational services. The concurrent control group did, however, report significantly

higher levels of job skills training. It is not clear if the differences were due to one

186

specific program administered to those under traditional probation or a conglomerate of
services administered across several different programs. The concurrent group did report
significantly more job skills training, but the impact of employment related services is
mixed. Lipsey and Wilson (1998) reported the effect of employment related services to
be mixed by generally positive for noninstitutionalized offenders, and weak or
nonexistent for institutionalized juvenile offenders (Table 13.8, 332).

In this research, subtypes of program services were aggregated into one summary
measure of program services (variable, “Total Services”) because of problems associated
with multi-collinearity. It would have been preferable to consider the independent effects
of the different measures, but it was not possible in this scenario. The results detailing
the relationship between program services and recidivism are difficult to explain. When
considered simultaneously with basic intake data, background variables and group
membership (see Table 27, Model B), the measure for total program services was
significantly predictive of recidivism but the relationship is in the unexpected direction.
Individuals who received higher levels of program services were over 2.5 times more
likely to recidivate. Similarly, the indicator of total program services was positively
associated with T3 self-reported delinquency, yet the relationship was non-significant.
Limitations of Current Research and Directions for Future Research

This research was not without limitations that should be considered when
interpreting the findings detailed previously. One of the most methodologically
challenging issues to this research has been the sample sizes, especially as they related to
the size of the experimental and concurrent control groups. Sample size is an important

consideration for detecting intervention effects because it is closely connected with

187

measures of statistical power such as significance tests and effect sizes. The samples for
this research involved 99 individuals considered members of the experimental group and
79 members of the concurrent control group. The limited numbers in each group reduced
the ability to perform additional analyses that may have firrther clarified some of the
relationships proposed earlier. For example, if the sample sizes were larger it would have
been meaningful to analyze the effects of the independent variables for the different study
populations as separate models instead of adding reference categories for population
membership. Analysis of separate models would have allowed statistical tests to
determine if the combined effects of the independent models behaved differently across
models (see Paternoster, Brame, Mazerolle, & Piquero, 1998).

Members of the experimental and concurrent control group were also not
randomly assigned to their respective groups. In this case, it would be difficult to discern
if these findings are generalizable to other groups of offenders. We know the concurrent
control group was different in terms of “risk” from the experimental group to some extent
so it was clear the selection process itself, as intended, funneled more serious offenders
into traditional probation services and less serious offenders into the experimental group.
It would be beneficial in an ideal sense to randomly assign groups of offenders to the
treatment and control groups, yet such random selection is often neither feasible nor
desirable from a community safety perspective.

It should also be noted that the juveniles included in this study were considered
relatively serious offenders who were state commitments. These juveniles were not
typical juvenile offenders coming before a court but those considered so serious that the

county court moved jurisdiction to State officials, and thus represented the tail end of the

188

continuum as some of the more serious juveniles. Including only the most serious
offenders most likely resulted in a restricted range on many of the delinquency scales.
Inclusion of less serious youth may have revealed different relationships between initial
risk, changes in risk, treatment effects, and recidivism. The lack of significant findings
especially as they related to measures of official recidivism might be explained

While the reader is cautioned that the lack of random assignment and restriction
in range in terms of seriousness of the population may limit the generalizability of the
findings in some fashion or another, it by no means invalidates the importance of the
findings. The design of the analysis plan controls for indicators of “seriousness” of the
offenders by including not only intake placement scores but also other background
characteristics and delinquency risk factors. It is these types of characteristics that often
differentiate more serious from less serious offenders. In this case, total risk score in
conjunction with offense type were the primary determinants between classifications into
the treatment versus control groups. Mandatory overrides to more serious classification
levels did occur in some instances because of serious family problems (see Appendix A),
but the placement guidelines were generally followed.

With the statistical controls considered in the models proposed in Chapter 4, these
findings are expected to be generalizable to a broad class of juvenile offenders. The
statistical controls included intake placement variables, instant offense type, background
characteristics, and T1 delinquency risk factors. The participants in the treatment and
control groups also included a wide range of offenders, most referred for serious offenses

(see Table 7).

189

The current research would also have been strengthened by more robust measures
of treatment services. Treatment services currently represented aggregations of self-
reported measures that reflected frequency of participation in certain types of services
across one or more programs. Respondents were asked to identify all of the programs by
name from which they received services during the identified time period. Respondents
were subsequently asked to indicate the frequency with which they attended individual or
group counseling services, educational tutoring, job-skills training, or recreational
services for each program identified (see “Program Treatment Model” in Chapter 3 for
fiiture discussion on measures).

When designing a similar program, it would make sense to develop agreements
with individual service providers to maintain logs of when individuals attend their
programs, the types of services received during the visit, and the duration of different
services (e. g., group therapy). This “official” measure of program services could service
either as a substitute for self-reported measures or as supplemental information. Where
possible, data collection procedures should also create finer distinctions between various
types of program services.

It would have been preferential to collect data from a non-delinquent group of
juveniles to compare to the delinquent groups. It was impossible to determine if the
youth included in this study were at one end of the continuum in terms of both initial risk
and change in risk or if their relative levels of risk were developmentally “normal.”
There were two core assumptions running throughout this manuscript. First, those at
greater initial risk would be more likely to recidivate because high levels of risk as

operationalized in the measures is not normal fi‘om the perspective of developmental

190

theory. Thus, those at greater risk would respond by participating in greater levels of
delinquency. The second assumption was that changes in risk were directly attributable
to services received from participation in the program. For example, the implicit
assumption was reductions in drug use for both populations were a result of effective
intervention services. Regardless if the reductions were attributable to reduced
opportunity to procure drugs or use drugs caused by increased surveillance, or
attributable to reductions in the psycho-physiological cravings (e.g., addiction) resultant
to therapy, it was assumed any apparent reductions are due to the treatment models
themselves.

There is a host of research that identifies “normal” developmental patterns in both
delinquent and non-delinquent behaviors (Elliot et al., 1985; Elliot et al., 1989;
Farrington, 1992; Furstenberg & Hughes, 1997 ; Loeber, 1996; Loeber & Hay, 1994;
Loeber et al., 1993; Thomberry, 1997b; Thomberry et al., 1994; Warr, 2002) and the
inclusion of a non-delinquent control group similar in age, gender, and socio-economic
status would provide the ability to see if T 1 delinquency risk factors or changes in those
risk factors were “norma ” issues that occurred in this age-specific deve10pmental stage.
Warr (2002), for example, provided a comprehensive assessment of how the changing
roles of peers and peer networks influence levels of delinquency throughout the life-
course. The use of a non-delinquent control group would advance the argument the
presence of T1 delinquency risk factors and changes in those risk factors are unique to the
study populations.

Finally, this study would have been aided by a longer follow-up for both the

structured interviews and measures of official recidivism. The eighteen-month follow-up

191

period was consistent with previous research investigating similar phenomenon, yet
following individuals over extended periods of time may have yielded relationships not
evident in the current analyses. Many longitudinal research designs consider the
relationships between early delinquency (before age 15) and criminal behavior in
adulthood. F arrington (1989), for example, reported early irnpulsivity diagnosed between
ages 8-10 to be predictive of self—reported violence at age 32. Similarly, Tremblay and
LeMarquand (2001) reported, among other findings, problem behavior (oppositional
behavior and physical aggression) during kindergarten predictive of involvement in both
property and violent crime at ages 11 and 12. Expansion of the follow-up would allow
for consideration of the issue of “staying power” of treatment effects, and consider the
long-term effects of T1 delinquency risk factors and change in risk factors on recidivism
for a larger part of the life course.
Policy Implications

Treatment in the juvenile justice system should be structured in manners that use
the least restrictive placement options that best facilitate goals of the system such as
reduced recidivism and reintegration into the community (Miller, 1991). Early treatment
of juvenile offenders should be one of the primary goals of the juvenile justice system. A
review of the literature leaves little reason to doubt that large-scale success in either
preventing delinquency or intervening after the onset of delinquent behavior has yet to be
realized. Private citizens, religious organizations, and government agencies have
invested undeterrninable levels of time and resources into the search for the “cure” for
delinquency. Since the early 19003 treatment strategies have taken many different forms,

some focusing on psychoses of individual offenders, others at peer networks or families,

192

and still others directed at social changes that increase the presence of legitimate
opportunities for residents of distressed communities. Despite the best of intentions, the
involvement of juveniles in crime, particularly serious crime, increased throughout much
of the twentieth century. While there has been some controversy about the accuracy of
official statistics, few can dispute that the involvement of juveniles in serious and violent
crime increased throughout this period. The most promising strategies involve
formulating intervention programs for younger offenders near the onset of delinquency
(Howell, 1998, 1995).

Recent policy initiatives have advocated the implementation of comprehensive
service delivery models that are developmentally specific and deliver intervention
services that realize the ontology of delinquency. The Office of Juvenile Justice and
Delinquency Prevention established a series of guidelines and published extensively on
model programs directed at this effort. Regardless of the exact form of the intervention
services, early and comprehensive intervention programs should be guided by the notion
of the least restrictive alternative. However, it is unclear at this point what “least
restrictive” actually means. Miller’s (1991) assessment of the inadequacies and harmful
effects of juvenile institutions in Massachusetts could be interpreted as support for a lack
of official involvement by the juvenile justice system. To do so would be a shortsighted
interpretation of this important work. Although Miller (1991) argued many young
offenders before juvenile courts today need little, if any, direct intervention by actors in
the criminal justice system, he also recognized there are a significant number of juvenile
offenders that do need intervention services and such services should be meaningful and

humane.

193

Yet the mandate for effective treatment in the least restrictive settings remains
reasonably broad. Evidence from the current study indicates the non-secured residential
model was more effective in producing the intended outcomes of reducing recidivism and
delinquency risk factors. Even in light of the fact that the clients in residential placement
were slightly more serious offenders, the treatment modality was more effective in
reducing recidivism. While those in residential placement received similar levels of
services during their commitment, their treatment was overall likely to include more
intensive supervision since their ability to leave the premises was restricted. This
provides support for the argument that the treatment modality itself matters and that all
interventions are not equal. In this case, “least restrictive” should not be interpreted as
meaning “non-restrictive.”

Policy should be concerned with establishing not only if programs work, but why
certain programs work while others do not. Determining why one program model is
more effective over another answers only part of the question and leaves the field of
juvenile intervention woefully lacking in viable alternatives. Lipsey and Wilson (1998)
advanced the argument that it is important to disaggregate treatment programs by
subtypes of services. The effectiveness of intervention programs is not only based on the
type of offenders, but the nature of the setting in which services are delivered
(institutional setting versus non-institutional setting) and the specific types of intervention
services delivered (educational, interpersonal counseling, family therapy, etc.). In the
case of this research, the global measure of program services was significantly related to
measures of official recidivism, yet the relationship was in the unexpected direction.

Meaning, those who received more services were also more likely to recidivate.

194

Related to the above, it is also important to determine if failure was caused by a
flawed conceptual foundation characterized by a misspecified understanding of the
problem or caused by implementation problems. There is no doubt that the Youth and
Family Studies program suffered from major implementation problems. For example, the
program design called for a “tracking” component that included hiring personnel to verify
the location and activities of clients when they were not involved in program activities.
Face-to-face and phone contacts were to be initiated at random times throughout the
week with special emphasis on “high crime” times such as nights and weekends. This
aspect of the program failed to be operational in any real sense for a substantial part of
the program. Clients were largely left completely unsupervised by program staff while
not involved in program-related firnctions (Bynum, Davidson 11, Beitzel, Nguyen, &
Wordes, 2000).

The program was also never able to integrate parents, guardians, or other family
members into program activities. Although multiple special events (e. g., ice cream
socials) were created in an attempt to create positive interaction between youth and their
families, few families took advantage of these opportunities. Finally, there were also
major problems with staff turnover during the first two years the YFS program was in
operation. This is important from a policy perspective because it is necessary to build
“startup time” into a problem model. Providing for a six-to-twelve month startup period
can assist with working out the “kinks” of a program. When implementing programs of
this nature sufficient time should be taken to plans its implementation and fix emergent

problems in the program model as they arise.

195

The notion of treatment is one of the cornerstones of the juvenile justice system
and it is philosophically grounded on providing safe treatment services to wayward
youth. Although the non-secured residential treatment was more effective in providing
the intended outcomes in this case, the day treatment model should not be abandoned. It
was notable that the day treatment model produced delinquency risk factor change scores
that were similar in magnitude to traditional probation. While members of the concurrent
control group were more serious, they proportionately experienced less (favorable)
change. The YFS program was an attempt to maintain the connections between clients
and their communities, including their families. This type of community-based
alternative may have been more effective in reducing delinquent behavior if given
enough time to work out the implementation problems but these findings suggested
otherwise. In this case, the cumulative evidence indicated support for the argument that
non-secure residential facilities may offer effective alternatives to more restrictive forms

of institutional treatment.

196

APPENDIX A

197

DELINQUENT YOUTH
CLASSIFICATION AND ASSIGNMENT REPORT
MICHIGAN DEPARTMENT OF SOCIAL SERVICES

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

WAYNE COUNTY
Name of Youth (Last, First, Middle): Admission Date:
Date of Birth: Soc. Sec. #:
Who Completed RlSk Report: Sex: [:1 N131c [:1 Female
(Circle One) Court Case Number: County:
1 = Juvenile Court
Residential Provider Number Worker Name:
2 = MDSS Intake & Court
Services
Supervisor’s Signature: Report Date:
3 = Reception & Assessment
Center ODS Case #: Site #: Stay #:
4 = DSW
5 = Other ODS Staff
MDSS Case #:
6=Oma
INITIAL SECURITY MATRIX
MOST SERIOUS
ADJUDICATED
OFFENSE CATEGORY RISK SCORE SECURITY LEVEL ASSIGNMENT
HIGH HIGH
CLASS 1 & II
MODERATE MEDIUM
LOW MEDIUM
HIGH LOW
CLASS 111
MODERATE COMMUNITY BASED
LOW COMMUNITY BASED
HIGH ' LOW
CLASS IV & V .
MODERATE COMMUNITY BASED
LOW COMMUNITY BASED
INITIAL SECURITY MATRIX
MANDATORY OVERRIDES
L] Manshughter L1 Negligent l-bmidde UFelmy Child Abuse Dcscwor 1)
[1 FA (301) UAW/ice (302) UPRIOR CIASS 1 or 111 [3131500 (306)

 

 

DISCRETDNARY OVERRIDE: ENTER CODE FROM G’ERATIONS HANDBOOK:

 

 

 

 

PLACEMENT SECURITY CATEGORY
DCOMMUNHY BASED BLOW 1:1 MEDIUM [3111011

 

198

 

 

MICHIGAN DEPARTMENT OF SOCIAL SERVICES
INITIAL DELINQUENCY RISK ASSESSMENT SCALE

 

 

 

 

 

 

 

 

1. Age At First Adjudication 7. Number Of Out-Of Home
11 Or Under 3 Placements
12-14 2 One Or Fewer 0
15 1 Two Or More 1
16 Or Over 0 8. Number Of Runaways From Prior

2. Number Of Prior Arrests Placements
None 0 None 0
One Or Two 1 One Or More 1
Three Or More 2 9. Last Grade Completed

3. Current School Status 10th Grade Or Higher 0
Attending Regularly, Occasional 9th Grade 1
Truancy 0 8'h Grade
Dropped Out Of School 1 10. Level Of Parent/Caretaker Control
Expelled/Suspended/Habitual Supervision Provided Even If
Truant 2 Ineffective 0

4. History Of Drug Use Little Or No Supervision Provided 1
No Known Use Or Exper Only 0 11. Peer Relationships
Regular Use 1 Good Support &

5. Current Offense Influence/Associates With Non-
Non-Assaultive Felony Or Delinquent Friends 0
Misdem 2 Non-Peer Oriented Or Some
All Others 0 Companions W/Del Orient. 2

6. Youth Was On Probation At Time Most Companions Involved In

Of Commitment Delinquent Behavior Or Gang
No 0 Involvement 3
Yes 1

TOTAL SCORE
RISK CATEGORY
RISK LEVELS
0-8 — LOW 9-1 1 — MODERATE 12-19 - HIGH
COMPLETED BY: I I I I I I I

199

STAFF CODE/LOAD #

 

 

APPENDIX B

200

Table B-1. Detailed Listing of Qualifying Offenses (n=369).

 

Experimental Concurrent Historical Total
Robbery Count 2 4 2 8
Col % 2% 4% 2% 2%
Assaultive Count 26 15 24 65
C01 % 20% 14% 18% 18%
Criminal Sexual Count
Conduct 2 12 3 17
C01 % 2% 11% 2% 5%
Arson Count 1 0 1 2
Col % 1% 0% 1% 1%
Burglary Count 10 13 17 40
C01 % 8% 12% 13% 11%
Drugs Count 26 20 l 0 56
C01 % 20% 19% 8% 15%
Receiving and Conceal. Count
Stolen Property 14 14 14 42
Col% 11% 13% 11% 11%
Larceny Count 8 3 9 20
C01 % 6% 3% 7% 5%
Weapons Count 1 6 3 14 33
C01 % 12% 3% 11% 9%
Retail Fraud Count 1 l 1 3
Col % 1% 1% 1% 1%
Malicious Dest. Of Count
Property 7 8 6 21
C01 % 5% 8% 5% 6%
Auto Theft Count 5 0 5 10
Col % 4% 0% 4% 3%
Trespass Count 0 2 0 2
Col % 0% 2% 0% 1%
Other Prop Count 0 1 0 1
Col % 0% 1% 0% 0%
Other Count 7 5 6 18
C01 % 5% 5% 5% 5%
Status/Disorder Count 7 5 19 3 1
Col % 5% 5% 15% 8%
Total Count 132 106 131 369
Col % 100% 100% 100% 100%

 

201

Table B-2. Multicollineari dia nostic tests for intake variables

 

 

 

 

 

 

Model A1
Tolerance VIF
Age Instant 0.596 1.679
Offense 0.859 1.164
Age lst Arrest 0.105 9.481
Prior Arrest 0.131 7.623
School History 0.122 8.215
Drug Use 0.246 4.060
Offense Serious. 0.090 11.136
Prob Time of
Offense 0.276 3.621
Prior Placement 0.268 3.728
Runaway Status 0.518 1.930
Last Grade 0.202 4.956
Parental
Commitment 0.349 2.865
Peer Relations 0.1 17 8.578
Total Risk Score 0.013 75.750
Model A;
Tolerance VIF
Age Instant 0.623 1.605
Offense 0.863 1.159
Age lst Arrest 0.565 1.769
Prior Arrest 0.818 1.223
School History 0.780 1.282
Drug Use 0.841 1.189
Offense
Seriousness 0.829 1 .206
Prob Time of
Offense 0.788 1.269
Prior Placement 0.892 1.121
Runaway Status 0.902 1.108
Last Grade 0.755 1.324
Parental
Commitment 0.955 1.048
Peer Relations 0.883 1.133

 

202

 

 

 

 

 

 

Model B1
Tolerance VIF

Age Instant 0.595 1.680
Offense 0.824 1 .214
Age 1st Arrest 0.105 9.529
Prior Arrest 0.130 7.718
School History 0.120 8.323
Drug Use 0.245 4.080
Offense Serious. 0.089 11.220
Prob Time of
Offense 0.273 3.659
Prior Placement 0.267 3.740
Runaway Status 0.518 1.930
Last Grade 0.199 5.020
Parental
Commitment 0.347 2.884
Peer Relations 0.116 8.632
Total Risk Score 0.013 76.995
Historical 0.726 1.377
Concurrent 0.668 1 .497

Model B;

Tolerance VIF

Age Instant 0.621 1.610
Offense 0.825 1 .212
Age lst Arrest 0.559 1.788
Prior Arrest 0.81 1 1.234
School History 0.767 1.304
Drug Use 0.812 1.232
Offense
Seriousness 0.808 1 .238
Prob Time of
Offense 0.783 1.277
Prior Placement 0.884 1.131
Runaway Status 0.891 1.122
Last Grade 0.751 1.331
Parental
Commitment 0.944 1.059
Peer Relations 0.871 1.148
Historical 0.727 1.375
Concurrent 0.679 1 .473

 

Table B-3. Multicollinearity diagnostic tests for intake variables(N=396).

 

 

Tolerance VIF
Age Instant 0.620 1.613
Offense 0.823 1.215
Age lst Arrest 0.559 1.789
Prior Arrest 0.799 1.251
School History 0.765 1.307
Drug Use 0.811 1.233
Offense Seriousness 0.804 1.244
Prob Time of Offense 0.768 1.302
Prior Placement 0.868 1.152
Runaway Status 0.890 1.123
Last Grade 0.747 1.338
Parental Commitment 0.944 1.060
Peer Relations 0.860 1.162
Days in Detention 0.863 1.158
Historical 0.701 1 .426
Concurrent 0.645 1.551

 

203

dia nostic tests for back round variables

 

Table B-4. Multicollineari

 

 

 

 

 

Model A Model B
Tolerance VIF Tolerance VIF

Age Instant 0.919 1.088 Age Instant 0.906 1.104
Intake Offense 0.973 1.027 Intake Offense 0.927 1.078
Total Intake 0.925 1.081 Total Intake 0.802 1 .246
Soc Dis 0.472 2.120 Soc Dis 0.471 2.122
Ses 0.453 2.208 Ses 0.453 2.208
Det Days 0.877 1.140 Det Days 0.971 1.281

Concurrent 0.695 1.438

 

 

204

Table B-5. Multicollinearig diagnostic tests for background variables and T,
Delinquency Risk Factors (n=178).

 

 

 

 

 

Model A
Tolerance VIF
Age Instant 0.918 1.089
Offense 0.973 1.027
Total Intake 0.893 1.120
T. Risk 0.956 1.046
Soc Dis 0.471 2.121
Ses 0.450 2.223
Det Days 0.877 1.141
Model A
Tolerance VIF
Age Instant 0.865 1.156
Offense 0.982 1.018
Total Intake 0.898 1.1 14
T1 SB 0.629 1.591
T3 DU 0.586 1.707
T; SL 0.751 1.332
Soc Dis 0.472 2.119
Ses 0.450 2.222
Det Days 0.925 1.081

 

 

 

 

 

 

Model B
Tolerance VIF
Age Instant 0.905 1.104
Offense 0.926 1.080
Total Intake 0.792 1.263
T; Risk 0.928 1.077
Soc Dis 0.471 2.123
Ses 0.450 2.223
Det Days 0.776 1.288
Concurrent 0.675 1.481
Model B
Tolerance VIF
Age Instant 0.850 1.177
Offense 0.924 1.082
Total Intake 0.772 1.295
T, SB 0.627 1.596
T, DU 0.578 1.730
T3 SL 0.713 1.402
Soc Dis 0.471 2.121
Ses 0.448 2.232
Det Days 0.841 1.189
Concurrent 0.646 1 .547

 

205

Table B-6. Multicollinearity diagnostic tests for background variables and
Delinquency Risk Factor Change Scores (n=178).

 

 

Model A
Tolerance VIF
Age Instant 0.907 1.102
Offense 0.928 1.078
Total Intake 0.878 1.139
AS Change 0.752 1.330
SB Change 0.559 1.788
AD Change 0.696 1.436
SE Change 0.909 1.100
PV Change 0.853 1.172
EP Change 0.675 1.481
FIP Change 0.866 1.155
DU Change 0.543 1.842
SL Change 0.578 1.732
CI Change 0.801 1.248
Soc Dis 0.465 2.151
Ses 0.451 2.216
Det Days 0.934 1.070

 

 

 

Tolerance VIF
Age Instant 0.900 1.111
Offense 0.874 1.144
Total Intake 0.768 1.302
AS Change 0.751 1.331
SB Change 0.557 1.794
AD Change 0.694 1.440
SE Change 0.907 1.102
PV Change 0.850 1.176
EP Change 0.675 1.481
FIP Change 0.865 1.156
DU Change 0.528 1.896
SL Change 0.577 1.732
CI Change 0.798 1.253
Soc Dis 0.465 2.151
Ses 0.450 2.222
Det Days 0.914 1.094
Concurrent 0.764 1.309

 

206

Table B-7. Multicollinearigy diagnostic tests for background variables and

Program Services (Disaggregated and Cumulative Measure) (n=178).

 

 

 

 

 

Model A1
Tolerance VIF
Age Instant 0.901 1.110
Offense 0.979 1 .022
Education 0.148 6.777
Job Skills 0.351 2.853
Recreational 0.123 8.107
Soc Dis 0.489 2.046
Ses 0.478 2.093
Det Days 0.904 1.107
Model A2
Tolerance VIF
Age Instant 0.921 1.086
Offense 0.978 1.023
Total Intake 0.961 1.041
Total Services 0.950 1.053
Soc Dis 0.489 2.043
Ses 0.480 2.083
Det Days 0.959 1.043

 

 

 

 

 

 

Model B;
Tolerance VIF
Age Instant 0.898 1.114
Offense 0.945 1.059
Education 0.146 6.871
Job Skills 0.339 2.954
Recreational 0. 123 8. 107
Soc Dis 0.488 2.049
Ses 0.47 8 2.094
Det Days 0.882 1.134
Concurrent 0.888 1.126
Model B;
Tolerance VIF
Age Instant 0.910 1.099
Offense 0.918 1.089
Total Intake 0.829 1.207
Total Services 0.944 1.05 9
Soc Dis 0.489 2.043
Ses 0.480 2.084
Det Days 0.935 1.069
Concurrent 0.800 1 .249

 

207

Table B-8. Multicollinearity diagnostic tests for complete model (n=178 1.

 

 

 

 

 

 

Model A Model B
Tolerance VIF Tolerance VIF
Age Instant 0.763 1.311 Age Instant 0.754 1.326
Offense 0.896 1.1 16 Offense 0.848 1.179
Total Intake 0.795 1.258 Total Intake 0.672 1.488
AS“ 0.384 2.602 AS“ 0.384 2.603
SBn 0.379 2.639 SBn 0.376 2.658
AD“ 0.379 2.641 AD“ 0.378 2.644
SET] 0.555 1.802 SE11 0.555 1.802
PVT. 0.350 2.860 PV—n 0.339 2.953
EPn 0.408 2.448 EPn 0.407 2.456
FIP“ 0.580 1.725 FIP“ 0.580 1.725
DU-n 0.354 2.824 DU—n 0.352 2.844
SLn 0.475 2.107 SLn 0.458 2.182
CIT] 0.468 2.139 CI“ 0.467 2.140
AS Change 0.433 2.310 AS Change 0.433 2.311
SB Change 0.384 2.606 SB Change 0.373 2.678
AD Change 0.445 2.247 AD Change 0.443 2.256
SE Change 0.544 1.838 SE Change 0.539 1.856
PV Change 0.433 2.309 PV Change 0.432 2.312
EP Change 0.491 2.036 EP Change 0.487 2.054
FIP Change 0.623 1.605 FIP Change 0.623 1.606
DU Change 0.391 2.555 DU Change 0.387 2.581
SL Change 0.431 2.321 SL Change 0.431 2.322
CI Change 0.529 1.892 CI Change 0.525 1.905
Total Total
Services 0.842 1.188 Services 0.840 1.190
Soc Dis 0.450 2.221 Soc Dis 0.450 2.222
Ses 0.427 2.339 Ses 0.425 2.351
Det Days 0.823 1.215 Det Days 0.710 1.409
Concurrent 0.575 1.740

 

 

208

APPENDIX C

209

%

 

 

 

 

 

 

 

 

 

 

.oo. ...o. .8. ooo. o8. .8. €9.95. .5
8.. 5....- 8..- ooo. ooo. 8. 8.5.9.8 828A. 8.5.8....
o8 8.. Sn 8- oow «ow z
ooo. ooo. 8o. 2.... ...u. .8. 69.95. ooo.

:9...- ..ooe. town. .8. .8. on..- 8.5580 828A. 833%
8.. .8 oo~ mom 8... Now 2
ooo. .8. .oo. woo. 25. t... 66.8.... on

too... .58.- 8~.~.- .om..- «8. 8o. 8.5.2.8 828A. 83.8..
8... 8m o8 mom o8 Now 2
o3. ~.~. 3o. ooo. ..o. woo. 8.5.5. .3
ooo. 8o.- .8... ooo.- o8. «8.. 8.5580 828.. €3.55
oom 8- oom mom .8 mom 2
. ooo. ooo. .8. 8m. ooo. 8.5.... on
. :8.” .58.- o8.- «8.- o8. 8.5580 e856“. 83.82
8- 8... 8... 3w 8... .8 z
ooo. . ooo. 8o. 3.... ooo. .885. .5

zoom.- . .58. 82. 8o. «3.. 8.5580 828A. 68:8
mom 8.. 8.. 8- oow «on 2
8o. ooo. . 8w. 8.. 3.. 8.5.5. on

tom-o.- ..o8. . .8. o... o. ... 8.5.2.8 828A. 2.3.5....
now 3~ mom .8 .8 8.. z
o8. 8o. o8. . 8.. ooo. 8.5.5. 3.»
E? .9... .8. . 3o. o8. 8.5.680 828d .nx.o.8w .8585: 95....
8m 8.. oou RN 8... 8- 2
8m. 3m. 8.. 8.. . v8. 625.5. .5
«8.. .8. o. .. 3o. . .8.- 55580 828d .8955. 5 88.5
wow .o~ No... 8m 8o 95 2
8... 8o. 3.. ooo. e8. . 8.5.5. on
o8. .3.- o. ..- o8. .8.- . 85.250 828d . .3555... 5 oo<

9. mx ex 8. «x 5.

23.52.50

210

i

 

 

 

 

 

 

 

 

 

 

08. 8. o8. 8o. 8«. 58. 8.5.5. 5.0
.«o. 58. 0.5. .8..- 58. 58.- 8.5.980 85555 .o«x558..0 5.“.
8« 8« 8« k« 8« 8« z
.8. .oo. ..oo. .5... o8. o8. 8.58. 5.0
.55.. L8.- ..«5... 5 .o.. «8. 8o. 8.5580 80.8.". 8.55880 am
8« 8« 58 t.« 8« 8« z
ooo. 5«... t... 8.. 8... 80. 65.5.5. 5.0
...5«.- 8o. 8.. «8. 53.. .8.- 8.5.80 8285 5.0.5880 >5
8« 8« 8« t-« 8« 8« z
58. 8«. .«5. .8. o8. ...«. 85.5.5. 5.0
...o. 58.- 58. ooo. .5... So. 8.5580 80555 $5885 00
8« 8« 8« k« 58 8« z
ooo. 8o. ooo. 8m. ««.. 58. 65.5.5. 5.0
:53.- ..8.. :53. 8o.- «8. .8. 8.5580 85555 5.55820 3
8« 8« 5o« «5 8« 8« z
8... ooo. .oo. 8.. 3.5. «8. 8.5.5. 5.0
«8. .85.- ..5««.- 8.. ..oo. 58. 8.5580 8055“. 5.55880 50
8« 8« 5o« t« 8« 8« z
.8. .8. ooo. v«o. 5«o. 8.. 85.5.5. 5.0
8.. 8... :58.- .8... .8... .8. 8.5580 8055.. 3.55880 0...
5o« 8« 8« 8« 8« «o« z
ooo. ooo. ooo. .8. 85. 58. 65.5.5. 5.0
..«o«. ...5«.- ...-08 «8.- ooo. .8.. 8.5580 80555 8.5.:0
8« 8« 8« 8« 8« «o« 2
8o. ooo. .8. «oo. 08. 80. 85.5.5. 5.0
8.. ..o«...- ..8«.- .L.«.. 8o.- «8. 8.5580 8555.. 3.5.50
5o« 8« 5o« 8« 5o« «8 z
ooo. 8o. ooo. 8o. .8. .8. 55.5.5. 5.0
:58. :85. 25.5. .50.. 8o.- .«.. 8.5580 80.8... 2.5.50
ox 8 ..x 5x «x .x

ocean—9:00

211

 

 

 

 

 

 

 

 

 

«8 So. 9% 8o. «8. m8. .Bfié gm
.8 .52. «8 :95. :8... E... 8.52.8 828a 85385 .380
SN 8N SN 8w 9m 8m 2
3w SN. m3 8v. 8n. Ev. A353 .3
23.. Bo. «no 53.- «no. «8.- 552.8 828a 55260 8.528
8a new mom Rm 8m 8m 2
8m. NE. 8.. 80. RN. m8. ABEE. gm
“8. 80.- 8a.- 2 P. «3.- 3.. 532.8 828a 893%
8m new mom km «8 RN 2
3m. .8. 3. m5. .2. 3.. 825-3 gm
.8. «8.- m8: .5... So; .8.- 8.5.2.8 Samoa 333 8m
8m 8m 8m RN «8 8m 2
8w. n8. 5. 2w. 9.6. 8o. .8._s-~v .9m
29. 8.. «8. 8o. 80. :8? 552.8 828a #28028 2.2an
8m 8m 8m kw 2m 8w 2
So. new. «8. 3n. .8. 8*. .Efié gm

:3... to. ..Na. “no. :0. no. .. 832.8 Samoa 83858 .0
8m wow wow RN «8 8w 2
3. 2o. «8. «.3. at 2w. A353 .9m
98. at. 8o. .3. 95.. 8a.. 5.388 838 393820 .6
8m 8m 8w RN 8N 8m 2
So So. So 3m. v8. n8. A353 gm

Li. .L 3”,- :8~.- 20.- «8.- 8o. 5.92.8 528a A .8855 :o

ox mx «x «x «x .x

30:82.00

212

 

 

 

 

 

 

 

 

 

 

9... 8o. .8. .8. .8. 62.9.3 .8
.8.. :8? . :8? :89. .8.. 88.9.8 8&3". 8.5.8:
8- 88 8a 88 8a 8- 2
So. So. .8. . 8o. «8. 62.9-... .9m
:81- :8... :8? . eh? .m... 88.88 8&8“. 655$
8- 88 wow 8- 8~ 8- z
«8. 8o. .8. 8°. . 8.. 62.9-3 .8
.8.. :3? .8.. :k? . 8.. 8.8.2.8 8&8; 85.5..
8- 8~ 28 wow 8- mow z
80. m8. .8. NS. 8.. . 62.98. .8
t .. m. ..- .8.. .N..- 8.. . 88.2.8 828.". Eerfiw
8m 8m 8m 8w 8- 88 2
8o. 8o. .8. 8o. 8o. 93. 8.5.8 .3
.9. .88.. 8.. :~.v.- :8... 8... 88.88 898a 65:9.
8m 8m 8m 8m 8- mom 2
So. So. .8. 8o. ”8. ~.~. 69.9.8 .8
.8...- :20 he..- :8... :8? .8.. 88.2.8 agnon. 65:8
8m 88 8- 88 8- 88 z
.8. So. .8. 8o. .8. «mo. 8.5.... gm
:8? :9». 8... .88. :8? .8... 88.2.8 888 65.5,.
8- mom 8- mom 8- 8~ z
«8. m8. 8... ..n. «8. Q8. 8.5-m. .8.-u,
:2? .8.. 8o. .8. .8..- 8o.- 8..m.e.8 828a 6598a .588... 9.9:.
8- mom 8- 88 8- 8~ z
88. .8. 2... Cu. 88. ..n. 8.8-3 gm
8.... 8? 8o. 88. «8. OS. 88.2.8 823.. 689.9... a 9.8.5
Now Now 8- ~o~ 8- ~o~ z
8.... .8. .8. «8. z... «8. 62.9.8 .8
«8. .~.. 8o. 8... .8. «8.- 88.2.8 888 25.58. a 8<

 

213

 

 

 

 

 

 

 

 

 

 

man. 88. So. 8.. 6.... 8... 69.9.3 .66
8o.- mmo. :2..- n. .. 80. 68.- 8.6.88 8989 696968 A...
8m 8- 88 88 8- 8~ z
.8. «8. 83 ooo. m8. So. 69.9.3 .3
:53 .3...- 8... :53- .8.. .6... 8.6.9.8 8989 6.3888 mm
26 8- 8~ 8- 88 88 2
«8. 8o. 53 .8. 8o. «5. 69.9.3 65
.6..- o~.. «8.- :83 :08.- amo. 8.6.9.8 8968 6.5888 >9
8- 88 8- 88 8- 8~ z
08. 8.. 8.. «R. 68. 8o. 69.9.3 .9».-
~.o.- 85. 2o. 8.... «N..- :96.- 8.6.980 898“. 6.5888 mm
8- 88 8m 8m 8- 8~ z
.8. «8. N8. .8. 8o. ..8. 69.9-3 .9m
So.- :...3 .3..- :-3 :83- 88.- 8.6.9.8 8968 6 .5888 9.
m8 8- 8~ SN 8- 8~ 2
8o. 8o. .3. 88. 8.. 5.. 69.9-3 .9m
:23 :83- So. :23- 8... 98. 8.6.9.8 8989 6.3888 mm
8- 8~ 8m 8m 8w 8- z
wwo. 68. $3 8.. 83 9:. 69.9-3 gm
.8.. .8... 2o. 8..- 2o. 5.. 8.6.9.8 898a. 6.5888 m<
8m 8m SN 88 8- 8~ z
8... 8o. 80. «8. m8. 8.. 69.9.3 .8
.8. :23- :83 .8..- .8.. 8.. 8.6.9.8 898.. 6 53.90
wow 96 8a 8- 88 8- z
. 8... m8. 8... «3.. 08. 69.9.3 9....-
. :98.- .8 .. :39..- .8.. t .. 8.6.9.8 898... 6.8.5.".
96 wow 8- 88 8- 8~ 2
8o. . 8o. 8o. 38. m8. 69.9-3 gm
:3... . :8? :83 :33- ... ..- 8.6.9.8 898... c 53.50

 

N;

214

 

 

 

 

 

 

 

 

8o. 08. .8. 8o. 3.. «.3 69.9.3 .8.»

:. .3- :.8. 23. 8.. 83. ..o.- 8.6.9.8 898... 68890 .968
.8 .8 .8 .8 .8 .8 z
83 83 .8. 83 88. 83 69.9.3 .86
3o.- ..3. 8o. 85. .8... .8. 8.6.9.8 898.. 63398 8.8.8
m8 8 8 8 88 88 z
83 .8. 8.. 83 6.3 8.. 69.9.3 .86
.8.- :8..- .8. .8.- 8a.. 8.. 8.6.9.8 898.. 638mm
8 8 m8 m8 8 8 z
8.. .8. .8. .8. 83 .8. 69.9-3 .86
.o... .. ... :2... 68. 68.- o. .. 8.99.8 898a 6338 8m
88 8 m8 8 8 88 z
9.3 .8. 83 83 83 6... 69.9.3 .86
3.3- 80. 95.- .3. m. ..- 8o.- ..o.6_9.8 898.. 6388.8... 698...
m8 m8 m8 m8 8 8 z
83 8o. .8. 83 83 53 69.9.3 .8.»
.8. .8.. .8..- 8o. 88.. .3. 8.6.9.8 898.. 696268 8
88 a8 8 m8 8 8 z
ooo. 8o. 83 80. 83 .... 69.9.3 .86

:83- .8.. 8o. ...... m8.- 8? 8.6.9.8 898.. ...-338.68 .8
m8 m8 8 m8 8 88 z
.8. 8o. 6.3 So. «.8. .8. 69.9.3 .86

8.3 :...3- .8. :33- 8.. .8.. 8.668 898a A .8868 no

~.x ..x 2x 6x

 

215

 

 

 

 

 

 

 

 

 

 

8m .8 8.. «8. .3. 8w. o8. ABE-N. .3
8o. 8.... 2o. .3. .- 8o. 2o. .88. 8.22.8 8&8; 8.5.5:
mom 8... 8- 8... 8- SN mm 2
8o. .8 ...... .8. N8. 8.. «mo. ABE-u. gm
:5...- :a-. 93.. :-~. :m.~.- 8.. .8..- 8_.~_2.8 828.. axfam
8- 8... 8- 8... 8- 8~ 8w 2
9... So. 2o. «8. 3.. 3.... m8. ABE-N. .gm
.8.. :8...- -..- :mo~.. o8. Sb. .8.. 88.2.8 8&2... 8x. .5
8w 8... 8- 8... 8- 8... 8- z
«6. ...... 8o. «8. a... 9.. 8.. 8.5.8 .3
...... awe. :8...- 8? 8o. 8.. 8.. 888.80 828a txrmw
88 8- 8~ 8- 8~ 8- 8... z
.8. So. 8... 8o. 8... .8. 8o. 69.9-3 ....m
.9... :.o~.- 3o. :2...- Nmo. 8.. 14.8. 8.92.8 828a 55.5...
.8 8m 8m 8- 8~ 8m 8w 2
.8. 8... m2. m8. 8o. .8. 8o. ABE-N. .gm
:BN: mmo. 8o.- :8.. :2? 8... :...~.. 88880 8&3. 65.5w
8~ 8m 8a 88 m8 8... 8- z
«8. t... .8. o8. .8. 08. So. 898...... .gm
:3..- 5.. 8o. :83. :8? 88.- :8? 88.2.8 898.. 3.5:?
t... R... km E kw kw 8- z
5.. 2.. ..8. 8m. 8.. c8. «8. 8.3-8 .3
.- .o. «8. ooo. 80.. 8o.- .8..- «3.- 88.2.8 888a 8x§8w .8683 9.95
8m 8m 8m 8w 8w 8- 8~ z
08. 8... 80. -.. 3.... 8o. 8... ABE-N. .9...
«8. ms.- 3... N8. 3... .8... 8c. 8.8880 828.. $.39... .a 88.5
8a 8- 8... 8- 8... 8m Now 2
SN. 2m. .3“. n8. «8. 3.. 8m. 88...... .3
m8. .8.- ...o. .8. 8o. .8. .3- 8....880 828a 25.8.2. .n 8<

 

9X 9X

5X

ex

mwx 3.x

2039—50

0

9.x

216

%

 

 

 

 

 

 

 

 

 

..w. .8. «8. «.8. con. 8.. 8.. 62.9.8 .2...
ms.- 3.. .8.- .8.. 8o. 28.. v. ... 8:238 828“. 688820 a:
8m «8 n8 8- 88 88 8- z
. 38. 8.. NS. 8°. .8. .8.. 8.9.8 .88
. .8..- 8.... .8..- :o.... :88 v. .. 88.280 828.. 6.8820 mm
98 n8 n8 8- 8~ 8..- 8n 2
38. . 8.. 8°. 8.. 8.. 98. 62.9.8 8.....
.8..- . 88. :8. .8.. 38.. «3.. 88.980 898.. 6.88820 E
m8 8w «8 8a 88 ...8 8- z
8.. 8.. . 3... ma. 8... 88.. 62.9-8 ....m
88.. 8... . 88. ...-o.- .8..- 8... 8:28.80 888.. 5.8820 mm
n8 «8 m8 «8 ...8- 8a 8- z
«.8. 8. ..B. . .8. .8. 88. 62.9.8 .85
8....- :88 88. . :8..- :8. .m... 8.8.280 898.". 6.88820 9..
m8 8- 8~ n8 n8 8... 8- z
88. 8.. 8.8 .8. . 88. .8. 62.9.8 .5
.8.... .8.- ...-o.- :8..- . :...8 8.. 8.8.280 828.. 6.8820 mm
8... 8- 8~ 8- nm... 8m 88 z
.8. 8.. 88. 8o. 88. . 3... 62.9.8 .2».
:88 .8.- .8..- :8. :88. . L... 8.8.980 8982". 6.88820 9..
m8 8- 88 8- 88 8- 88 z
.8.. 8m. 88.. 8o. .8. 8.... . 62.9.8 ....m
c. .. 88.- 88. .m... 8.. ...... . 8.9280 898m 6.x..t0
8- 88 8m 8m 8- 8... 88 z
.8. ...-o. 98. .8. 8... «8. 8... 62.9.8 .2m
:88 «a..- «.8.. o.o.- :88 .8.. .8. 8.8.280 898.. ...-58.5...
88 8- 8... 8m 8- 8... 8- z
88. 88. 8... N8. .8. .8. 88. 62.9.8 .20.
..v..- 8.. 8.... :28 :80. e......- :2n.. 88880 828.. .. 58.25

 

-.m|.x-..- - -- ...... ..x? 2.x . 8...... - 3...... : 8.x

 

217

”mm

«ms.
90...

now

am...
35.-

mmm

:ng

now

mam

mam

:mmm.

62.9.8 .86
882280 828a
z

62.9-8 ....m
8.2280 828..
2

62.9.8 .3
8.8.280 829:.
2

62.9.8 ....w
8.2280 828..
z

62.9-8 gm
8.8.280 828..
2

62.9.8 .2...
8.8.280 828..
2

62.9.8 .98.
8.8.280 828a.
2

62.9.8 .85
882280 828..

 

6888.0 8980

622.8 8.892...

Gmxvmmm

$9“va 00w

AvmeooEow E9528

89909.ch .0

...-8889.0 ..m

..waocaco no

 

218

%

 

 

 

 

 

 

 

 

 

.oo. 80. 3o. 08. 0.0. 8o. 69.3.0. 0.0
:09. 03.- .2... 08. .8. :2? 8.8880 828.. 850..."...
80 80 80 80 80 80 z
.00. 08. o00. 00o. ooo. 3.. .3030. 0.0
00o. ...o. 08. ..0.. :30: 0. .. 8.8880 80.8.. 85:80
80 000 000 80 80 000 z
000. 00o. 000. 000. ...o. ...0. 68.8.0. 0.0
03.. 0.... 00o.. 08.. 00.. 00o. 8.8880 80800 35.5..
80 0o0 80 80 80 80 2
So. 0... .00. .... .0o. 000. 80.8.0. 0.0
0. .. 00o.. ...o.. 08.. .00.. 0.8.- 8.8880 8088.. €0.50
80 80 80 0o0 80 80 z
.00. 000. 08. 0:. ooo. 000. 88.8.0. 0.0
.8. 0.o.. :3... 00o. :0..0. 00o. 58880 8088 6.0.5..
80 80 0o0 80 80 80 z
.00. 0.o. .00. 0.o. ooo. 00... 68.8.0. 0.0
08.. 00.. to. .0... :20. 0.8. 88880 8088.. 50.50
80 80 80 80 80 80 z
0...... 00.. 08. 000. .8. 00... 60.8.0. 0.0
08.- 00o. :0.0. moo. :000.- 03. 8.8880 80.8.. .395...
0.0 0.0 R0 :0 .00 t0 2
0.o. 000. 30. 03. 0.0. 0.o. 08.8.0. 0.0
.03. 08. .0o. .00.. 0.o.- .00... 8.8880 80.8.. 5.880 88.888 988.
000 000 000 000 000 000 z
.0.. 000. .00. .0... .00. 000. 8.8.0. 0.0
So. 00o. ..o. 0..o.. 08.- 08. 8.8880 80880 .0539... 8 88.5
000 000 000 000 000 000 z
00... 8o. 8.. 0.0. 000. 0.0. 82.30. 0.0
.0o.- :8..- 0o..- o0o.- 00o. 00o.- 8.8880 8088.. :5888... 8 80..

 

mNX

VNX

max

219

 

 

 

 

 

 

 

 

 

 

..0. 03. ooo. o... .3. 8.8.0. 0.0

...o. .0. ..- ....0. 08.. 00o. . 8.8880 8080.. 80500800 0.".
000 000 000 000 000 000 z

.00. 0... 0o... ooo. ooo. ..0. 8.8.0. 0.0

08.- .0o.- o0o.. :00... :000. 0.o.. 8.8880 880.. 85.80800 am.
000 000 000 000 000 000 z

000. 00.. .3. 0o... 00o. .8. 80.8.0. 0.0

03 000 ..0.. o0o. .00... 3.. 8.8880 8080.. 85.80.85 >0
000 000 000 000 000 000 z

03. ..0. 00.. 0.o. .8. 000. 80.8.0. 0.0

08.- 00o. 000. .00.. :00..- 08.- 8.8880 80800 50.80800 00
000 000 000 000 000 000 z

0.... 000. 80. 20. 8o. 0.o. 8.8.0. 0.0

03.- 0S ..0.-0. 30.. ..0.0.- .00.. 8.8880 8080 6.0080800 9..
000 000 000 000 000 000 z

00.. 8o. ..8. ooo. 8o. 80. 88.8.0. 0.0

.3.- :30. :.0..- ..0.0.- ..000. 00o. 8.8880 80.800 6.500800 00
000 000 000 000 000 000 z

80. 3o. .8. 0.o. ooo. 00.. 80.8.0. 0.0

08.. ..0... :00..- .0....- :30. 0.o. 8.8880 8.0800 8.500800 0...
80 80 000 80 80 80 z

0.o 00.. ooo. .3. .oo. 8.. 8.8.0. 0.0
..... 0.o. 85.. 0o..- :000. ... ..- 8.8880 8080... 8.5...0
80 80 80 80 80 80 z

00.. 0.0. 000. ooo. .8. 000. 8.8.0. 0.0

.o. 03.- .8. ..00.... ..0.0. 08.. 8.8880 8080.. 0.00.50
80 000 80 80 000 000 2

.0o. B0. 000. 00o. ooo. 0.0. 8.8.0. 0.0

... ..- 00o. .00.. .00.. :03- 00o. 8.8880 8080.. . . 30.50

 

220

 

 

 

 

 

 

 

 

o... o3. ..00. .0o. 08. 000. 60.8.0. 0.0
0o.. 0 .o.- ..0.. .00.. :3... 000.- 8.8880 8080... 80508.0 .980
000 0.0 0.0 000 0.0 000 z
..00. 00.. .00. 00.. 0.... 000. 60.8.0. 0.0
o0o. .00. 00o. 0.o.- 00o.- 08. 8.8.080 8080.. £060.80 8.80.00
000 000 000 000 000 000 z
ooo. 08. .00. 000. 00o. .00. 80.8.0. 0.0

....00. .0o. 00o. 00o. .00.. 00o.- 8.8.080 8080.. 80.0000
000 000 000 000 000 000 z
. 3... o... 0.0. 000. ..0. 80.8.0. 0.0
. 03. 08.. Eo. 00o. ...-o.- 8.8880 8080.. $050.0 80
000 000 000 000 000 000 z
..0... . 000. 08. ooo. 03. 80.8.0. 0.0
03. . 0.o. :00.. :20.- .0. ... 8.8880 8080.. $0308.80 889..
000 000 000 000 000 000 z
o... 0o0. . 0.... 08. 8o. 80.8.0. 0.0
08.. 0.o. . 03. ..000.- :20. 8.8880 8..-80.. 89.89.20 .0
000 000 000 000 000 000 z
0.0. 08. 0.... . ooo. o... 80.8.0. 0.0
0.o. :00.. 03. . :3...- 00o.- 8.8880 8080.. 00500800 ..0
000 000 000 000 000 000 z
000. ooo. ooo. ooo. . .3. 80.8.0. 0.0
3o. :30. ...-000 :3...- . 00o. 8.8880 8080.. $0500.88 :0

 

mNX

221

 

 

 

 

 

 

 

 

 

 

 

.00. 30. 80.8.0. 0.0
o.o.- ooo. .8. 8.8880 8080.. 850.8...
80 .00 80 z
o8. 000. 000. 80.8.0. 0.0
00.. 08. .3.- 8.8880 80800 80.80
.00 80 80 z
..0.. 0.o. 0.0. 80.8.0. 0.0
08.. .00... 08.- 8.8880 8200.. 8.0.50
.8 B0 80 z
0.0. 000. 00.. 80.8.0. 0.0
to.- .8. .o.. 8.8880 8080.. $0 .50
B0 80 000 z
.00. 30. .00. 80.8.0. 0.0
.8.. 08.. ..0o. 8.8880 8080.. 8.0.5...
80 80 80 2
So. 0.0. 0.0. 80.8.0. 0.0

:0... 00o. 03.- 8.8880 80800 8.0.50
80 B0 80 z
000. 03.. 00.. 80.8.0. 0.0
3o. ..0o.- 08.. 8.8880 80800 50.5..
000 000 .00 2
8o. 00... 00o. 80.8.0. 0.0

:03. .3.- 0. .. 8.8880 8080.". 800.80 .5080... 0.8....
0.0 0.0 000 z
08. 000. .00. 80.8.0. 0.0

:00..- 00o. 0.o.- 8.8880 80800 .0008... 8 00.8.5
000 000 000 z
000. .00. 00o. 80.8.0. 0.0
0.o.- 00o.- 8.. 8.8880 8080.... 20.8.0... 8 00..

00x 00x 00x

ago—«0.2.50

222

1

 

 

 

 

 

 

 

 

 

 

000. .00. .00. 80.8.0. 0.0
08.. 8o. 08.. 8.8.080 8080.. 80.000800 0.0
80 80 000 z
0.0. o8. 03. 80.8.0. 0.0
08.- .8. 0.o.- 8.8.080 80.000 8.000800 ..0.
80 80 000 z
80. 80. 000. 80.8.0. 0.0
08. 08. 08 8.8.080 8080.. 8.000020 >0
«.0 0.0 000 z
8.. 30. 30. 80.8.0. 0.0
8.. 08.. 0.o. 8.8880 8080.. £50885 m0
000 80 000 z
.0... 000. 80 80.8.0. 0.0
.3.- 000. 08.- 8.8.080 8080.. 8.000800 9..
80 0.0 000 z
00.. 3.. 00.. 80.8.0. 0.0
08.- 08.. 0.o.- 8.8.080 8080.. 8.000800 00
80 000 000 z
000. 000. 000. 80.8.0. 0.0
08.- 3o. 03.- 8.8.080 80.00... 8.000020 0<
S0 B0 80 z
000. 000. 0.o. 80.8.0. 0.0
08.. 0.o.- .0... 8.8.080 880.. 8505.0
80 80 80 z
ooo. 000. 000. 80.8.0. 0.0

... .0.- ..3 - 03.- 8.8880 80800 050.50
.o0 80 80 z
ooo. 80. .8. 80.8.0. 0.0

:.00. 03. :00..- 88880 80800 .. 50.50

00x 00x 00..

28.8880

223

 

80.8.0. .20. 00.0 05 8 88080.0 0. 080880 ..

80.8.0. 8.6. .00 0... 8 880.88 0. 88.080 ...

 

 

 

 

 

 

 

.8. 000. 80.8.0. 0.0
. 00.. 8.. 8.8.080 8080.. 8000090 8.80
80 80 0.0 2
.0o. . .3. 80.8.0. 0.0
o0.. . . ...... 8.8.080 80800 ..0.0080 8.80.00
80 80 000 z
08. .3. . 80.8.0. 0.0
8.. ..0.. . 8.8.080 8080.. 800000
0.0 80 000 z
o... ..00. ooo. 80.8.0. 0.0
8.. o0o. ....00. 8.8880 8080.. 80.005 80
80 80 000 z
9.0. 00.. 80. 80.8.0. 0.0
0.o.- .8. .8. 8.8.080 8080.. 80008.80 0889..
80 0.0 000 2
3o. .3. .00. 80.8.0. 0.0
..0.. 08. 08.- 8.8.080 8200.. 80.000800 .0
80 0.0 000 z
.8. 8.. 000. 80.8.0. 0.0
.00.. ..o. 08.. 8.8.080 80800 80000800 .0
80 0.0 000 z
08. 0.... 08. 80.8.0. 0.0
:3... 08.. .00.. 8.8.080 8080.. 29.00800 :0

 

 

224

APPENDIX D

225

000800000... 0.02 .0000 0
80D 00.0.5 .0000 0
030.280. 00000=bm 0

 

 

00.000 0.03 «do .00t0fl080 080.0 00300.... 00.00
00.0000 00_0> 02%.: 000300.: 00.00 00 000.002 8.0% 003003—
02000 0.0-E 0&0 .00.-58.00 0.0080020 <00 00.00 0.00. 00.0000
030.. ..0.—mi 0.0080020 0.00.. 00000 ..0 0000.02 E08000... 00.00
00.000 0.05. <00 80:03.00 0000000
.0080. ..0 0.0.. .0 02.00000 00 00.... 0:00.50 00.0000 0 00w000w..”0.00
0:_0> 000w;— ..00.5.E.000 ..0 0:0. .0 00.8080 0 z... 0 .0
00.000 ..0.”. <00 80.89.00 00000 000.000 0 ..0 003 0000000 .0080.
00.0000 00_0> 0003.0 800000.000 0000000 .0050 000000200m 0000000
00.000 0.000 <00 80083.00 00: ma... ..0 3000.0 00080—0000 w
00.00.00. 00_0> 00.3: .00: 00.6 ..0 0.8.0.: 003 0.5
00.000 ..0..— (£0 .00.-89.00 .00.—00 0.
00.0.0000 00.0000 0:.0> ..0.—w... 0.00.0 .00.-.00 000.000 8.0% .0000m
00.000 0.02 <00 .00.-53.00 0.00.00 00.00
000.0 00.0000 030.. 0000.3 .000-:0 00.00 00 00000.2 0.00.5. 00.00
00.000 0.02 «£0 .00.-53.00 5:0»
.0w0000 00.0000 00.0> 00.3.: 02.00.00.000 .0. 0w< 00.000.03.00. 00. 09..
3.0000 90 .00.-88.00 0000000 .0080. ..0 003 0000000 0000000 .0800.
00.000 0.05. n 8.000000 0.00m EU 000000000 500» 0020 00.0000 0000000
0=_0> 00.3.: 0000000 .0080. 00.. 00.0.00 .0 0w< 000.0... .0 09..
0030......» 0000000000—
.m 2- 0000000200 .00.-5000.200 m... Btwwwmmmo
005 .0000 0000000 .0085 ..0.. ...-.0000. 0000 0500.0
0. €2.30... ..0 00030—0... 80.000080053000010 0.03.2000.
00.00.00.» 000000009
.25.:-
00. 00:022-030.. 036% 00wfi<> 2000000000 000<E<> 02.2 000.00;
40.30.557.00

 

H4:-H 67:90.0 ”mu—Eds

226

 

 

227

808808... 0.0... .0000 0
0.0m. 8.08. .0000 0
030.>00.0. 00.0.0000m 0
0.0.0. 000003. N 00.6003. .m m... .. ... 00000..000 0000m 0%.0 0000.. 80008.00 0. 0.000005 09.000 00
0.0.0. 00000.00. n 0000000. .m m... .. ... 0000....000 0.00m 0w0000 000.0 > .0808... 0w0000 >0
0.0... 00000000. n 0000000. .m m... .. .. 00008800 0000m 0w0000 800.00....0m 0w0000 0w0000 mm
0.0.0. 000000.. n 0000003. .m m... .. H 000000000 0w0000 000038.00. 00030... 000....< 0.000080. 0m0000 Q<
0.0.0. 000000.. n 0000000.. .m m... .. ... 00008800 0000m 0w0000 0030000. .0000m 0>..0w0Z 0w0000 mm
0.0... 000000.. n 0000000.. .m m... .. ... 0.500.800 0000m 0w0000 .0000m 00030... 000....< 0>..0w0Z 0w0000 m<
00.00... 0.0.x .m . ... 00008800 000.00.. 0.0.x 00000.00..0Q . .. ..0 0.00... 0.0.0. . ..
00.000 0.0.00.0... .m . ... 000000000 .0080>.0>0. 0.808800 . ....0
00.00... 0.0.0. .m .... 000000000 0.003.. 0.0... 00.00005 ......m
00.000. 0.0.... .m .... 00008800 00.. 0.0.9.0000? .05..
00.00... 0>..00.000 .m . ... 000000000 .0080>.0>0. 0.000.... 0>...000 . ..0....
00.00.. 0.0.... .m . .0 000080000 00000 80008.00 0. 0.800%. . 0.00
00.00.. 03.008... .m . ... 00008800 000.0 > .0.000000 . ...>0
00.00.. 0.0.00.0... .m . .. 00000..000 800.00...0m . ...mm
00.00.. 0.0.... .m . .0 00000.800 3000.00.30 00030... 000....< 0.00.0300. . 0.0...
00.00... 0.0.... .m . H 00008800 003000.. .0000m 0>..0w07. . ...mm
00.00.. 0.0... .m .... 00000.800 .0000m 00030.0 000....< 0>..0w07. ....m<
.000000 .00.00w0.00 0.00.00.00.01. 000.0 .000000 80000000 .8800 800000000
.8800 .00..0w0.00 0.0000082). 00000 .000000 .00.00.0.... .000000 .00.00.0....
00.00 0 00.00w0.0 0.. 00.000w 00.000
... 0. .0. <00 . . 0 00.0> 000w... .00000 0.0.. .030.....0.00.0 00.0.0. .000.”V 00.0.0. .08...
00.00 0. 00.00w0.0 00000 0.
... 0. .y. <00 . . 0 0000 0.8820000 00.0000 00.0> 000$... .000..0.00.00.00 00000.00. 0000
00.00 0 00.00w0.0 80808800 0.00.00 .030 00.00. 0.
n. 0. D. <00 . . 0 00.0> 00.3.0. ..0.0.000 0%.8000000000”. ..0 .0”... 80808800 .0000000
00.00.. 0.0... 4.9.0 .00.00m0.00 00.0.0800 000% 00.0.0800
0030. 00.0000 00.0> .0000. 00.0.0800 000.» .00.. 00000 .00..
.230:
00 0.0020...<..00 wwwwm0 00wflfl<> 80.0.0000”... 000<E<> 02$. 000<E<>
J<DEHUZOU

 

”ma—m3? 07..ch HAM—<3.»

acoemmomfix 015M 580 u

 

 

8mm 8.85 580 p
955385 38885 a
x51
c8581 u moatow $5 5 _ ._. $895280 3508M $533 88,—. .8 885 353$ 38%
Sun 580 3.85580 a:-Bo=om 5.82 2 ”~835qu 5 .930 B8B 995 «on
*3. Sun 89:.
88885 n Egon nwi 8880 93— 32.5880 8mm Bod 885895302 .8 88322 58>?— 882$
55¢ @8885 Sun— 88%
u zoznficwwhoma 55$ £550 32 32.5580 noofionnwfiz .8 =585§w8ma 38m 59 com
m 8880 n
-0 Ema BMW—z“ n can” Trw Hm MP ._ H 3835280 588mm 55¢ 5 09:50 8.“ 585 09850 55d
:5— 38085 n 3885 a a ._ H 58:58.80 080m 05850 8050235 558880 paw—8:0 _0
*5. 3058: N 558503— ~m a J h 38:50.80 080m uwcan0 $=o>m £5 .emmobm owSBU Am
07.2 3283— M c5883. 5 a ._ ... 32.5280 080m 03.20 3.: wand—8582 098:0 DD
me 83.385 n 3885 E Mb .2. 38:58.80 28m amp—~50 8058235 358m 358m 0320 E..—
me “8263— " :5883— 5 mh ._ H $835280 088m 058:0 floom 8285—09 3 088qu ow550 mm
me 3.3085 n 8385 a El ._ H 32.5880 Boom $550 mos—“3 3885 ow550 >m
.238 8:28 ”,8:
Oh 5.520;!qu <h<n qu<—¢<> ZO—gmnvmmn mq§> m2<z mam<a<>
44.3.750200

 

H1:—<5. UZ—GOU max—33>

228

BIBLIOGRAPHY

ADAM. (2000). 1999 Annual Report on Drug Use Among Adult and Juvenile Arrestees.
Washington, DC: National Institute of Justice,.

Adams, K. (1997). Developmental aspects of adult crime. In T. P. Thomberry (Ed.),
Developmental theories of crime and delinquency (pp. 309-342). New Brunswick:
Transaction Publishers.

Agnew, R. (1992). Foundation for a general strain theory of crime and delinquency.
Criminology, 30(1), 47-87.

Agresti, A., & Finlay, B. (1997). Statistical methods for the social sciences. Upper Saddle
River, NJ: Prentice-Hall, Inc.

Anderson, E. (1999). Code of the street. New York: W.W. Norton & Company.
Aries, P. (1962). Centuries of childhood. Librairie Plon: Paris.

Attar, B. K., Guerra, N. G., & Tolan, P. H. (1994). Neighborhood disadvantage, stressful
life events, and adjustment in urban elementary-school children. Journal of
Clinical Child Psychology, 23, 391-400.

Baker, R. L., & Mednick, B. R. (1984). Influences on human development : a
longitudinal perspective. Boston: Kluwer-Nijhoff Publishers.

Bartholomew, L. K., Parcel, G. S., & Kok, G. (1998). Intervention mapping: a process for
developing theory and evidence based health education programs. Health
Education and Behavior, 25(5), 545-563.

Benda, B. B., & Toombs, N. J. (2002). Two preeminent theoretical models: a
proportional hazard rate analysis of recidivism. Journal of Criminal Justice,
30(1), 1-12.

Bennett, W. J ., DiIulio, J. J ., & Walters, J. P. (1996). Body count: moral poverty and how
to win America '3 war against crime and drugs. New York: Simon & Schuster.

Bernard, T. J. (1992). The cycle of juvenile justice. New York: Oxford University Press.

Berry, B. J. L., & Kasarda, J. D. (1977). Contemporary urban ecology. New York:
Macmillan.

Birkbeck, C., & LaFree, G. (1993). The situational analysis of crime and deviance.
Annual Review of Sociology, 19, 113-137.

229

Bittner, E. (1976). Policing juveniles: the social context of common practice. In M. K.
Rosenheim (Ed.), Pursuing justice for the child (pp. 69-93). Chicago: The
University of Chicago Press.

Blumstein, A., Cohen, J ., & F arrington, D. P. (1988a). Criminal career research: its value
from criminology. Criminology, 26(1), 1-35.

Blumstein, A., Cohen, J ., & F arrington, D. P. (1988b). Longitudinal and criminal career
research: further clarifications. Criminology, 26(1), 57-74.

Blumstein, A., Cohen, J ., Roth, J. A., & Visher, C. A. (1986). Criminal careers and
"career criminals ". Washington, DC: National Academy Press.

Blumstein, A., Farrington, D. P., & Moitra, S. (1985). Delinquency careers: innocents,
desisters, and persisters. In M. Tonry & N. Morris (Eds.), Crime and Justice: An
Annual Review of Research (Vol. 6, pp. 187-219). Chicago, IL: University of
Chicago Press.

Bourgois, P. I. (1995). In search of respect: selling crack in El Barrio. Cambridge:
Cambridge University Press.

Brewer, D. D., Hawkins, J. D., Catalano, R. F., & Neckerman, H. J. (1995). Preventing
serious, violent, and chronic juvenile offending. In J. C. Howell & B. Krisberg &
J. D. Hawkins & J. J. Wilson (Eds.), Serious, violent, and chronic juvenile
oflenders (pp. 61-141). Thousand Oaks, CA: Sage Publications.

Brooks, J. S., Cohen, R, Whiteman, M., & Gordon, A. S. (1992). Psychosocial risk
factors in the transition from moderate to heavy use or abuse of drugs. In M. D.
Glanta & R. W. Pickens (Eds), Vulnerability to drug abuse (pp. 359-388).
Washington, DC: American Psychological Association.

Burgess, R. L., & Akers, R. L. (1966). A differential association-reinforcement theory of
criminal behavior. Social Problems, 14(2), 128-147.

Bursik Jr., R. (1988). Social disorganization and theories of crime and delinquency:
problems and prospects. Criminology, 26(4), 519-551.

Bynum, T. 8., Davidson 11, W. S., Beitzel, T. G., Nguyen, H. X., & Wordes, M. (2000).
Alternatives to residential placement for PA. 150 youth. East Lansing, MI:
Michigan State University.

Carmines, E. G., & Zeller, R. A. (1979). Reliability and validity analysis (Vol. 17).
Thousand Oaks, CA: Sage Publications.

Catalano, R. F., Arthur, M. W., Hawkins, J. D., Berglund, L., & Olson, J. J. (1998).
Comprehensive community— and school-based interventions to prevent antisocial
behavior. In R. Loeber & D. P. F an'ington (Eds), Serious and Violent Juvenile
Oflenders (pp. 248-283). Thousand Oaks: Sage Publications.

230

Catalano, R. F ., & Hawkins, J. D. (1996). The social development model: a theory of
antisocial behavior. In J. D. Hawkins (Ed.), Delinquency and Crime (pp. 149-
197). New York: Cambridge University Press.

Cattarello, A. M. (2000). Community-level influences on individuals' social bonds, peer
associations, and delinquency: a multilevel analysis. Justice Quarterly, 17(1), 33-
60.

Chase-Lansdale, P. L., Gordon, R. A., Brooks-Gunn, J ., & Klebanov, P. K. (1997).
Neighborhood and family influences on the intellectual and behavioral
competence of preschool and early school-age children. In J. Brooks-Gunn & G.
J. Duncan & J. L. Aber (Eds.), Neighborhood poverty (Vol. II, pp. 79-118). New
York: Russell Sage Foundation.

Cicchetti, D., Toth, S., & Bush, M. (1988). Developmental psychopathology and
incompetence in childhood: suggestions for intervention. In B. B. Lahey & A. E.
Kazdin (Eds.), Advances in clinical child psychology (V 01. 11, pp. 1-71). New
York and London: Plenum Press.

Cloward, R., & Ohlin, L. E. (1960). Delinquency and opportunity: a theory of delinquent
gangs. New York: Free Press.

Cohen, A. K. (1955). Delinquent boys: The culture of the gang. Glencoe, 11.: The Free
Press.

Cole, G. F., & Smith, C. E. (2001). The American system of criminal justice (Ninth ed.).
Belmont, CA: Wadsworth.

Conger, R. D., & Rueter, M. (1996). Siblings, parents, and peers: a longitudinal study of
social influences in adolescent risk for alcohol use and abuse. In G. H. Brody
(Ed.), Sibling relationships: their causes and consequences (V 01. 10, pp. 1-30).
Stamford, CT: Ablex Publishing Corp.

Cook, P. J ., & Laub, J. H. (1998). The unprecedented epidemic in youth violence. In M.
Tonry & M. H. Moore (Eds.), Youth Violence (Vol. 24, pp. 27-64). Chicago: The
University of Chicago Press.

Cooley, C. H. (1912). Human nature and social order. New York: Scribner's.

Costello, B. J. (2000). Techniques of neutralization and self-esteem: a critical test of
social control and neutralization theory. Deviant Behavior, 21(4), 307-329.

Crutchfield, R. D., & Pitchford, S. R. (1997). Work and crime: the effects of labor
stratification. Social Forces, 76(1), 93-118.

Danegger, A. E., Cohen, C. B, Hayes, C. D., & Holden, G. A. (1999). Juvenile
accountability incentive block grants: strategic planning guide (N CJ 172846).

231

Washington, DC: United States Department of Justice, Office of Juvenile Justice
and Delinquency Prevention.

Davidson, W. S., Redner, R., Amdur, R. L., & Mitchell, C. M. (1990). Alternative
treatments for troubled youth: the case of diversion from the justice system. New
York: Plenum Press.

Demaris, A. (1992). Logit modeling (Vol. 86). Newbury Park: Sage Publications.

Dembo, R., Dertke, M., Schmeidler, J ., & Washbum, M. (1987). Prevalence, correlates
and consequences of alcohol and other drug use among youths in a juvenile
detention center. Journal of Prison and Jail Health, 6(2), 97-127.

Dembo, R., Schmeidler, J ., Nini-Gough, B., & Manning, D. (1998). Sociodemographic,
delinquency-abuse history, and psychosocial functioning differences among
juvenile offenders of various ages. Journal of Child and Adolescent Substance
Abuse, 8(2), 63-78.

Dembo, R., Washbum, M., Wish, E. D., Schmeidler, J ., Getreu, A., Berry, E., Williams,
L., & Blount, W. R. (1987). Further examination of the association between heavy
marijuana use and crime among youths entering a juvenile detention center.
Journal of Psychoactive Drugs, 19(4), 361-373.

Dembo, R., Williams, L., Fagan, J ., & Schmeidler, J. (1993). The relationships of
substance abuse and other delinquency over time in a sample of juvenile
detainees. Criminal Behaviour and Mental Health, 3, 158-179.

Denno, D. W. (1990). Biology and violence: from birth to adulthood. Cambridge, UK:
Cambridge University Press.

Eddy, J. M., & Gribskov, L. S. (1998). Juvenile justice and delinquency prevention in the
United States: the influence of theories and traditions on policies and practices. In
T. P. Gullotta & G. R. Adams & R. Montemayor (Eds.), Delinquent violent youth:
theory and interventions (pp. 12-52). Thousand Oaks: Sage Publications.

Ellickson, P., & McGuigan, K. A. (2000). Early predictors of adolescent violence.
American Journal of Public Health, 90(4), 566-572.

Elliot, D. S. (1994). Serious violent offenders: onset, developmental course, and
tennination--The American Society of Criminology 1993 presidential address.
Criminology, 32(1), 1-21.

Elliot, D. S., & Ageton, S. S. (1980). Reconciling race and class differences in self-
reported and official estimates of delinquency. American Sociological Review, 45,
95-1 10.

Elliot, D. S., & Huizinga, D. (1984). The relationship between delinquent behavior and
ADM problems (28). Bolder, CO: Behavioral Research Institute.

232

Elliot, D. S., Huizinga, D., & Ageton, S. S. (1985). Explaining delinquency and drug use.
Newbury Park, CA: Sage Publications.

Elliot, D. S., Huizinga, D., & Menard, S. (1989). Multiple problem youth: delinquency,
substance abuse, and mental health problems. New York: Springer-Verlag.

Elliot, D. S., & Menard, S. (1996). Delinquent friends and delinquent behavior: temporal
and developmental patterns. In J. D. Hawkins (Ed.), Delinquency and Crime (pp.
28-67). New York: Cambridge University Press.

Elliott, D. S. (1994). Serious violent offenders: onset, developmental course, and
termination--The American Society of Criminology 1993 presidential address.
Criminology, 32(1), 1-21.

Elliott, D. S., & Menard, S. (1996). Delinquent friends and delinquent behavior: temporal
and developmental patterns. In J. D. Hawkins (Ed.), Delinquency and Crime (pp.
28-67). New York: Cambridge University Press.

Erickson, M. L., & Jensen, G. F. (1977). Delinquency is still group behavior!: toward
revitalizing the group. Journal of Criminal Law and Criminology, 68(2), 262-273.

Fagan, J ., & Pabon, E. (1990). Contributions of delinquency and substance use to school
dropout among inner-city youths. Youth and Society, 21(3), 306-354.

Fagan, J ., & Wilkinson, D. L. (1998). Guns, youth violence, and social identify. In M. T.
a. M. H. Moore (Ed.), Youth violence (Vol. 24, pp. 105-188). Chicago: University
of Chicago Press.

Farrington, D. P. (1985). Predicting self-reported and official delinquency. In D. P.
Farrington & R. Tarling (Eds.), Prediction in criminology (pp. 150—173). Albany,
NY: State University of New York Press.

Farrington, D. P. (1989). Early predictors of adolescent aggression and adult violence.
Violence and Victims, 4, 79-100.

Farrington, D. P. (1992). Explaining the beginning, progress and ending of antisocial
behavior from birth to adulthood. In J. McCord (Ed.), Facts, framework, and
forecasts (V 01. 231-252). New Brunswick, NJ: Transaction Publishers.

Farrington, D. P. (1993). Understanding and preventing bullying. In M. Tonry (Ed.),
Crime and justice: a review of research (V 01. 17, pp. 381-458). Chicago:
University of Chicago.

Farrington, D. P. (1995). The development of offending and antisocial behaviour from
childhood: key findings from the Cambridge Study in Delinquent Development.
Journal of Child Psychology and Psychiatry, 36(3), 929-964.

233

 

F arrington, D. P. (1996). The explanation and prevention of youthful offending. In J. D.
Hawkins (Ed.), Delinquency and Crime (pp. 68-148). New York: Cambridge

University Press.

F arrington, D. P. (1998). Predictors, causes, and correlates of male youth violence. In M.
Tonry & M. H. Moore (Eds.), Youth violence (V 01. 24, pp. 421—476). Chicago:
The University of Chicago Press.

F arrington, D. P., & Tarling, R. (1985). Criminological prediction: an introduction. In D.
P. Farrington & R. Tarling (Eds.), Prediction in criminology (pp. 2-33). Albany,
NY: State University of New York Press.

Federal Bureau of Investigations. (1954). Crime in the United States. Washington, DC:
US. Government Printing Office.

Federal Bureau of Investigations. (1961). Crime in the United States. Washington, DC:
US. Government Printing Office.

Felson, M., & Gottfredson, M. R. (1984). Social indicators of adolescent activities near
peers and parents. Journal of Marriage and the Family, 46, 709-714.

Fox, J. (1991). Regression diagnostics (V 01. 79). Thousand Oaks: Sage Publications.

Furstenberg, F. F., Cook, T. D., Eccles, J ., Elder Jr., G. H., & Sameroff, A. (1999).
Managing to make it: urban families and adolescent success. Chicago: The
University of Chicago Press.

Furstenberg, F. F., & Hughes, M. E. (1997). The influence of neighborhoods on
children's development: a theoretical perspective and research agenda. In J.
Brooks-Gunn & G. J. Duncan & J. L. Aber (Eds.), Neighborhood poverty (V 01. II,
pp. 23-47). New York: Russell Sage Foundation.

Furstenberg Jr., F. F., Cook, T. D., Eccles, J ., Elder Jr., G. H., & Sameroff, A. (1999).
Managing to make it: urban families and adolescent success. Chicago: The
University of Chicago Press.

Furstenberg Jr., F. F., & Hughes, M. E. (1997). The influence of neighborhoods on
children's development: a theoretical perspective and research agenda. In J. L.
Aber (Ed.), Neighborhood poverty (V 01. II, pp. 23-47). New York: Russell Sage
Foundation.

Glueck, S. (1959). Problem of delinquency. Boston: Houghton Mifflin Company.

Glueck, S., & Glueck, E. T. (1934). One thousand juvenile delinquents (V 01. 1).
Cambridge: Harvard University Press.

Glueck, S., & Glueck, E. T. (1950). Unraveling juvenile delinquency. New York:
Commonwealth Fund.

234

Glueck, S., & Guleck, E. (1930). 500 criminal careers. New York: Alfred A. Knopf.

Glueck, S., & Guleck, E. (1970). Toward a typology of juvenile offenders. New York:
Grune 8c Stratton, Inc.

Goldstein, P. J. (1995). The drugs/violence nexus: a tripariate conceptual framework. In
J. A. Inciardi & K. McElrath (Eds.), The American drug scene (pp. 255-264). Los
Angeles: Roxbury Publishing Company.

Goring, C. (1913). The English convict. Montclair, NJ: Patterson Smith.

Gorman—Smith, D., Tolan, P. H., Loeber, R., & Henry, D. B. (1998). Relation of family
problems to patterns of delinquent involvement among urban youth. Journal of
Abnormal Child Psychology, 26(5), 319-333.

Gottfredson, D. C. (2001). Schools and delinquency. New York: Cambridge University
Press.

Gottfredson, D. C., McNeil III, R. J ., & Gottfredson, G. D. (1991). Social area influences
on delinquency: A multilevel analysis. Journal of Research in Crime and
Delinquency, 28(2), 197-226.

Gottfredson, M. R., & Hirschi, T. (1986). The true value of lambda would appear to be
zero: an essay on career criminals, criminal careers, selective incapacitation,
cohort studies, and related topics. Criminology, 24(2), 213-234.

Gottfredson, M. R., & Hirschi, T. (1990). A general theory of crime. Stanford, CA:
Stanford University Press.

Greenberg, D. F. (1985). Age, crime, and social explanation. American Journal of
Sociology, 91(1), 1-21.

Grobsmith, E. (1989). The relationship between substance abuse and crime among Native
American inmates in the Nebraska Department of Corrections. Human
Organization, 48, (4), pp. 285 297., 48(4), 285-297.

Habermas, J. (1988). Legitimation crisis (T. McCarthy, Trans). Boston: Beacon Press.

Halpem-Felsher, B. L., Connell, J. P., Spencer, M. B., Aber, J. L., Duncan, G. J .,
Clifford, E., Crichlow, W. E., Usinger, P. A., Cole, S. P., Allen, L., & Seidman,
E. (1997). Neighborhood and family factors predicting educational risk and
attainment in African American and White children and adolescents. In J. Brooks-
Gunn & G. J. Duncan & J. L. Aber (Eds.), Neighborhood poverty (V 01. II, pp.
146-173). New York: Russell Sage Foundation.

Hawkins, J. D., Catalano, R. F., & Miller, J. Y. (1992). Risk and protective factors for
alcohol and other drug problems in adolescence and early adulthood: Implication
for substance abuse prevention. Psychological Bulletin, [12(1), 64-105.

235

Hawkins, J. D., Herrenkohl, T., Farrington, D. P., Brewer, D., Catalano, R. F ., & Harachi,
T. W. (1998). A review of predictors of youth violence. In R. Loeber & D. P.
Farrington (Eds.), Serious and Violent Juvenile Offenders (pp. 106-146).
Thousand Oaks: Sage Publications.

Healy, W. (1915). The individual delinquent: a text-book of diagnosis and prognosis for
all concerned in understanding oflenders. Boston: Little, Brown.

Heilbrun, K., Brock, W., Waite, D., Lanier, A., Schmid, M., Witte, G., Keeney, m.,
Westendorf, M., Buinavert, L., & Shumate, M. (2000). Risk factors for juvenile
criminal recidivism: the postrelease community adjustment of juvenile offenders.
Criminal Justice and Behavior, 27(3), 275-291.

Hill, K. G., Howell, J. C., Hawkins, J. D., & Battin-Pearson, S. R. (1999). Childhood risk
factors for adolescent gang membership: results from the Seattle Social
Development Project. Journal of Research in Crime and Delinquency, 36(3), 300-
322.

Hindelang, M. J. (1974). The Uniform Crime Reports revisited. Journal of Criminal
Justice, 1, 1-18.

Hindelang, M. J ., Hirschi, T., & Weis, J. G. (1981). Measuring delinquency (V 01. 123).
Beverly Hills: Sage Publications.

Hindelang, M. J ., Hirschi, T., & Weis, J. G. (1991). Correlates of delinquency: The
illusion of discrepancy between self-reported and official measures. In W. E.
Thompson & J. E. Bynum (Eds.), Juvenile delinquency: classic and contemporary
readings (pp. 86-110). Boston: Allyn and Bacon.

Hirschi, T. (1969). Causes of delinquency. Berkeley: University of California Press.

Hoffman, J. P., & Cerbone, F. G. (1999). Stressful life events and delinquency escalation
in early adolescence. Criminology, 37(2), 343-373.

Houston, J. (1996). What works: the search for excellence in gang intervention programs.
Journal of Gang Research, 3(3), 1-16.

Howell, J. C. (1998). Promising programs for youth gang violence prevention and
intervention. In R. Loeber & D. P. F arrington (Eds.), Serious and violent juvenile
oflenders. Thousand Oaks: Sage Publications.

Howell, J. C. (2001). Juvenile justice programs and strategies. In R. Loeber & D. P.
Farrington (Eds.), Child delinquents (pp. 305-321). Thousand Oaks: Sage
Publications, Inc.

Howell, J. C. (Ed). (1995). Guide for implementing the comprehensive strategy for

serious, violent, and chronic juvenile oflenders. Washington, DC: Office of
Juvenile Justice and Delinquency Prevention.

236

Huizinga, D., & J akob-Chien, C. (1998). The contemporaneous co-occurrence of serious
and violent juvenile offending and other problem behaviors. In R. Loeber & D. P.
Farrington (Eds.), Serious and Violent Juvenile Offenders (pp. 47-67). Thousand
Oaks: Sage Publications.

Huizinga, D., Loeber, R., & Thomberry, T. P. (1994). Urban delinquency and substance
abuse (NCJ 143454). Washington, DC: OJJDP.

Huizinga, D., Loeber, R., Thomberry, T. P., & Cothem, L. (2000). Co-occurrence of
delinquency and other problem behaviors (NCJ182211). Washington, DC: Office
of Justice Programs, Office of Juvenile Justice and Delinquency Prevention.

J ang, S. J. (1998). Self-esteem, delinquent peers, and delinquency: a test of the self-
enhancement thesis. American Sociological Review, 63(4), 586-598.

Jensen, G. F., and Dean G. Rojek. (1992). Delinquency and youth crime (2nd ed).
Prospect Heights: Waveland Press, Inc.

Johnson, B. D., Andres Golub, and Jeffrey F agan. (1995). Careers in crack, drug use,
drug distribution, and nondrug criminality. Crime and Delinquency, 41(3), 275-
295.

Kandall, D. B. (1982). Epidemiological and psychosocial perspectives on adolescent drug
use. Journal of Child Psychiatry, 21 , 328-347.

Kaplan, H. B., & Damphousse, K. R. (1997). Negative social sanctions, self-derogation,
and deviant behavior: main and interaction effects in longitudinal perspective.
Deviant Behavior, 18, 1-26.

Kaufman, P., Chen, Z., Choy, S. P., Ruddy, S. A., Miller, A. K., Fleury, J. K., Chandler,
K. A., Rand, M. R., Klaus, P., & Planty, M. G. (2000). Indicators of school crime
and safety, 2000 (NCES 2001-017/NCJ-184176). Washington, DC.: US.
Departments of Education and Justice.

Keenan, K., Shaw, D., Delliquadri, E., Giovannelli, J ., & Walsh, B. (1998). Evidence for
the continuity of early problem behaviors: application of a developmental model.
Journal of Abnormal Child Psychology, 26(6), 441-454.

Kelley, B. T., Huizinga, D., Thomberry, T. P., & Loeber, R. (1997). Epidemiology of
serious violence. Washington, DC: Office of Justice Program, Office of Juvenile
Justice and Delinquency Prevention.

Kelly, D. (1982). Creating school failure, youth crime, and deviance. Los Angeles:
Trident Shop.

Keppel, G. (1991). Design and analysis: a researcher's handbook (Third ed). Upper
Saddle River, NJ: Prentice Hall.

237

Komhauser, R. (1978). Social sources of delinquency. Chicago: University of Chicago
Press.

Krisberg, B., and James Austin. (1978). The children of Ishmael. Palo Alto: Mayfield
Publishing Company.

Kuhn, T. S. (1970). The structure of scientific revolutions (Vol. 11). Chicago: The
University of Chicago Press.

Le Blane, M., & Loeber, R. (1998). Developmental criminology updated. In M. Tonry
(Ed.), Crime and Justice: a review of research (V 01. 23, pp. 115-198). Chicago:
University of Chicago Press.

Leadbeater, B. J ., Blatt, S. J ., & Quinlan, D. M. (1995). Gender-linked vulnerabilities to
depressive symptoms, stress, and problem behaviors in adolescents. Journal of
Research on Adolescence, 5, 1-29.

Leger, R. G. (1981). Labeling and its consequences in a closed social system. British
Journal of Criminology, 21(2), 109-122.

Lewis, D. O. (1989). Toward a theory of the genesis of violence: a follow-up study of
delinquents. Journal of the American Academy of Child and Adolescent
Psychiatry, 28(3), 431-436.

Lewis-Beck, M. S. (1980). Applied regression: an introduction (V 01. 22). Thousand
Oaks: Sage Publications.

Lipsey, M. (1999). Can intervention rehabilitate serious delinquents? Annals of the
American Academy of Political and Social Science, 564, 142-166.

Lipsey, M. W. (1998). Design sensitivity: statistical power for applied experimental
research. In L. Bickrnan & D. J. Rog (Eds.), Handbook of applied social research
methods (pp. 39-68). Thousand Oaks: Sage Publications.

Lipsey, M. W. (2000, July 16-19). What 5 00 intervention studies show about the eflects
of intervention on the recidivism of juvenile ofi'enders. Paper presented at the
Annual Conference on Criminal Justice Research and Evaluation, Washington,
DC.

Lipsey, M. W., & Derzon, J. H. (1998). Predictors of violent or serious delinquency in
adolescence and early childhood. In R. Loeber & D. P. Farrington (Eds.), Serious
and Violent Juvenile Offenders (pp. 86-105). Thousand Oaks: Sage Publications.

Lipsey, M. W., & Wilson, D. B. (1998). Effective intervention for serious juvenile

offenders. In R. Loeber & D. P. F arrington (Eds.), Serious and Violent Juvenile
Oflenders (pp. 313-345). Thousand Oaks: Sage Publications.

238

Lipton, D., Martinson, R., & Wilks, J. (1975). The effectiveness of correctional
treatment: a survey of treatment evaluation studies. New York: Praeger
Publishers.

Liska, A. E. (1987). Perspectives on deviance (2nd ed). Englewood Cliffs: Prentice-Hall,
Inc.

Loeber, R. (1988). Natural histories of conduct problems, delinquency, and associated
substance abuse: evidence for developmental perspectives. In B. B. Lahey & A.
E. Kazdin (Eds.), Advances in clinical child psychology (Vol. 11, pp. 73-124).
New York and London: Plenum Press.

Loeber, R. (1990). Development and risk factors of juvenile antisocial behavior and
delinquency. Clinical Psychology Review, 10, 1-41.

Loeber, R. (1996). Developmental continuity, change, and pathways in male juvenile
problem behaviors and delinquency. In J. D. Hawkins (Ed.), Delinquency and
Crime (pp. 1-27). New York: Cambridge University Press.

Loeber, R., & Dishion, T. (1983). Early predictors of male delinquency: a review.
Psychological Bulletin, 93, 68-99.

Loeber, R., F arrington, D. P., & Washbush, D. A. (1998). Serious and violent juvenile
offenders. In R. Loeber & D. P. F anington (Eds.), Serious and Violent Juvenile
Oflenders (pp. 13-29). Thousand Oaks: Sage Publications.

Loeber, R., & Hay, D. (1994). Developmental approaches to aggression and conduct
problems. In M. Rutter & D. F. Hay (Eds.), Development through life .° a
handbook for clinicians (pp. 488-515). Oxford: Blackwell Scientific.

Loeber, R., & Hay, D. (1997). Key issues in the development of aggression and violence
from childhood to early adulthood. Annual Review of Psychology. 48, 371-410.

Loeber, R., & Snyder, H. N. (1990). Rate of offending in juvenile careers: findings in
constancy and change in lambda. Criminology, 28(1), 97-109.

Loeber, R., & Stouthamer-Loeber, M. (1996). The development of offending. Criminal
Justice and Behavior, 23(1), 12-24.

Loeber, R., Van Kammen, W. B., Stouthamer-Loeber, M., & Farrington, D. P. (1996,
November 20-23). Synchrony and sequence in the joint development of
delinquency and substance use. Paper presented at the American Society of
Criminology, Chicago, IL.

Loeber, R., Wung, P., Keenan, K., Giroux, B., Stouthamer-Loeber, M., Van Kammen, W.
B., & Maughan, B. (1993). Developmental pathways in disruptive child behavior.
Development and Psychopathology, 5(1-2), 103-133.

239

Lundman, R. J. (2001). Prevention and control of juvenile delinquency. New York:
Oxford University Press.

Maguin, E., Hawkins, J. D., Catalano, R. F., Hill, K., Abbot, R., & Herrenkohl, T. (1995).
Risk factors measured at three ages for violence at age 1 7-18. Paper presented at
the American Society of Criminology, Boston, MA.

Maltz, M. D., Gordon, A. S., McDowall, D., & McCleary, R. (1980). An artifact in
pretest-posttest designs: how it can mistakenly make delinquency programs look
effective. Evaluation Review, 4(2), 225-240.

Maxwell, S. R., & Maxwell, C. D. (2000). Examining the "criminal careers" of
prostitutes within the nexus of drug use, drug selling, and other illicit activities.
Criminology, 38(3), 787-809.

Mazerolle, P., Brame, R., Paternoster, R., Piquero, A., & Dean, C. (2000). Onset age,
persistence, and offending versatility: comparisons across gender. Criminology,
38(4), 1143-1172.

McCarthy, B., & Hagan, J. (2001). Mean streets: situational delinquency among
homeless youth. In R. C. Smandych (Ed.), Youth crime: varieties, theories, and
prevention (pp. 142-159). Toronto: Harcourt Canada.

McCluskey, C. P. (2002). Understanding Latino delinquency: the applicability of strain
theory by ethnicity. New York: LF B Scholarly Publishing.

McGarrell, E. F. (1988). Juvenile correctional reform: two decades of police and
procedural change. Albany: State University of New York Press.

McGarrell, E. F. (2001). Restorative justice conferences as an early response to young
oflenders (N CJ 1 87769). Washington, DC: United States Department of Justice,
Office of Juvenile Justice and Delinquency Prevention.

Mears, D. P., Ploeger, M., & Warr, M. (1998). Explaining the gender gap in delinquency:
peer influences and moral evaluations of behavior. Journal of Research in Crime
and Delinquency, 35(3), 251-266.

Melby, J. N., Conger, R. D., Conger, K. J ., & Lorenz, F. O. (1993). Effects of parental
behavior on tobacco use by young male adolescents. Journal of Marriage and the
Family, 55(2), 439-454.

Menard, S. (1995). Applied logistic regression analysis (Vol. 106). Thousand Oaks, CA:
Sage Publications.

Mennel, R. (1973). Thorns and thistles: juvenile delinquents in the United States, 1825-
1940. Hanover, NH: University Press of New England.

240

Miller, J. G. (1991). Last one over the wall: the Massachusetts experiment in closing
reform schools. Columbus: Ohio State University Press.

Moffitt, T. B. (1994). Natural histories of delinquency. In E. Weitekamp & H. J. Kemer
(Eds.), Cross-national research on human development and criminal behaviour.
Netherlands: Kluwer Academic Publishers.

Moffitt, T. E. (1997). Adolescence-limited and life-course-persistent offending: a
complementary pair of developmental theories. In T. P. Thomberry (Ed.),
Developmental theories of crime and delinquency. New Brunswick, New Jersey:
Transaction Publishers.

Nagin, D. S., & Farrington, D. P. (1992). The onset and persistence of offending.
Criminology, 30(4), 501-523.

National Research Council. (2001). Understanding dropouts: statistics, strategies, and
high-stakes testing. Washington, DC: National Academy Press.

‘

National Research Council and Institute of Medicine. (2001). Juvenile crime, juvenile
justice. Washington, DC: National Academy Press.

Nettler, G. (1984). Explaining crime (Third ed). New York: McGraw-Hill.

Office of Juvenile Justice and Delinquency Prevention. (2001). The 8 percent solution
(NCJ 192272). Washington, DC: United States Department of Justice.

Olson-Raymer, G. (1984). National juvenile justice policy: myth or reality? In S. H.
Decker (Ed.), Juvenile justice policy: analyzing trends and outcomes (Vol. 7, pp.
19-58). Thousand Oaks: Sage Publications.

Olweus, D. (1993). Bullying at school. Oxford, UK: Blackwell.

Osgood, D. W., O'Malley, P. M., Bachman, J. G., & Johnston, L. D. (1989). Time trends
and age trends in arrests and self-reported illegal behavior. Criminology, 27(3),
389-414.

Paternoster, R., Brame, R., Mazerolle, P., & Piquero, A. (1998). Using the correct
statistical test for the equality of regression coefficients. Criminology, 36(4), 859-
866.

Petersilia, J ., & Turner, S. (1990). Comparing intensive and regular supervision for high-
risk probationers: early results from an experiment in California. Crime and
Delinquency, 36(1), 87-111.

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

241

Platt, A. (1991). The child-saving movement and the origins of the juvenile justice
system. In R. J. Berger (Ed.), The sociology of juvenile delinquency (pp. 9-26).
Chicago: Nelson-Hall.

Platt, A. M. (1977). The child savers : the invention of delinquency. Chicago: University
of Chicago Press.

Reiss, A. J. (1986). Co-offender influences on criminal careers, Criminal careers and
"career criminals " (Vol. 2, pp. 121-160). Washington, DC: National Academy
Press.

Rosenberg, M., Schooler, C., & Schoenbach, C. (1989). Global self-esteem and specific
self-esteem: different concepts, different outcomes. American Sociological
Review, 60(1), 141-156.

Rotrnan, E. (1986). Do criminal offenders have a constitutional right to rehabilitation?
Journal of Criminal Law and Criminology, 77(4), 1023-1068.

Rowe, D. C., & Farrington, D. P. (1997). The familial transmission of criminal
convictions. Criminology, 35(1), 177-201.

Sampson, R. J ., & Groves, W. B. (1989). Community structure and crime: testing social-
disorganization theory. American Journal of Sociology, 94(4), 774—802.

Sampson, R. J ., & Laub, J. H. (1997). A life-course theory of cumulative disadvantage
and the stability of delinquency. In T. P. Thomberry (Ed.), Developmental
theories of crime and delinquency (pp. 133-161). New Brunswick, New Jersey:
Transaction Publishers.

Schlossman, S., & Sedlak, M. (1983). The Chicago Area Project revisited. Crime and
Delinquency, 29(3), 398-462.

Scholte, E. M. (1992). Identification of children at risk at the police station and the
prevention of delinquency. Psychiatry, 55, 354-369.

Sellin, T., 8c Wolfgang, M. E. (1969). Measuring delinquency. In T. Sellin & M. E.
Wolfgang (Eds.), Delinquency: selected studies (pp. 1-10). New York: John
Wiley and Sons, Inc.

Sharp, 8., & Thompson, D. (1994). The role of whole-school policies in tacking bullying
behavior in schools. In P. K. Smith & 8. Sharp (Eds.), School bullying: insights
and perspectives (pp. 57-83). London: Routledge.

Shavit, Y., & Rattner, A. (1988). Age, crime, and the early life course. American Journal
Of Sociology, 93(6), 1457-1470.

Shaw, C., & McKay, H. (1931). The natural history of a delinquent career. Chicago:
University of Chicago Press.

242

Shaw, C., & McKay, H. (1942). Juvenile delinquency and urban areas. Chicago:
University of Chicago Press.

Shaw, C., & McKay, H. (1969). Juvenile delinquency and urban areas (Revised ed).
Chicago: University of Chicago Press.

Short, J. F. J ., & Nye, I. F. (1957). Reported behavior as a criterion of deviant behavior.
Social Problems, 5, 207-213.

Simons, R. L., Wu, C.-I., Conger, R. D., & Lorenz, F. (1994). Two routes to delinquency:
differences between early and late starters in the impact of parenting and deviant
peers. Criminology, 32(2), 247-275.

Smith, P. K., & Sharp, S. (1994). School bullying: insights and perspectives. London:
Routledge.

Snyder, H., & Sickmund, M. (1999). Juvenile oflenders and victims: 1999 national
report. Washington, DC: Office of Juvenile Justice and Delinquency Prevention.

Snyder, H. N. (1997). Juvenile arrests 1996 (NCJ 167578). Washington, DC: Office of
Juvenile Justice and Delinquency Prevention.

Spergel, I. (1964). Racketville, Slumtown, Haulburg. Chicago: The University of Chicago
Press.

Spergel, I. A., & Curry, G. D. (1990). The youth gang problem. In C. R. Huff (Ed.),
Gangs in America (lst ed., pp. 288-309). Newbury Park: Sage Publications.

Stanton, C. A., & Meyer, A. L. (1998). A comprehensive review of community-based
approaches for the treatment of juvenile offenders. In T. P. Goullotta & G. R.
Adams & R. Montemayor (Eds.), Delinquent violent youth (pp. 205-229).
Thousand Oaks: Sage Publications.

Steffensmeier, D. J ., Allan, E. A., Harer, M. D., & Streifel, C. (1989). Age and the
distribution of crime. American Journal of Sociology, 94(4), 803-831.

Stice, E., Myers, M. G., & Brown, S. A. (1998). Relations of delinquency to adolescent
substance use and problem use: A prospective study. Psychology of Addictive
Behaviors, 12(1), 136-146.

Strom, R. D., Bernard, H. W., & Strom, S. K. (1987). Human development and learning.
New York: Human Sciences Press, Inc.

Sullivan, M. L. (1989). Getting paid. Ithaca, NY: Cornell University Press.

Sutherland, E. H., & Cressey, D. R. (1978). Principles of criminology. Philadelphia, PA:
Lippincott.

243

Sykes, G. M., & Matza, D. (1957). Techniques of neutralization: a theory of delinquency.
American Sociological Review, 22, 664-670.

Tannenbaum, F. (1938). Crime and the community. Boston, MA: Ginn.

Teplin, L. A. (2001). Assessing alcohol, drug, and mental disorders in juvenile detainees
(Fact Sheet FS-200102). Washington, DC: United States Department of Justice,
Office of Juvenile Justice and Delinquency Prevention.

Thomas, C. W., & Bishop, D. M. (1984). The effect of formal and informal sanctions on
delinquency: a longitudinal comparison of labeling and deterrence theories. The
Journal of Criminal Law and Criminolog, 75(4), 1222-1245.

Thomberry, T. P. (1987). Toward an interactional theory of delinquency. Criminology,
25(4), 863-891.

Thomberry, T. P. (1996). Empirical support for interactional theory: a review of the
literature. In J. D. Hawkins (Ed.), Delinquency and Crime (pp. 198-235). New
York: Cambridge University Press.

Thomberry, T. P. (1997a). Introduction: some advantages of developmental and life-
course perspectives for the study of crime and delinquency. In T. P. Thomberry
(Ed.), Developmental theories of crime and delinquency (V 01. 7, pp. 1-10). New
Brunswick: Transaction Publishers.

Thomberry, T. P. (Ed). (1997b). Developmental theories of crime and delinquency.

Thomberry, T. P., Huizinga, D., & Loeber, R. (1995). The prevention of serious
delinquency and violence: implications from the program of research on the
causes and correlates of delinquency. In J. C. Howell & B. Krisberg & J. D.
Hawkins & J. J. Wilson (Eds.), Serious, violence, and chronic juvenile oflenders
(pp. 213-237). Thousand Oaks: Sage Publications.

Thomberry, T. P., Lizotte, A. J ., Krohn, M. D., Famworth,.M., & Jang, S. J. (1994).
Delinquent peers, beliefs, and delinquent behavior: a longitudinal test of
interactional theory. Criminology, 32(1), 47-79.

Thomberry, T. R, Moore, M., & Christenson, R. L. (1991). The effects of dropping out
of high school on subsequent criminal behavior. In R. J. Berger (Ed.), The
sociology of juvenile delinquency (pp. 207-219). Chicago: Nelson-Hall.

Thrasher, F. M. (1936). The gang (2nd ed). Chicago: University of Chicago Press.

Tittle, C. R., Villemez, W. J ., & Smith, D. A. (1978). The myth of social class and
criminality: an empirical assessment of the empirical evidence. American
Sociological Review, 43(5), 643-656.

244

Tolan, P. H., & Gorrnan-Smith, D. (1998). Development of serious and violent offending
careers. In R. Loeber & D. P. Farrington (Eds.), Serious and Violent Juvenile
Oflenders (pp. 68-85). Thousand Oaks: Sage Publications.

Tremblay, R. B., & LeMarquand, D. (2001). Individual risk and protective factors. In R.
Loeber & D. P. F arrington (Eds.), Child delinquents (pp. 137-164). Thousand
Oaks: Sage Publications.

US. Department of Justice: Bureau of Justice Statistics. (1999). Sourcebook of criminal
justice statistics. Washington, DC: Government Printing Office.

Veysey, B. M., & Messner, S. F. (1999). Further testing of social disorganization: An
elaboration of Sampson and Groves's 'community structure and crime'. Journal of
Research in Crime and Delinquency, 36(2), 156-174.

Voelkl, K. E., Welte, J. W., & Wieczorek, W. F. (1999). Schooling and delinquency
among white and African American adolescents. Urban Education, 34(1), 69-88.

Vold, G. B., Bernard, T. J ., & Snipes, J. B. (1998). Theoretical criminology (Fourth ed).
New York: Oxford University Press.

Vuchinich, S., Bank, L., & Patterson, G. R. (1992). Parenting, peers, and the stability of
antisocial behavior in preadolescent boys. Developmental Psychology, 28(3), 510-
521.

Warr, M. (2002). Companions in crime: the social aspects of criminal conduct. Chicago:
Cambridge University Press.

Weis, J. G. (1988). Issues in the measurement of criminal careers. In A. Blumstein (Ed.),
Criminal careers and "career criminals " (V 01. H, pp. 1-51). Washington, DC.

Wells, L. B., & Rankin, J. H. (1988). Direct parental controls and delinquency.
Criminology, 26(2), 263-286.

Welsh, W. N., Greene, J. R., & Jenkins, P. H. (1999). School disorder: the influence of
individual, institutional, and community factors. Criminology, 37(1), 73-115.

Welsh, W. N., Stokes, R., & Greene, J. R. (2000). A macro-level model of school
disorder. Journal of Research in Crime and Delinquency, 3 7(3), 243-283.

White, H. R., Loeber, R., Stouthamer—Loeber, M., & F arrington, D. P. (1999).
Developmental associations between substance use and violence. Development
and Psychopathology, 11(4), 785-803.

Wikstrdm, P.-O. H., & Loeber, R. (2000). Do disadvantaged neighborhoods cause well-
adjusted children to become adolescent delinquents? a study of male juvenile
serious offending, individual risk and protective factors, and neighborhood
context. Criminology, 38(4), 1109-1142.

245

Wilson, J. J ., & Howell, J. C. (1993). Comprehensive strategy for serious, violent, and
chronic juvenile offenders. Washington, DC: United States Department of Justice,
Office of Juvenile Justice and Delinquency Prevention.

Wilson, J. J ., & Howell, J. C. (1995). Comprehensive strategy for serious, violent, and
chronic juvenile offenders. In J. C. Howell & B. Krisberg & J. D. Hawkins & J. J.
Wilson (Eds.), A sourcebook: serious, violent, and chronic juvenile offenders.
Thousand Oaks, CA: Sage Publications.

Wilson, J. Q. (1968). Varieties of police behavior. Cambridge, Massachusetts: Harvard
University Press.

Wilson, J. Q. (1983). Thinking about crime (Revised ed). New York: Vintage Books.

Wilson, J. Q., & Hermstein, R. J. (1985). Crime and human nature. New York: Simon
and Schuster.

Winner, L., Lanza-Kaduce, L., Bishop, D. M., & Frazier, C. E. (1997). The transfer of
juvenile to criminal court: Reexamining recidivism over the long term. Crime and
Delinquency, 43(4), 548-563.

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

Wolfgang, M. E., Thomberry, T. P., & Figlio, R. M. (1987). From boy to man, fi'om
delinquency to crime. Chicago and London: University of Chicago Press.

Woodson, R. L. (1981). A summons to life: mediating structures and the prevention of
youth crime. Cambridge: Ballinger.

Yule, W. (1981). The epidemiology of child psychopathology. In B. B. Lahey & A. E.
Kazdin (Eds.), Advances in clinical child psychology (V 01. 4, pp. 1-51). New
York and London: Plenum Press.

Zimring, F. E. (1981). Kids, groups and crime: some implications of a well-known secret.
Journal of Criminal Law and Criminology, 72(3), 867-885.

Zimring, F. E. (1998). Toward a jurisprudence of youth violence. In M. Tonry & M. H.
Moore (Eds.), Youth Violence (Vol. 24, pp. 477-501). Chicago: The University of
Chicago Press.

246

MICHIGAN TTTTTTTT

lllllllllwwg